CN116205382B

CN116205382B - Prediction method and prediction device for electricity consumption, electronic device and electronic equipment

Info

Publication number: CN116205382B
Application number: CN202310494933.XA
Authority: CN
Inventors: 赵云; 肖勇; 蔡梓文; 陆煜锌; 王浩林
Original assignee: CSG Electric Power Research Institute
Current assignee: CSG Electric Power Research Institute
Priority date: 2023-05-05
Filing date: 2023-05-05
Publication date: 2023-08-22
Anticipated expiration: 2043-05-05
Also published as: CN116205382A

Abstract

The application provides a prediction method and a prediction device for electricity consumption, an electronic device and electronic equipment. The method comprises the following steps: acquiring historical electricity consumption data and environment information corresponding to the historical electricity consumption data to obtain first historical electricity consumption data, decomposing the first historical electricity consumption data by using an aggregate empirical mode decomposition method, and at least obtaining multi-order electricity consumption data components; generating each stage of electricity utilization data component into an input data set, and inputting the input data set into an LSTM neural network corresponding to each stage of electricity utilization data component so as to process the input data set; and obtaining each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, adding at least a plurality of power consumption prediction results to obtain a total power consumption prediction result in a preset time period in the future, and adjusting the power supply quantity of the power supply system according to the total power consumption prediction result in the preset time period in the future. The application solves the problem of inaccurate prediction of the electricity consumption.

Description

Prediction method and prediction device for electricity consumption, electronic device and electronic equipment

Technical Field

The present application relates to the field of electricity consumption prediction, and in particular, to a method for predicting electricity consumption, a prediction apparatus, a computer-readable storage medium, an electronic apparatus, and an electronic device.

Background

The stable supply of electric energy is a necessary guarantee for the development of the current society economy and the normal life of people, and the accurate electric quantity prediction can provide reliable guidance for electric quantity production and power supply scheduling, so that the power supply quality of an electric power system is improved.

For the power consumption data, the time sequence distribution has certain statistical characteristics, firstly, the power consumption data is superposition of comprehensive power consumption of all industries, and for industrial power consumption, although different areas have different power consumption, the industrial power consumption in a certain area usually has certain periodicity by taking the circumference as a period, namely, the power consumption in a working day is relatively stable, and the power consumption in a weekend is obviously reduced; for domestic electricity, the periodicity is relatively unobvious, and the electricity consumption is related to holidays, weather conditions and the like, and generally presents larger fluctuation; therefore, after the power consumption conditions of industrial power consumption, resident power consumption, commercial power consumption, government institutions, public institutions and the like are superposed, the total power consumption in the area has certain periodicity and certain volatility, and the power consumption prediction task is challenged.

At present, automatic prediction of electricity consumption by means of a machine learning method is a main stream direction of related research, and conventional electricity consumption prediction methods include regression analysis, markov models, support vector regression, time sequence analysis models and the like. With the development of artificial intelligence algorithms, machine learning models (such as a multiple-output support vector regression machine and a neural network) are applied to the task, with the development of machine learning, for power prediction, a multi-scale prediction model is often used to extract time series features so as to improve prediction accuracy, and wavelet analysis and fourier transform are also often used for the prediction of electricity consumption. However, the construction of the wavelet functions is relatively complex, which would impose additional computational burden. The learner predicts the power consumption at different time scales after decomposing the sequence data. However, these electricity consumption prediction models are difficult to capture the fluctuation of data while fitting periodicity, so that the prediction result of electricity consumption is inaccurate, reliable guidance cannot be provided for electricity production and power supply scheduling, and the power supply quality of a power system cannot be guaranteed.

Therefore, a method capable of analyzing the periodