CN116862036A

CN116862036A - Load prediction method and device

Info

Publication number: CN116862036A
Application number: CN202210300910.6A
Authority: CN
Inventors: 徐灏; 龚兰平; 邱辉; 张泽宇
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2022-03-25
Filing date: 2022-03-25
Publication date: 2023-10-10

Abstract

The application discloses a load prediction method and a load prediction device, which are used for predicting the load of a station based on less sampling data, and obtaining more accurate predicted load through less calculated amount under the condition that the battery load can not be acquired in real time. The method comprises the following steps: acquiring first sampling data, wherein the first sampling data comprise load power obtained by calculating electric energy consumed by a first station through discharge sampling in a first period, and the first period is any one period in a plurality of periods divided in advance; load information of a first station in a second period is obtained according to the first sampling data and the historical sampling data, the second period is the next period after the first period, the load information comprises a power value of the first station in the second period, and the historical sampling data comprises data obtained by sampling in a period before the first period; and determining a charging and discharging strategy according to the load information, wherein the charging and discharging strategy comprises a strategy for controlling the first station to execute charging and discharging in a second period.

Description

Load prediction method and device

Technical Field

The present application relates to the field of communications, and in particular, to a load prediction method and apparatus.

Background

With the development of technology, most devices need to be charged and discharged. Taking the fifth generation mobile communication technology (5 th-generation mobile communication technology, abbreviated as 5G) as an example, the increase of the energy consumption of the base station in the 5G scene brings great challenges to the cost of operators, and the battery charging and discharging time of the base station can be optimized by predicting the load of the base station, namely the power value of the battery of the base station when discharging, so as to achieve the purpose of saving the electric charge. However, the power supply of different base stations or the battery is usually different manufacturers, and the power supply detection cannot be directly performed to obtain the real-time load of the battery, so that the battery cannot be reasonably charged. Therefore, how to load a single cell in a situation where power detection cannot be directly performed is a problem to be solved.

Disclosure of Invention

The application provides a load prediction method and a load prediction device, which are used for predicting the load of a site based on less sampling data, and obtaining more accurate predicted load through less calculated amount under the condition that the battery load cannot be acquired in real time. In a first aspect, the present application provides a load prediction method, which is applied to an electrical power consumption system, where the electrical power consumption system includes at least one station and a power supply device, each station is provided with a battery, and the power supply device is configured to provide charging power for the at least one station, and includes: acquiring first sampling data, wherein the first sampling data comprise load power obtained by calculating electric energy consumed by a first station through discharge sampling in a first period, and the first period is any one period in a plurality of periods divided in advance; load information of a first station in a second period is obtained according to the first sampling data and the historical sampling data, the second period is the next period after the first period, the load information comprises a power value of the first station in the second period, and the historical sampling data comprises data obtained by sampling in a period before the first period; and determining a charging and discharging strategy according to the load information, wherein the charging and discharging strategy comprises a strategy for controlling the first station to execute charging and discharging in a second period.

The load change condition of the station in each period can be predicted in an iterative mode, and particularly in a scene that the real-time monitoring of the battery load of the station is difficult to realize, the accurate prediction can be realized by sampling less data, so that the charging and discharging time of the station is reasonably planned, the electricity cost of the station is reduced, the electricity consumed by sampling can be reduced, and the utilization rate of the electricity is improved.

In one possible embodiment, the foregoing obtaining load information of the first station in the second period according to the first sampling data and the historical sampling data includes: fitting according to the first sampling data and the historical sampling data to obtain a load prediction model; and acquiring load information of the first station in a second period according to the load prediction model, wherein the second period is the next period after the first period, and the load information comprises the power value of the first station in the second period.

Therefore, in the embodiment of the application, when the load prediction is performed on the next period, the load prediction model can be fitted by using the sampling data of the current period and the sampling data of the previous period or periods, then the load of the next period is predicted by using the load prediction model, and the load change condition of the station in each period can be predicted in an iterative mode. Especially in the scene that is difficult to realize the battery load real-time supervision to the website, can sample less data and can realize accurate prediction to rationally plan the charge-discharge time of website, reduce the power consumption cost of website, and can reduce the electric power that consumes because of the sampling, improve the utilization ratio of electric power.

In one possible implementation manner, the power utilization system includes a plurality of sites, and the load prediction model is obtained by fitting the first sampling data and the historical sampling data, and includes: fitting according to sampling data of a plurality of stations to obtain a group load curve, wherein the group load curve is used for representing load changes of the plurality of stations in a first period; and fitting an individual load curve according to the first sampling data and the historical sampling data, wherein the individual load curve represents the load change of the first station in the first period, and the group load curve and the individual load curve form a load prediction model.

Therefore, in the embodiment of the application, the load change curves of the group and the individual can be respectively fitted, so that the load of each site can be predicted by combining the load curves of the group and the individual in the subsequent prediction.

In a possible implementation manner, the acquiring load information of the first station in the second period according to the load prediction model may include: outputting a first predicted sequence of the first station in a second period according to the group load curve; outputting a second predicted sequence of the first station in a second period according to the individual load curve; and fusing the first prediction sequence and the second prediction sequence to obtain load information.

Therefore, in the embodiment of the application, when the load prediction is carried out on one station, the sampling data of other stations in the power utilization system can be fused, so that the predicted load of a single station is more accurate.

In one possible embodiment, the acquiring a group load curve includes: if the first period is the first period of the plurality of periods, clustering the sampling data of the plurality of sites to obtain at least one category of data; and fitting the data of at least one category respectively to obtain a load curve corresponding to each category, wherein the group load curve comprises the load curve of each category in the at least one category.

Therefore, in the embodiment of the application, after the load power of the first period is obtained by sampling, the sampled data of a plurality of stations are clustered, which is equivalent to classifying the load change conditions of the plurality of stations into a plurality of types, so that when a curve is fitted, a load curve conforming to the load change conditions of the stations can be obtained, and the accuracy of the subsequent prediction result is improved.

In a possible implementation manner, clustering the sampled data of the plurality of sites to obtain at least one category of data may include: a burr sequence is screened out from the sampling data of a plurality of stations to obtain stable data, wherein the burr sequence is data with deviation between adjacent sampling points being larger than preset deviation; clustering the stable data to obtain data of at least one category; the method may further include: and determining the category corresponding to the burr sequence according to the distance between the burr sequence and at least one category.

Therefore, in the embodiment of the application, the burr sequences can be selected when clustering is performed, so that the influence of the burr sequences on the fitting result is reduced, and the burr sequences are classified after clustering, so that the utilization rate of the burr sequences is improved.

In a possible implementation manner, after the foregoing obtaining, according to the load prediction model, the load information of the first station in the second period, the method further includes: acquiring an output value of a load prediction model at a first time point, and acquiring a load value of a first site at the first time point; and if the deviation between the output value and the load value is larger than the preset deviation value, the load information is adjusted, and the adjusted load information is obtained.

Therefore, in the embodiment of the application, whether the predicted result has regular deviation can be judged, so that the adjustment is performed in time, and the accuracy of the final predicted result is improved.

In one possible embodiment, the foregoing obtaining load information of the first station in the second period according to the first sampling data and the historical sampling data includes: if the first sampling data and the historical sampling data accord with preset conditions, acquiring load information of the first station in a second period according to the average value of the data included in the first sampling data and the historical sampling data; the preset condition comprises that a stable value is smaller than a first threshold value, and the stable value comprises at least one of variance, sample entropy, approximate entropy or fuzzy entropy of each time point in the first sampling data and the historical sampling data.

Therefore, in the embodiment of the application, when the change rule of the acquired data representation is stable, the average value of the sampled data can be directly used for predicting the load, so that accurate prediction can be completed based on a simple algorithm, the prediction calculation amount is reduced, and the prediction efficiency is improved.

In one possible embodiment, the acquiring the first sampled data includes: if the first period is the first period of the plurality of periods, sampling is carried out according to a preset first number of sampling points to obtain first sampling data; if the first period is not the first period of the plurality of periods, determining a second number of sampling points according to the load information obtained in the previous period, and sampling according to the second number of sampling points to obtain first sampling data, wherein the first number is larger than the second number.

Therefore, in the embodiment of the application, more sampling points can be sampled in the first period, and the sampling data of the previous period can be used in an iterating way when the subsequent prediction is performed, so that fewer sampling points can be sampled in the non-first period, and the power consumption generated by the sampling in the subsequent period is reduced.

In a possible implementation manner, in the foregoing process of sampling in the first period, the method may further include: when the battery of the first station is discharged, sampling is carried out according to a preset interval, and second sampling data are obtained; and adding the second sampling data to the first sampling data to obtain new first sampling data.

In the embodiment of the application, when the battery of the station is actively discharged, sampling can be performed, so that more data can be acquired under the condition that additional discharging is not needed, and the accuracy of the subsequent predicted load is improved.

In a second aspect, the present application provides a load predicting apparatus, applied to an electrical power consumption system, the electrical power consumption system including at least one station and a power supply device, each station being provided with a battery, the power supply device being configured to supply charging power to the at least one station, the load predicting apparatus comprising:

the system comprises an acquisition module, a sampling module and a sampling module, wherein the acquisition module is used for acquiring first sampling data, the first sampling data comprise load power obtained by calculating electric energy consumed by a first station through discharge sampling in a first period, and the first period is any period in a plurality of periods divided in advance;

the prediction module is used for acquiring load information of the first station in a second period according to the first sampling data and the historical sampling data, wherein the second period is the next period after the first period, the load information comprises a power value of the first station in the second period, and the historical sampling data comprises data obtained by sampling in a period before the first period;

and the determining module is used for determining a charging and discharging strategy according to the load information, wherein the charging and discharging strategy comprises a strategy for controlling the first station to execute charging and discharging in a second period.

In one possible implementation, the prediction module is specifically configured to: fitting according to the first sampling data and the historical sampling data to obtain a load prediction model; and acquiring load information of the first station in a second period according to the load prediction model, wherein the second period is the next period after the first period, and the load information comprises the power value of the first station in the second period.

In one possible implementation, the power consumption system includes a plurality of stations, and the prediction module is specifically configured to: fitting according to sampling data of a plurality of stations to obtain a group load curve, wherein the group load curve is used for representing load changes of the plurality of stations in a first period; and fitting an individual load curve according to the first sampling data and the historical sampling data, wherein the individual load curve represents the load change of the first station in the first period, and the group load curve and the individual load curve form a load prediction model.

In one possible implementation, the prediction module is specifically configured to: outputting a first predicted sequence of the first station in a second period according to the group load curve; outputting a second predicted sequence of the first station in a second period according to the individual load curve; and fusing the first prediction sequence and the second prediction sequence to obtain load information.

In one possible implementation, the prediction module is specifically configured to: if the first period is the first period of the plurality of periods, clustering the sampling data of the plurality of sites to obtain at least one category of data; and fitting the data of at least one category respectively to obtain a load curve corresponding to each category, wherein the group load curve comprises the load curve of each category in the at least one category.

In one possible implementation, the prediction module is specifically configured to: a burr sequence is screened out from the sampling data of a plurality of stations to obtain stable data, wherein the burr sequence is data with deviation between adjacent sampling points being larger than preset deviation; clustering the stable data to obtain data of at least one category; and determining the category corresponding to the burr sequence according to the distance between the burr sequence and at least one category.

In one possible implementation, after the prediction module obtains the load information of the first station in the second period according to the load prediction model, the prediction module is further configured to: acquiring an output value of a load prediction model at a first time point, and acquiring a load value of a first site at the first time point; and if the deviation between the output value and the load value is larger than the preset deviation value, the load information is adjusted, and the adjusted load information is obtained.

In one possible implementation, the prediction module is specifically configured to: if the first sampling data and the historical sampling data accord with preset conditions, acquiring load information of the first station in a second period according to the average value of the data included in the first sampling data and the historical sampling data; the preset condition comprises that a stable value is smaller than a first threshold value, and the stable value comprises at least one of variance, sample entropy, approximate entropy or fuzzy entropy of each time point in the first sampling data and the historical sampling data.

In one possible implementation manner, the acquiring module is specifically configured to: if the first period is the first period of the plurality of periods, sampling is carried out according to a preset first number of sampling points to obtain first sampling data; if the first period is not the first period of the plurality of periods, determining a second number of sampling points according to the load information obtained in the previous period, and sampling according to the second number of sampling points to obtain first sampling data, wherein the first number is larger than the second number.

In one possible implementation manner, during the sampling in the first period, the obtaining module is further configured to: when the battery of the first station is discharged, sampling is carried out according to a preset interval, and second sampling data are obtained; and adding the second sampling data to the first sampling data to obtain new first sampling data.

In a third aspect, an embodiment of the present application provides a load prediction apparatus, including: the processor and the memory are interconnected by a line, and the processor invokes the program code in the memory to perform the processing-related functions in the load prediction method according to any one of the first aspect. Alternatively, the load predicting means may be a chip.

In a fourth aspect, embodiments of the present application provide a load predicting device, which may also be referred to as a digital processing chip or chip, the chip comprising a processing unit and a communication interface, the processing unit obtaining program instructions via the communication interface, the program instructions being executed by the processing unit, the processing unit being configured to perform a process-related function as in the first aspect or any of the alternative embodiments of the first aspect.

In a fifth aspect, embodiments of the present application provide a computer readable storage medium comprising instructions which, when run on a computer, cause the computer to perform the method of the first aspect or any of the alternative embodiments of the first aspect.

In a sixth aspect, embodiments of the present application provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform the method of the first aspect or any of the alternative embodiments of the first aspect.

Drawings

FIG. 1 is a schematic diagram of an electrical system according to the present application;

FIG. 2 is a schematic diagram of another power utilization system according to the present application;

fig. 3 is a schematic view of an application scenario provided by the present application;

fig. 4 is a schematic diagram of another application scenario provided in the present application;

fig. 5 is a schematic diagram of another application scenario provided in the present application;

FIG. 6 is a schematic flow chart of a load prediction method according to the present application;

FIG. 7 is a schematic flow chart of another load prediction method according to the present application;

fig. 8 is a schematic diagram of another application scenario provided in the present application;

FIG. 9 is a flowchart of another load prediction method according to the present application;

FIG. 10 is a flowchart of another load prediction method according to the present application;

FIG. 11 is a flowchart of another load prediction method according to the present application;

FIG. 12 is a flowchart of another load prediction method according to the present application;

fig. 13 is a schematic structural diagram of a load prediction device according to the present application;

fig. 14 is a schematic structural diagram of a load prediction device according to the present application.

Detailed Description

The following description of the technical solutions according to the embodiments of the present application will be given with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

The load prediction method provided by the application can be applied to an electric system, the electric system can comprise power supply equipment and electric equipment, the number of the power supply equipment can be one or more, the number of the electric equipment can be one or more, and the power supply equipment can be used for providing power for the electric equipment.

For example, the architecture of the power consumption system provided by the application may be shown in fig. 1, where the power consumption system may include a power supply device and one or more electric devices.

The power supply device may be used to provide power to the powered device. The electrical consumer may be powered by gasoline, diesel, other petroleum-based fuels, propane, other compressed gas-based fuels, ethanol, solar panels, batteries, or other sources of electricity.

The electric equipment can be provided with a battery, the electric equipment can acquire electric power from the power supply equipment and charge the battery, and the battery discharges to provide energy required by work for the electric equipment. The battery may be a lithium ion battery, a lead acid battery, or the like. One or more battery packs of such batteries may be configured with a power supply to provide power to various components of the powered device.

For ease of understanding, the consumer is referred to herein as a site, i.e., one or more sites may be included in the consumer system.

In general, to facilitate charge management for individual sites, control nodes may be provided in the power usage system, or may be referred to as load prediction devices. Specifically, the control node may be set separately, for example, a server, a terminal, or a cloud server may be set separately as the control node, or one of the sites may be set as the management node, or a gateway in the communication system may be set as the control node. The method provided by the application can be executed by the control node, namely, the load prediction method provided by the application can be executed by a server, a terminal, a cloud server or a site which are arranged independently.

The power utilization system provided by the application can be deployed in various scenes, such as the scenes of communication system power supply, vehicle power supply, household appliance power supply or terminal power supply, and the like, and is exemplified by specific scenes.

Such as a wireless network, a wired network, or a combination of wireless and wired networks, etc. The wireless network includes, but is not limited to: a fifth Generation mobile communication technology (5 th-Generation, 5G) system, a Long term evolution (Long term evolution, LTE) system, a global system for mobile communication (global system for mobile communication, GSM) or code division multiple access (code division multiple access, CDMA) network, a wideband code division multiple access (wideband code division multiple access, WCDMA) network, the internet of things, wireless fidelity (wireless fidelity, wiFi), bluetooth (blue), zigbee (Zigbee), radio frequency identification technology (radio frequency identification, RFID), long Range (Lora) wireless communication, near field wireless communication (near field communication, NFC). The wired network may include a network of fiber optic communications or coaxial cables, etc.

Accordingly, the station provided by the present application may be a base station (eNodeB, eNB) in long term evolution (long term evolution, LTE) within a communication network, a base station (gndeb, gNB) in a new radio, NR), and so on. For example, from a product morphology, a base station is a device with a central control function, such as a macro base station, a micro base station, a hot spot (pico), a home base station (femto), a Transmission Point (TP), a Relay (Relay), an Access Point (AP), and the like.

In some other scenarios, such as a vehicle charging or home charging scenario, the station may be a powered device, such as a vehicle, a home appliance, or a terminal, and the corresponding control node may be one of the powered devices, or a separately configured node.

For example, the power system architecture may be as shown in fig. 2, which may include power supply devices, cloud servers (otherwise known as cloud management systems), gateways, and sites.

The cloud server is used for processing data acquired from the site, predicting the load of the site based on the processed data and calculating the sampling point of the next period. After the load is predicted and the sampling point is calculated, the information of the sampling point strategy table can be issued to the gateway, and after the charge and discharge strategy is obtained based on the load, the charge and discharge strategy is issued to the gateway.

The gateway establishes connection with the station, and can send instructions to the station to control the station to complete operations such as charging or discharging. After receiving the sampling point policy table issued by the cloud server, the gateway can control the site to charge and discharge in the next period based on the information of the sampling points included in the sampling point policy table, namely the sampling period, so as to complete sampling. After receiving the charge and discharge policy table, the charge and discharge of the station can be controlled based on the charge and discharge policy table.

The station can be used for collecting current or voltage and the like during self charging and discharging, and the collected data can be reported to the cloud server through the gateway, so that the cloud server can know the load of the station based on the received data, namely the load power of the station during discharging. Specifically, the site may report the sampling data periodically, or report the sampling data to the cloud server after sampling is completed at each sampling point.

It should be noted that, the foregoing gateway is an optional device, and the steps executed by the gateway may be integrated in a site or may be integrated in a cloud server, and may be specifically determined according to an actual application scenario, and the present application is merely illustrative.

Some common approaches for load prediction for a site may include: directly averaging according to the load value of the same time point of the history to carry out load prediction; fitting based on multiple linear regression, and continuously optimizing by using test data; modeling and predicting each base station load respectively by directly utilizing a time sequence prediction algorithm such as ARIMA and the like or other machine learning algorithms; and performing machine learning model training by using other local point data, and performing transfer learning training and prediction on the new local point data.

However, the situation that the power supply and the battery are not the same is very common, real-time load data are difficult to directly acquire, and accurate prediction of the load of the base station is required to be realized for realizing intelligent peak-shifting charge and discharge. Because the load and time and other related variables are not necessarily in a linear relation, and different base station time sequence modes are different, under the condition of less data, the modeling accuracy of the first three modes is not high, the requirement on the data integrity is high, and each base station needs to be independently modeled and optimized. The fourth approach relies on other local data and is difficult to handle irregularly sampled and sparsely sampled scenes. The load prediction method provided above cannot be adapted to the situation of uneven time points, and is difficult to capture the periodicity law under the situation of small data volume. And under the condition that site load data cannot be directly obtained, frequent load sampling through discharging causes power charge loss to reduce peak-shifting income.

Therefore, the application provides a load prediction method, which is used for accurately predicting the load condition of the station under the condition that the load data of the station cannot be obtained in real time directly through the power supply monitoring function or the data is less, so that the charge and discharge time is reasonably planned, and the running cost of the station is reduced. In addition, the prediction can be performed by using fewer discharge times, the discharge times are reduced to reduce extra power consumption, and the prediction cost is reduced.

First, for easy understanding, some application scenarios of the load prediction method provided by the present application are described in an exemplary manner.

Power supply system in scene one and communication system

Illustratively, taking a 5G communication system as an example, as shown in fig. 3, the communication system may include a control node, a gateway, base stations (base stations 1 to N as shown in fig. 3), and a power supply device for supplying power to each base station.

Wherein, can set up the battery in the basic station, can use the battery to supply power for the basic station. The battery is usually required to be charged, and a power supply, namely a power supply device, can be arranged for each base station, and the battery is charged through the power supply device, so that the electric quantity of the battery can ensure the normal operation of the base station.

The control node can control sampling, charging and discharging of the station through the gateway. For example, a charging instruction or a discharging instruction can be sent to the station through the gateway, and current or voltage information collected by the station is received through the gateway, so that load information of the station is obtained.

In a communication network, a base station may be fixed in location after being deployed at a certain location, and the base station may be in a state of long-term operation, so that the consumption of a battery by the base station is also large, and frequent charging is usually required periodically. And the load may be different at different base stations in general, and the cost of electricity generation is different. The charge and discharge time of the base station battery can be optimized by using the peak-valley electricity price difference, so that the purpose of saving electricity charge is achieved. The battery is usually used for discharging and supplying power in the peak period of electricity price, and is charged in the peak period of low electricity price, and meanwhile, the mode of reserving spare electricity quantity is guaranteed, so that the purpose of safely and reliably utilizing spare electricity quantity of the spare electricity battery and saving electricity charge is achieved.

Generally, 5G has large overall and static energy consumption, a large number of base stations, uneven distribution of user services in space-time, and waste of networks meeting peak demands is generated at idle time. The conventional fixed time peak shifting mode is difficult to realize optimal resource utilization efficiency. In addition, the power supply equipment and the battery of the base station can be different manufacturers, the market occupation of the mixed-match scene of the lithium battery and the lead-acid battery is large, and the load data of the base station can not be obtained in real time through the power monitoring function, namely, the power of the electric energy generated, converted or consumed by the base station in operation can not be obtained directly. And in the case of some less data, such as a scene with only one day of sampling data, it is difficult to achieve better prediction effect by using the machine learning algorithm.

Therefore, by the load prediction method provided by the application, the load condition of the base station can be accurately predicted under the condition that the load data of the base station cannot be obtained in real time directly through the power supply monitoring function or the data is less, so that the charge and discharge time is reasonably planned, and the running cost of the base station is reduced. In addition, the prediction can be performed by using fewer discharge times, the discharge times are reduced to reduce extra power consumption, and the prediction cost is reduced.

Scene two, vehicle charging scene

For example, a scenario of vehicle charging may be as shown in fig. 4.

Wherein a battery may be provided in the vehicle, through which battery the energy required for the operation of the vehicle is provided. The user can drive the vehicle to the stake of charging and charge, fills electric pile and can set up in places such as parking area, charging station. The vehicle may need to be charged at the same or different charging piles per cycle.

In general, the electricity prices of different electricity utilization times are also different, for reducing the electricity price of the vehicle, for example, the electricity price of electricity supplied by the commercial power and the electricity price of electricity supplied by solar energy are different, electricity may be supplied by the commercial power or a combination of the commercial power and the solar energy during the peak period or at night, and electricity may be supplied by the solar energy during the peak period. Thus, the charging can be performed in the low-price peak period and the discharging can be performed in the high-price peak period. The specific charge and discharge time of the vehicle can be determined through the load of the vehicle, so that the normal driving of the vehicle is ensured on the basis of saving the driving cost of the vehicle. While the power supply and the charging pile of different vehicles may be different manufacturers, it is difficult for the separately provided control node to collect the battery load of the vehicle in real time, and if the sample collected for only one vehicle is less, it is also difficult to accurately predict the battery load of the vehicle by means of machine learning.

Therefore, by the load prediction method provided by the application, the load condition of the vehicle can be accurately predicted under the condition that the battery load of the vehicle cannot be obtained in real time or the data is less, so that the charge and discharge time of the vehicle can be accurately planned, and the driving cost of the vehicle is reduced.

Scene three, household charging scene

Generally, referring to fig. 5, in a home charging scenario, an energy storage device (i.e., a site) may be provided, as well as a home appliance, wherein the energy storage device is provided with a battery to store energy for the home appliance. The household appliance can be powered by mains supply, solar energy, and electric energy provided by the energy storage device. The household appliances may include televisions, drinkers, floor sweeping robots, intelligent rice cookers, refrigerators, ovens, or household vehicles. For example, solar power may be insufficient during peak periods of electricity consumption, and mains electricity or an energy storage device may be used to supply electricity, while the energy storage device may be charged by solar energy during peak periods of electricity consumption, so electricity prices for electricity consumption may be different during different periods. Specifically, a terminal or other control nodes can be independently arranged to plan charging and discharging time points of the household appliances and the energy storage device. However, different appliances may be produced by different manufacturers, and a terminal or a control node may not be able to learn the load of the appliance in real time, and the load of the appliance may be changed by environmental changes or user selections, so that a specific charge and discharge time of the appliance may not be accurately planned.

By the load prediction method provided by the application, the load condition of the electric appliance can be accurately predicted under the condition that the battery load of the electric appliance cannot be obtained in real time or the data is less, so that the charging and discharging time of the electric appliance can be accurately planned, and the use cost of the electric appliance is reduced.

It should be noted that the foregoing description is merely illustrative of an application scenario of the load prediction method provided by the present application, and the present application is not limited to this scenario, and may be applied to other scenarios requiring charge and discharge.

The following describes the flow of the load prediction method provided by the present application in detail in conjunction with the foregoing communication system and application scenario.

First, referring to fig. 6, a flow chart of a load prediction method provided by the present application is as follows.

It should be noted that, the present application is exemplified by taking one of the sites (for convenience of distinction as the first site) as an example, and the first site mentioned below may be replaced by any one site in the power consumption system, which is not limited to this aspect of the present application.

Further, the period may be divided in advance, for example, 12 hours as one period, 1 day as one period, 2 days as one period, or the like. The present application is exemplified by one of the periods, which is divided into the first period for convenience.

601. First sample data is acquired.

The first sampled data may include a calculated load power of consumed electrical energy collected by the first station when discharging during the first period.

Specifically, if the method provided by the application is executed by the set control node, the control node can directly receive or receive the data reported by the first site through the gateway to obtain the first sampling data. If the method provided by the application is executed by the station, the station can directly acquire the data acquired by the station, so as to obtain the first sampling data.

More specifically, the battery may be discharged at each sampling point by the station, and the current or voltage of the battery at the time of discharge or the like may be collected. If the method provided by the application is executed by the set control node, the station can report the current or voltage of each sampling point to the control node, and the control node calculates according to the received current or voltage to obtain the load power of the station when discharging at each sampling point. Of course, the station can also calculate the load power by sampling the obtained current or voltage, and report the load power directly to the control node.

If the first period is the first period for sampling the first station, sampling may be performed according to a first number of sampling points in the period, and if the first period is not the first period for sampling the first station, sampling may be performed according to a second number of sampling points, where the first number is greater than the second number. For example. If the first period is the first sampling period, sampling can be performed according to 96 sampling points to obtain complete sampling data; if the first period is not the first sampling period, 12 sampling points can be selected for sampling, so as to obtain relatively incomplete sampling data. Therefore, in the embodiment of the application, the sampling can be performed at a plurality of time points in the first period, and the sampling points can be reduced in the subsequent period, so that the power consumption caused by sampling is reduced.

Optionally, in each period, the discharging condition of the first station may be detected, and if the battery of the first station discharges during use, sampling may be performed at preset intervals, and the current, the voltage, the power value, and the like of the battery may be collected to obtain second sampling data. And adding the second sampling data to the first sampling data to obtain more complete new sampling data. It can be understood that during the operation of the first station, the battery of the first station may be passively discharged, in which case, the discharging condition of the battery of the first station may be monitored, and the current, voltage or power value of the battery may be collected, so as to obtain the load condition under the passive discharging condition. Therefore, in the embodiment of the application, when the first station passively discharges, sampling can be performed without separately discharging, so that the discharge information of the first station is fully utilized, and the power consumption for separate sampling is reduced.

In addition, if the first period is not the first period for sampling the first station, the sampling point of the first period may be calculated according to the load information predicted from the previous period. For example, a point with a larger difference point between the cluster estimated value of the previous period and the fitting estimated value of the single station can be used as the sampling point of the first period. In this scenario, the data acquired in the first period may be used to correct the points with larger differences, so that the fitting result obtained in the current period is more accurate.

602. And acquiring load information of the first station in the second period according to the first sampling data and the historical sampling data.

After the first sample data is obtained, the first sample data and the historical sample data are used to predict load information of the first station in the second period. If the first period is not the first period for sampling the first station, the historical data may be data collected in a period before the first period, and if the first period is the first period for sampling the first station, the historical data may be null, i.e. the historical data is not used in the first period to predict the load information of the second period.

In one possible scenario, whether the first sampled data and the historical sampled data are stable or not may be determined, for example, whether the first sampled data and the historical sampled data meet a preset condition or not may be determined, and if so, load information of the first station in the second period may be directly predicted according to the average value of the first sampled data and the historical sampled data. The stable value of the data may be represented by using at least one of variance, sample entropy, approximate entropy, or fuzzy entropy of each time point in the first sample data and the historical sample data, and the preset condition may be that the stable value is smaller than a first threshold, that is, the regular stability of the data is represented. It can be understood that when the rule represented by the sampled data is stable, for example, the relationship representing the load change with time is in a multi-segment linear relationship, the average value of the first sampled data and the historical sampled data can be calculated, the relationship of the load change with time is fitted by using the average value, and the load change condition of the first station in the second period is predicted according to the relationship. Therefore, in the embodiment of the application, under the condition of stable data, the load prediction can be performed in a mean value calculation mode, the calculated amount is reduced, and the predicted load can be obtained efficiently and rapidly.

Of course, when the first sampled data and the historical sampled data are stable, the load change condition of the first station in the second period can also be predicted by adopting the following prediction modes, and the mode of predicting the load information of the first station in the second period when the data are stable is not limited by the application.

Specifically, the first sample data and the historical sample data may be used to fit a load prediction model, which is used to predict load information of the first station in a next period (for convenience of distinction, into a second period), which may include predicted power values of the first station at respective sample points in the second period.

For example, the load information may be represented as a sequence, and the sequence may include load power of the first station at a plurality of time points in the second period, that is, a power value for consuming electric energy; alternatively, the load may be directly expressed as a variation relationship of load and time.

Alternatively, multiple sites may be included within the power system and the load prediction model may include a group load curve and an individual load curve. The group load curve may represent a time-varying load of the plurality of sites as a whole over the first period, and the group load curve may be fitted using sampled data of the plurality of sites. The individual load curve is obtained by fitting the first sampling data and the historical sampling data and is used for representing the change condition of the load of the first station in the first period along with time.

Optionally, if the load prediction model may include a group load curve and an individual load curve, when the load information of the first station in the second period is predicted by the load prediction model, the prediction results of the group load curve and the individual load curve may be fused to obtain the load information of the first station in the second period. Specifically, the group load can be used to output a first predicted sequence of the first station in the second period, the individual load curve is used to output a second predicted sequence of the first station in the second period, and then the first predicted sequence and the second predicted sequence are fused, so that the load information of the first station in the second period can be obtained. Therefore, in the embodiment of the application, the load information of the single site can be predicted by combining the load change conditions of the group and the single site, so that a more accurate prediction result is obtained.

Optionally, if the first period is the first period of sampling the first site, the sampled data of the plurality of sites in the power consumption system may be clustered to obtain data of at least one class, and then the data of each class is used to fit the data of each class to obtain a load curve corresponding to each class, where the load curve of at least one class may form the load curve in the group. Therefore, in the embodiment of the application, the collected data of a plurality of stations can be clustered to realize classification of the stations, and the load curve of each category is fitted according to the category, so that the obtained load curve is more adaptive to the load change condition of the stations, and the accuracy of the subsequent prediction result is further improved.

In one possible implementation manner, in the process of clustering the sampled data of the multiple sites, at least one group of burr sequences can be screened from the sampled data of the multiple sites to obtain residual stable data, namely, data with deviation between adjacent sampled points being greater than preset deviation, and then the residual stable data is used for clustering to obtain at least one class of data. Subsequently, the distance between each group of spike sequences and each class may be calculated and the class to which each group of spike sequences belongs determined. If the distances between each group of burr sequences and each category can be compared, the burr sequences are classified into the category closest to the burr sequences, so that the data obtained by sampling can be fully utilized, and the accuracy of the subsequent prediction result is improved.

In one possible implementation, the load of the station may change with time, or regular drift may occur with data accumulation, so it may be determined whether regular drift occurs, so that the load information is adjusted based on the drift result, so as to obtain accurate load information. If the deviation is larger than the preset deviation value, the load information obtained through prediction can be adjusted, and the adjusted load information is obtained. Therefore, even under the condition of abrupt change of the load of the base station, the method provided by the application can obtain accurate predicted load information.

603. And determining a charging and discharging strategy according to the load information.

After the load information of the first station in the second period is predicted, the charging and discharging strategy of the first station in the second period can be determined according to the load information, so that the charging and discharging time point of the first station is controlled through the charging and discharging strategy, and the electricity cost of the first station is reduced.

In general, after the load change condition of the first station is known, the charging time point and the discharging time point of the first station can be determined, so that the first station is charged when the electricity price is kept at a low peak, and is discharged when the electricity price is at a high peak, and the electricity consumption cost of the first station is reduced. Therefore, in the embodiment of the application, the load change condition of the station in each period can be predicted in an iterative mode, and particularly in the scene that the real-time monitoring of the battery load of the station is difficult to realize, the accurate prediction can be realized by sampling less data, so that the charging and discharging time of the station is reasonably planned, the electricity cost of the station is reduced, the electricity consumed by sampling can be reduced, and the utilization rate of the electricity is improved.

The foregoing describes the flow of the load prediction method provided by the present application, and for convenience of understanding, the flow of the load prediction method provided by the present application is described in more detail below in conjunction with a specific application scenario.

First, a more complete flow of the present application is illustrated, and referring to fig. 7, a flow chart of another load prediction method provided by the present application is shown.

First, the full sampling 701 is performed in the first cycle, that is, the full data sampling is performed for the charge and discharge of the battery in the first cycle, which is different from the sampling in the subsequent cycle in that the number of sampling points in the first cycle is greater than the number of sampling points in the subsequent cycle. For ease of distinction, the samples of the first cycle are referred to as complete samples, and the samples of the subsequent cycle are referred to as incomplete samples.

The data processing 702 is then complete for the first cycle. That is, after the first period is completely sampled, the sampled data obtained by sampling the first period may be processed, such as filtering abnormal data in the sampled data, clustering the sampled data, and the like.

After processing the sampled data for the first complete cycle, the load prediction model 703 may be fitted based on the processed data, i.e., the load over time.

Then, a load prediction result 704 of the next cycle is obtained based on the load prediction model. The load prediction result of this next cycle may be used to determine a charge-discharge strategy.

And, after the load prediction result of the next cycle is obtained, the incomplete sampling point 705 of the next cycle may be calculated based on the load prediction result of the next cycle.

The next cycle of incomplete data sampling 706 is then performed based on the incomplete sample points of the next cycle, and the next round of prediction is continued using the data obtained by the incomplete data sampling and the sample data of the previous cycle, e.g., the load prediction model is re-fitted using the data obtained by the incomplete data sampling and the sample data of the previous cycle, i.e., steps 703-706 are repeatedly performed until no adjustment of the charge-discharge strategy is required.

Therefore, in the embodiment of the application, the complete data sampling is only carried out in the first period, the complete data sampling is not needed in the subsequent period, the power consumed by sampling is reduced, and the accurate load information can be predicted and obtained on the basis of less data, so that a charge-discharge strategy matched with the load can be obtained, and the power consumption cost of a station is reduced.

More specifically, the load prediction method provided by the present application may be deployed in a control node, and may also be deployed in each site, where the difference is that, when deployed in the control node, the load of a single site may be predicted using the sampled data of multiple sites, and when deployed in a single site, the load prediction may not be performed using the sampled data of other sites, and in the following, different scenarios are exemplarily described respectively.

1. Deployed at control node

For example, when the method provided by the application is deployed at the control node, the architecture of the corresponding power utilization system can be shown in fig. 8, and in the power utilization system, a server or a terminal or the like can be set as the control node to manage charging and discharging of a plurality of sites.

Specifically, as shown in fig. 9, for example, the control node may collect information related to the actual load of the station, such as information of current, voltage, or power value during discharging, through the gateway.

Furthermore, the load of the nodes may vary, often in different environments, such as different dates, places, weather, etc. For example, the time period of use of a possible node increases on holidays, and the load thereof may also increase; the geographic location is a commercial area or is closer to the commercial area, the load on the site may be greater than a site farther from the commercial area, etc. Therefore, the control node can acquire external input data in other modes, such as acquiring external environment information, such as date, map or weather information, through other servers, so that when the load change curve is fitted, the external environment information can be combined to fit to obtain more accurate load change conditions.

After receiving the actual load information and the external input data, the control node can store the received data in the space-time calculation database, and then, carry out load prediction and sampling point calculation of the next period based on the data stored in the space-time calculation database.

After load prediction and sampling point calculation, information preprocessing, such as data complement, can be performed on the calculated data. And then customizing a charging and discharging strategy based on the preprocessed data and battery information reported by the gateway, and issuing a charging strategy to the gateway, so that the gateway is used for controlling the station to charge or discharge according to the charging and discharging strategy.

A more detailed flow is described below by way of example, and referring to fig. 10, another flow diagram of a load prediction method provided by the present application is described below.

1001. The first cycle is a complete data sample.

When there is no site-related data in the control node at the first period, the period may be divided in advance and then sampling is performed from the first period.

Firstly, the period division may be performed, for example, dividing each 8 hours into one period, dividing each 12 hours into one period, dividing each day into one period, dividing each month into one period, dividing each 3 months into one period, and the like, and the present application is not limited thereto.

When the control node is in the first period, the control node has no load information of the station, and the station can be completely sampled. For example, the whole sampling pattern may be divided into a complete sampling and a non-complete sampling, with the difference that the number of sampling points of the complete sampling is greater than that of the non-complete sampling.

For example, the embodiment of the present application takes one cycle as one day, the complete sampling is to set 96 sampling points, sampling is performed every 15 minutes, and the incomplete sampling is to set 12 sampling points as an example. In the first period, the sampling points are 96, namely 96 samples are sampled, 70% of electric quantity is reserved as discharge capacity by default, sampling is carried out in a mode of discharging for 1 minute every 15 minutes, instantaneous current and voltage values of battery discharge are collected, average power is calculated by using average values of current and voltage in the discharge period, the average power is used as load data of the current sampling point, and 96 samples in a complete period are obtained after 96 times of discharging.

More specifically, in the case where the control node cannot read the load of the site in real time, a control instruction or a charging policy may be sent to the site through the gateway, so as to control the site to discharge at the sampling point. In the discharging process, the discharging condition of the station, such as current or voltage, can be monitored by controlling the station, so that the load change condition of the station is calculated.

Optionally, in some cases, the data interface of the battery management system (battery management system, BMS) of the battery cannot be obtained, so that the discharging current and the discharging voltage of the battery cannot be directly read, and an ammeter or a similar device needs to be additionally installed to monitor the discharging quantity of the battery, and load data sampling is realized by using the electric energy loss estimation load of short-time discharging monitored by the ammeter. Even if the battery data (such as current, voltage and the like) cannot be directly read, load sampling is indirectly obtained through the ammeter, so that accurate sampling data is obtained.

1002. Data cleaning and complementation.

After 96 samples are obtained in the first period of sampling, outliers can be removed, so that interference of the outliers on prediction is eliminated.

In particular, three times the standard deviation, the 3 sigma principle (x _t >Mu+3σ or x _t <Sample of μ -3σ, x _t For the t-time samples, μ is the sequence mean, σ is the sequence variance), the data are screened to screen outliers in 96 samples.

If the missing value exists in the sampled data or the outlier is removed, the missing value can be complemented so as to meet the requirement of first day data fitting. Specifically, the method can use linear interpolation, front-back value filling or spline fitting interpolation and other modes to finally ensure that each site on the first day has 96 sampling point complete information.

1003. Whether the data is stable or not is determined, if so, step 1004 is executed, and if not, step 1009 is executed.

The first day has 96 pieces of complete information of sampling points, the data are clustered after data processing, whether the time sequence of one day is stable or not is judged, mean value prediction can be directly utilized for the stable time sequence, model fitting is not needed, step 1009 is executed, and curve fitting is needed for non-stable data, namely step 1004 is executed.

Specifically, there may be various ways to determine whether the sample is stable, for example, whether a certain index of the sample is less than a first threshold, where the index may include at least one of variance, sample entropy, approximate entropy, or fuzzy entropy. For example, using variance as an example, it is possible to determine whether the variance is smaller than a threshold value at each time pointWhether the sequence is stable or not can be predicted better if the daily average value is used, namely the time sequence is considered to be stable, and if the daily average value cannot be used for accurately predicting the daily load, namely the time sequence is considered to be unstable. For example, the first daily load { x over a site _t ,t＝1,2,…,96}，(mu is the average value of 96 sampling points of the sequence, x _t The load value at time t, s is x _t A ratio of greater than 5% deviation from μ), the stabilizing sequence is set to s >90%, the unstable site is set as s<＝90％。

1004. And (5) clustering data.

In the embodiment of the application, in order to obtain the model which accords with the load change rule, the samples can be clustered, so that the model is constructed for different categories, the loads of the stations in different categories are predicted, and the more accurate predicted load is obtained.

In step 1004, the clustering is performed only in the first prediction process, and the data used in the previous iteration can be directly used in the prediction process of the subsequent period without performing the clustering again.

Alternatively, the burr sequence may be removed first, as the burr sequence has a larger influence on the clustering result. For example, the index c=q1/Q3 (Q1, Q3 are the top and bottom quarter points of the sequence) can be used to distinguish the data spike situation, and if the ratio c is greater than a certain threshold, the adjacent load in the daily load is considered to have severe fluctuation, and is marked as a spike sequence; ratio ofIf the load is smaller than a certain threshold value, the adjacent load fluctuation in the daily load is considered to be smaller, and the non-burr sequence is recorded.

The data for the unstable, non-bursty sites may be clustered based on scene and timing morphology. For example, the kmeans algorithm may be used to classify sites with similar spatio-temporal characteristics (scene where the site is located, site timing morphology).

Specifically, the data is compressed to [0,1 ] by first normalizing the timing]At the same time protectHolding other properties of the data unchanged; to determine the optimal number of clusters, the sum of squares variation in the group is used to be less than a certain threshold value for setting (t is the time point, k is the category, i is the site, x _it For the load of station i at time t, c _kt A cluster center point of the k-th class at the time t). />

After clustering, each class of site data has a corresponding class number, and the burr site sequence which does not participate in clustering determines the class number by calculating the distance between the sequence and the clustering center of each class, and the class number with the smallest distance is the class number. Specifically, the distance may be measured in a variety of ways, such as by Euclidean distance, cosine distance, manhattan distance, etc., to the burr sequence from each category. For example, the distance may be calculated using the euclidean distance, for example, x is a sequence to be determined, y is a cluster center sequence of a certain class, time points 1 to n (n=96), and the distance d (x, y) is calculated:

the burr sequences can then be categorized into categories closest to them, thereby improving the utilization of the data.

1005. Population and individual load curve fitting and fusion prediction are performed.

In general, because sequences within one cycle of the same class of stations have similarities, each class of stations can be fitted to make the prediction result more robust to interference from single cycle sample fluctuations.

The manner in which the group load curve, the individual load curve, and the predicted load are each described in detail is exemplified below.

1. Population load curve fitting

For example, the regression order under each category may be determined using AIC criteria in a polynomial fit. Aic=2k+nln (RSS), where k is the number of model parameters, n is the number of samples, and RSS is the sum of squares of the residuals. The final order of each class is determined based on the two test AIC reduction being less than a certain threshold.

For all loads under a certain class, polynomial fitting with time points as independent variables can be used, and a fitting curve under each class is recordedWherein k is the category number, t is the time point, and the estimated value of the next period population curve is obtained after inverse normalization is +.>Spline curve fitting or other curve fitting methods can be used for the fitting method.

2. Individual load curve fitting

Individual curves may be fitted to a single site sequence to predict the next cycle timing. For example, the first period complete sample value can be directly used as the next period predicted value, assuming that the first period complete sample is { x } _t T is {1,2, …,96}, thenWherein->Estimating a value for the complete load of the next period; polynomial fitting may also be performed using time points as independent variables, as with population curves. Spline curve fitting or other curve fitting methods can be used for the fitting method.

3. Fusion prediction

And the group rule and the individual rule are fused and predicted, and under the condition of insufficient information quantity, a similar sequence can provide supplementary information for the individual curve, so that the fitting precision is improved. Specifically, 96 time points load Est1 for the next period of population fit curve estimation and 96 time points negative for the next period of individual fit curve estimation can be usedThe load Est2 performs a weighted average as the final result. I.e. the final load estimation result of the next periodAlpha is an adjustable coefficient. Alpha can be learned through historical load training data, and alpha with highest precision is obtained through searching and is used as alpha in final prediction; when α is difficult to determine, the arithmetic average value of α=0.5 may be directly used for output.

Therefore, in the embodiment of the application, the data of a plurality of stations can be clustered, so that the individual load curve is fitted according to the category, the data quantity aiming at a single station is improved, the accuracy of the obtained load curve is improved, and the accuracy of the predicted load is further improved.

1006. Whether drift occurs is determined, if so, step 1007 is performed, and if not, the next periodic load prediction result 1008 is output.

After the predicted load is obtained, in order to improve the accuracy of the predicted load, whether a situation of regular drift exists can be judged, if the situation of regular drift exists due to accumulation of data, the concept drift of the data possibly occurs, namely, the rule of the data changes, so that the model accuracy is reduced, the data adapt to a new rule by carrying out drift adjustment on the data with drift, and therefore, a fitted curve can estimate the future load more accurately.

Specifically, it is first necessary to determine whether or not the data has drifted (comparing the predicted value with the actual sampling value). For example, it is known that a certain day of a certain site i is estimated asThe real load of the sampling point on the same day is { y } _t ,t∈T _s }, whereinNumber of samples T _s And is more than or equal to 12. The method can judge whether the current day estimation is accurate or not by assuming the load predicted valueAnd true value y _t Deviation of->A proportion of deviation less than 5->If m is less than or equal to 0.5, the drift occurs, and if m>0.5, no drift occurs.

In addition, whether the data has concept drift can be judged by using a machine learning model, a sliding window characteristic difference method and the like. If the observed data drift, the subsequent predictions need to ignore the data before drift and readjust the predictions.

1007. And (5) drift adjustment.

After confirming that there is a regular drift, the predicted load obtained in step 1005 may be adjusted in order to improve the accuracy of the predicted load.

Specifically, for example, the predicted value after the drift may be adjusted to a new rule by means of difference and proportion adjustment in the following manner: for site i, at T _m The day has prediction drift, T _m Day sampling data is { y } _t ,t∈T _s T (th) _m Data of day-1 (non-first day is 96 time-point load values estimated based on curve fitting, first day is 96 real sample load values) is { z _t ,t＝1,2,…,96}，T _m Daily sampling load averageT (th) _m -1 day and T _m Daily sampling load mean at the same time point +.>The difference between them->T (th) _m Day's complete load estimation curve y _t Adjust to-> T _m -1 day load curve ratio of time points to total day load +.>Then T is _m The estimated value after +1 day load adjustment is +.>And aiming at the drifting site, predicting and inputting the third day and later after drifting, and only keeping historical sampling data from the drifting day for curve fitting, and obtaining a final prediction result according to fusion prediction of a clustering curve and the latest site fitting curve.

1008. The next period load prediction results.

The load prediction result may include a load change condition of the single station in the next period.

1009. Mean value prediction is utilized.

If it is determined in step 1003 that the step data is stable, the average value of the loads of the respective sites may be calculated, and the average value may be used as the predicted load of the site, thereby reducing the calculation amount.

1010. The next periodic incomplete sampling point is calculated.

After the predicted load is obtained, the sampling point for the next cycle can be calculated.

Taking 15-minute interval load sampling interval as an example, assuming that one period is one day, the complete sampling of one day is 96 sample points, and starting from the second day, 12 sampling points are preset, namely 12 time points are selected from 96 sampling points of one day to sample.

By way of example, a specific sampling rule may be as shown in fig. 11. Samples were taken every 15 minutes with one cycle a day. Discharging 96 times in the first day to collect a complete sample of one day; the 96 sampling points on the first day are then used to predict the load of 96 points which are complete on the second day, the sampling points on the second day are calculated, and sampling is performed on the calculated sampling points on the second day. In addition, in the sampling process, the battery can be passively discharged, and sampling can be performed every 15 minutes to obtain sampling data. On the nth day, 94 loads on the nth day are predicted by using the historical loads on the first n-1 days, 12 sampling points on the nth day are calculated, sampling is performed based on the calculated 12 sampling points, and the like.

Specifically, for example, the sampling point on the next day may predict according to the load on the first day, and reject the point of low value (i.e. low information amount) until the target number of load monitoring points is obtained, i.e. the sampling point on the next day.

In this embodiment, a point with a large difference between the cluster fitting estimation value of the next day and the single station fitting estimation value is used as the sampling point of the next day. The calculation mode may be diff=abs (est_1-est_2)/(est_1+est_2), where est_1 is a cluster estimation result, est_2 is a single-station estimation result, diff is an array of 96 points, and the 12 points with the largest diff are selected as sampling points of the next day. In the environment where the computing resources are limited or the cluster estimation is not available, 12 time points can be selected as sampling points of the next day directly through a random sampling mode.

In general, in a normal off-peak discharge period, load value sampling is directly obtained while discharging; and in the non-charging and non-discharging stage, discharging for 1 minute according to the calculated sampling time points to sample, and finally obtaining 12 or more actual load samples.

1011. The next period is a non-complete data sample.

After the predicted load of the next cycle is used to determine the sampling point of the next cycle, that is, when sampling is performed in the next cycle, incomplete sampling is performed at the calculated sampling point, and then load prediction of the next cycle can be performed.

For example, after 12 sampling points are determined according to the predicted load, discharging can be performed at the 12 sampling points, and the current and voltage values are monitored to obtain the load power of the battery.

1012. And issuing a charging strategy.

In the prediction process for each period, after the load prediction result of the next period is obtained, a charging strategy can be determined according to the load prediction result, for example, the load prediction of the next period is performed at a fixed time point of each period, and the incomplete sampling point of the next period is calculated. And calculating the required standby electricity quantity of the battery at each station in the next period according to the load prediction result and the standby electricity duration, calculating the optimal charge-discharge strategy according to the electricity price meter, and simultaneously issuing the optimal charge-discharge strategy and the discharge strategy required by the incomplete sampling point to the intelligent gateway equipment to control the battery to perform charge-discharge.

Therefore, in the embodiment of the application, complete sampling is performed in the first period, load prediction can be performed in the subsequent period based on the data acquired in the previous period, cold start can be realized in a scene with less data, and the method is suitable for a scene with fewer samples. And the number of sampling points in the subsequent period is smaller than that in the first period, so that the power consumption generated by sampling can be reduced. It can be understood that the number of discharge sampling times can be reduced through load model fitting and incomplete sampling point calculation, and the electricity charge loss caused by sampling is reduced while load prediction is realized to support site battery standby power calculation and peak-staggering charge-discharge strategy calculation. And under the condition that only one period of complete data is sampled, the load prediction of the next period is carried out so as to start the functions of the station battery standby power calculation, peak-staggering charge-discharge strategy calculation and the like, and the station electric charge cost is reduced. And load prediction can be performed in a rolling way by combining load curve fitting with incomplete sampling point calculation without carrying out complete data sampling on each period, so that the requirement of load prediction on load data acquisition is greatly reduced. The load of the later period is predicted in the modes of data clustering, classification curve fitting, individual curve fitting and fusion prediction, so that the accurate prediction of the complete load of the second period can be realized under the condition that only one period is completely sampled, and the data quantity required to be accumulated for starting a prediction system is reduced. And supporting rolling prediction in scenes with different numbers of sampling points and time points of each period through fitting the relation between the time points and the load.

For example, taking a 300AH capacity battery as an example, taking a period of one day, the first day is completely sampled at 15min intervals, and then sampling is carried out 12 times a day, each time a discharge is carried out for 1 minute, the sampling life loss is only 1.4%, and the electricity fee loss is only 0.05%. Through actual data simulation test, test samples with load prediction errors within 10% of the incomplete sampled data can reach more than 90%, and compared with the method for predicting by using the complete data sampling, the method is reduced by about 5%, and the data input requirement of intelligent peak-shifting charge-discharge strategy solving of the battery is met.

2. Deployed at sites

The foregoing describes a flow of deploying the load prediction method provided by the present application to a control node, and a flow of deploying the method provided by the present application to a site may refer to fig. 12, as follows.

1201. The first cycle is a complete data sample.

1202. Data cleaning and complementation.

1203. If the data is stable, go to step 1208, if not, go to step 1204.

1204. And (5) fitting individual load curves and carrying out fusion prediction.

1205. If drift occurs, step 1206 is performed, and if not, step 1207 is performed.

1206. And (5) offset adjustment.

1207. The next period load prediction results.

1208. Mean value prediction is utilized.

1209. The next periodic incomplete sampling point is calculated.

1210. The next period is a non-complete data sample.

1211. And issuing a charging and discharging strategy.

The descriptions of steps 1201-12011 and steps 1001-1003, 1005-1012 are omitted here for the similar steps.

The differences between the deployment at the control node and the deployment at the site are mainly that: the control node may acquire the sampled data of multiple sites, so that the sampled data of multiple sites may be used, whereas a single site may generally only acquire the data acquired by itself, so that the data acquired by itself may be used to fit an individual load curve, i.e. a load variation curve of itself, without fitting a group load curve, and then the individual load curve may be used to predict the load of the single site in the next cycle.

Of course, in some possible scenarios, a single station may acquire information of other stations, and the steps of performing the station may refer to the foregoing steps 1001-1012, which are not described herein.

Therefore, in the embodiment of the application, for a single site, the cold start can be realized by using one period of sampling data, and the method can be suitable for a scene with small data quantity. And the complete sampling is carried out in the first period, the load prediction can be carried out in the subsequent period based on the data acquired in the previous period, the cold start can be realized in the scene with less data, and the method is suitable for the scene with less samples. And the number of sampling points in the subsequent period is smaller than that in the first period, so that the power consumption generated by sampling can be reduced.

The foregoing describes the flow of the load prediction method provided by the present application, and the following describes an apparatus for executing the load prediction method provided by the present application.

Referring to fig. 13, the present application provides a schematic structural diagram of a load prediction device.

The load prediction device can be applied to an electric system, the electric system comprises at least one station and power supply equipment, a battery is arranged in each station, the power supply equipment is used for providing charging power for at least one station, the load prediction device can be a control node which is arranged independently or one station, and the load prediction device can be arranged according to actual application scenes, and the application is not limited to the control node.

The load prediction apparatus may include:

an obtaining module 1301, configured to obtain first sampling data, where the first sampling data includes load power obtained by calculating electric energy consumed by discharge sampling in a first period by a first station, and the first period is any one period of a plurality of periods divided in advance;

the prediction module 1302 is configured to obtain load information of the first station in a second period according to the first sampling data and the historical sampling data, where the second period is a next period after the first period, the load information includes a power value of the first station in the second period, and the historical sampling data includes data obtained by sampling in a period before the first period;

The determining module 1303 is configured to determine a charging and discharging policy according to the load information, where the charging and discharging policy includes a policy for controlling the first station to perform charging and discharging in the second period.

In one possible implementation, the prediction module 1302 is specifically configured to: fitting according to the first sampling data and the historical sampling data to obtain a load prediction model; and acquiring load information of the first station in a second period according to the load prediction model, wherein the second period is the next period after the first period, and the load information comprises the power value of the first station in the second period.

In one possible implementation, the power consumption system includes a plurality of sites, and the prediction module 1302 is specifically configured to: fitting according to sampling data of a plurality of stations to obtain a group load curve, wherein the group load curve is used for representing load changes of the plurality of stations in a first period; and fitting an individual load curve according to the first sampling data and the historical sampling data, wherein the individual load curve represents the load change of the first station in the first period, and the group load curve and the individual load curve form a load prediction model.

In one possible implementation, the prediction module 1302 is specifically configured to: outputting a first predicted sequence of the first station in a second period according to the group load curve; outputting a second predicted sequence of the first station in a second period according to the individual load curve; and fusing the first prediction sequence and the second prediction sequence to obtain load information.

In one possible implementation, the prediction module 1302 is specifically configured to: if the first period is the first period of the plurality of periods, clustering the sampling data of the plurality of sites to obtain at least one category of data; and fitting the data of at least one category respectively to obtain a load curve corresponding to each category, wherein the group load curve comprises the load curve of each category in the at least one category.

In one possible implementation, the prediction module 1302 is specifically configured to: a burr sequence is screened out from the sampling data of a plurality of stations to obtain stable data, wherein the burr sequence is data with deviation between adjacent sampling points being larger than preset deviation; clustering the stable data to obtain data of at least one category; and determining the category corresponding to the burr sequence according to the distance between the burr sequence and at least one category.

In one possible implementation, after the prediction module 1302 obtains the load information of the first station in the second period according to the load prediction model, the prediction module 1302 is further configured to: acquiring an output value of a load prediction model at a first time point, and acquiring a load value of a first site at the first time point; and if the deviation between the output value and the load value is larger than the preset deviation value, the load information is adjusted, and the adjusted load information is obtained.

In one possible implementation, the prediction module 1302 is specifically configured to: if the first sampling data and the historical sampling data accord with preset conditions, acquiring load information of the first station in a second period according to the average value of the data included in the first sampling data and the historical sampling data; the preset condition comprises that a stable value is smaller than a first threshold value, and the stable value comprises at least one of variance, sample entropy, approximate entropy or fuzzy entropy of each time point in the first sampling data and the historical sampling data.

In one possible implementation, the obtaining module 1301 is specifically configured to: if the first period is the first period of the plurality of periods, sampling is carried out according to a preset first number of sampling points to obtain first sampling data; if the first period is not the first period of the plurality of periods, determining a second number of sampling points according to the load information obtained in the previous period, and sampling according to the second number of sampling points to obtain first sampling data, wherein the first number is larger than the second number.

In one possible implementation, during the sampling in the first period, the obtaining module 1301 is further configured to: when the battery of the first station is discharged, sampling is carried out according to a preset interval, and second sampling data are obtained; and adding the second sampling data to the first sampling data to obtain new first sampling data.

Referring to fig. 14, a schematic structural diagram of another electronic device provided by the present application is as follows.

The electronic device may include the aforementioned wearable device, terminal, vehicle, or the like, and may include a processor 1401, a memory 1402, and a transceiver 1403. The processor 1401 and memory 1402 are interconnected by wires. Wherein program instructions and data are stored in memory 1402.

The memory 1402 stores program instructions and data corresponding to the steps of fig. 6 to 12.

The processor 1401 is configured to perform the method steps performed by the first device or the electronic device as described in any of the embodiments of fig. 6-12.

A transceiver 1403 for performing the steps of receiving or transmitting data performed by the first device or the electronic device as described in any of the embodiments of fig. 6-12.

There is also provided in an embodiment of the present application a computer-readable storage medium having stored therein a program for generating a vehicle running speed, which when run on a computer, causes the computer to perform the steps in the method described in the embodiments shown in the foregoing fig. 6-12.

Alternatively, the aforementioned electronic device shown in fig. 14 is a chip.

The embodiment of the application also provides an electronic device, which may also be called a digital processing chip or a chip, wherein the chip comprises a processing unit and a communication interface, the processing unit obtains program instructions through the communication interface, the program instructions are executed by the processing unit, and the processing unit is used for executing the method steps executed by the electronic device shown in any embodiment of fig. 6-12.

The embodiment of the application also provides a digital processing chip. The digital processing chip has integrated therein circuitry and one or more interfaces for implementing the above-described processor 1401, or the functions of the processor 1401. When the memory is integrated into the digital processing chip, the digital processing chip may perform the method steps of any one or more of the preceding embodiments. When the digital processing chip is not integrated with the memory, the digital processing chip can be connected with the external memory through the communication interface. The digital processing chip implements the actions executed by the electronic device in the above embodiment according to the program codes stored in the external memory.

Embodiments of the present application also provide a computer program product which, when run on a computer, causes the computer to perform the steps performed by an electronic device in a method as described in the embodiments of fig. 6-12 above.

The electronic device provided by the embodiment of the application can be a chip, and the chip comprises: a processing unit, which may be, for example, a processor, and a communication unit, which may be, for example, an input/output interface, pins or circuitry, etc. The processing unit may execute the computer-executable instructions stored in the storage unit, so that the chip in the server performs the device search method described in the embodiment shown in fig. 6 to 12. Optionally, the storage unit is a storage unit in the chip, such as a register, a cache, etc., and the storage unit may also be a storage unit in the wireless access device side located outside the chip, such as a read-only memory (ROM) or other type of static storage device that may store static information and instructions, a random access memory (random access memory, RAM), etc.

In particular, the aforementioned processing unit or processor may be a central processing unit (central processing unit, CPU), a Network Processor (NPU), a graphics processor (graphics processing unit, GPU), a digital signal processor (digital signal processor, DSP), an application specific integrated circuit (application specific integrated circuit, ASIC) or field programmable gate array (field programmable gate array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. The general purpose processor may be a microprocessor or may be any conventional processor or the like.

The processor referred to in any of the foregoing may be a general purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits for controlling the execution of the programs of the methods of fig. 6-12 described above.

It should be further noted that the above-described apparatus embodiments are merely illustrative, and that the units described as separate units may or may not be physically separate, and that units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the embodiment of the device provided by the application, the connection relation between the modules represents that the modules have communication connection, and can be specifically implemented as one or more communication buses or signal lines.

From the above description of the embodiments, it will be apparent to those skilled in the art that the present application may be implemented by means of software plus necessary general purpose hardware, or of course by means of special purpose hardware including application specific integrated circuits, special purpose CPUs, special purpose memories, special purpose components, etc. Generally, functions performed by computer programs can be easily implemented by corresponding hardware, and specific hardware structures for implementing the same functions can be varied, such as analog circuits, digital circuits, or dedicated circuits. However, a software program implementation is a preferred embodiment for many more of the cases of the present application. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a readable storage medium, such as a floppy disk, a U-disk, a removable hard disk, a Read Only Memory (ROM), a random access memory (random access memory, RAM), a magnetic disk or an optical disk of a computer, etc., including several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to execute the method according to the embodiments of the present application.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product.

The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present application, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by a wired (e.g., coaxial cable, fiber optic, digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be stored by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a Solid State Disk (SSD)), or the like.

The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Finally, it should be noted that: the foregoing is merely illustrative embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily think about variations or substitutions within the technical scope of the present application, and the application should be covered. Therefore, the protection scope of the application is subject to the protection scope of the claims.

Claims

1. A load prediction method, applied to an electrical system, the electrical system including at least one site and a power supply device, each site having a battery disposed therein, the power supply device being configured to provide charging power to the at least one site, comprising:

acquiring first sampling data, wherein the first sampling data comprise load power obtained by calculating electric energy consumed by a first station through discharge sampling in a first period, and the first period is any period in a plurality of periods divided in advance;

acquiring load information of the first station in a second period according to the first sampling data and historical sampling data, wherein the second period is the next period after the first period, the load information comprises a power value of the first station in the second period, and the historical sampling data comprises data obtained by sampling in a period before the first period;

and determining a charging and discharging strategy according to the load information, wherein the charging and discharging strategy comprises a strategy for controlling the first station to execute charging and discharging in the second period.

2. The method of claim 1, wherein the obtaining load information of the first station during the second period based on the first sample data and the historical sample data comprises:

Fitting according to the first sampling data and the historical sampling data to obtain a load prediction model;

and acquiring load information of the first station in a second period according to the load prediction model, wherein the second period is the next period after the first period, and the load information comprises a power value of the first station in the second period.

3. The method of claim 2, wherein the power system includes a plurality of sites therein, and the fitting the load prediction model based on the first sampled data and the historical sampled data includes:

fitting according to the sampling data of the plurality of stations to obtain a group load curve, wherein the group load curve is used for representing the load change of the plurality of stations in the first period;

and fitting an individual load curve according to the first sampling data and the historical sampling data, wherein the individual load curve represents the load change of the first station in the first period, and the group load curve and the individual load curve form the load prediction model.

4. A method according to claim 3, wherein said obtaining load information of the first station during the second period according to the load prediction model comprises:

Outputting a first predicted sequence of the first station in the second period according to the group load curve;

outputting a second predicted sequence of the first station in the second period according to the individual load curve;

and fusing the first prediction sequence and the second prediction sequence to obtain the load information.

5. The method of claim 3 or 4, wherein the obtaining a population load curve comprises:

if the first period is the first period of the plurality of periods, clustering the sampling data of the plurality of sites to obtain at least one category of data;

and fitting the data of the at least one category respectively to obtain a load curve corresponding to each category, wherein the group load curve comprises the load curve of each category in the at least one category.

6. The method of claim 5, wherein clustering the sampled data for the plurality of sites to obtain at least one category of data comprises:

screening burr sequences from the sampling data of the multiple stations to obtain stable data, wherein the burr sequences are data with deviation from adjacent sampling points being larger than preset deviation;

Clustering the stable data to obtain data of at least one category;

the method further comprises the steps of:

and determining the category corresponding to the burr sequence according to the distance between the burr sequence and the at least one category.

7. The method of any of claims 2-6, wherein after the obtaining load information for the first station during a second period according to the load prediction model, the method further comprises:

acquiring an output value of the load prediction model at a first time point, and acquiring a load value of the first site at the first time point;

and if the deviation between the output value and the load value is larger than a preset deviation value, the load information is adjusted to obtain adjusted load information.

8. The method of claim 1, wherein the obtaining load information of the first station during the second period based on the first sample data and the historical sample data comprises:

if the first sampling data and the historical sampling data accord with preset conditions, acquiring load information of the first station in a second period according to the average value of the data included in the first sampling data and the historical sampling data;

The preset condition comprises that a stable value is smaller than a first threshold value, and the stable value comprises at least one of variance, sample entropy, approximate entropy or fuzzy entropy of each time point in the first sampling data and the historical sampling data.

9. The method of any of claims 1-8, wherein the acquiring the first sampled data comprises:

if the first period is the first period of the plurality of periods, sampling is carried out according to a preset first number of sampling points to obtain first sampling data;

if the first period is not the first period of the plurality of periods, determining a second number of sampling points according to the load information obtained in the previous period, and sampling according to the second number of sampling points to obtain the first sampling data, wherein the first number is larger than the second number.

10. The method according to any one of claims 1-9, wherein during the sampling of the first period, the method further comprises:

when the battery of the first station is discharged, sampling is carried out according to a preset interval, and second sampling data are obtained;

And adding the second sampling data to the first sampling data to obtain new first sampling data.

11. A load predicting apparatus, characterized by being applied to an electric power consumption system including at least one station and a power supply device, each station having a battery disposed therein, the power supply device being configured to supply charged electric power to the at least one station, comprising:

the device comprises an acquisition module, a sampling module and a sampling module, wherein the acquisition module is used for acquiring first sampling data, the first sampling data comprise load power obtained by electric energy calculation consumed by a first station through discharge sampling in a first period, and the first period is any period in a plurality of periods divided in advance;

the prediction module is used for obtaining load information of the first station in a second period according to the first sampling data and historical sampling data, the second period is the next period after the first period, the load information comprises a power value of the first station in the second period, and the historical sampling data comprises data obtained by sampling in a period before the first period;

and the determining module is used for determining a charging and discharging strategy according to the load information, wherein the charging and discharging strategy comprises a strategy for controlling the first station to execute charging and discharging in the second period.

12. The apparatus according to claim 11, wherein the prediction module is specifically configured to:

13. The device according to claim 12, wherein the power consumption system comprises a plurality of stations, and the prediction module is specifically configured to:

14. The apparatus according to claim 13, wherein the prediction module is specifically configured to:

15. The apparatus according to claim 13 or 14, wherein the prediction module is specifically configured to:

16. The apparatus according to claim 15, wherein the prediction module is specifically configured to:

clustering the stable data to obtain data of at least one category;

17. The apparatus of any of claims 12-16, wherein after the prediction module obtains the load information of the first station for a second period according to the load prediction model, the prediction module is further configured to:

18. The apparatus according to claim 11, wherein the prediction module is specifically configured to:

19. The apparatus according to any one of claims 11-18, wherein the acquisition module is specifically configured to:

20. The apparatus of any one of claims 11-19, wherein, during sampling of the first period, the acquisition module is further configured to:

21. A load predicting device comprising a processor coupled to a memory, the memory storing a program that when executed by the processor, implements the method of any one of claims 1 to 10.

22. A computer readable storage medium comprising a program which, when executed by a processing unit, performs the method of any of claims 1 to 10.

23. A load predicting device comprising a processing unit and a communication interface, the processing unit obtaining program instructions via the communication interface, the program instructions, when executed by the processing unit, implementing the method of any one of claims 1 to 10.

24. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the steps of the method of any of claims 1 to 10.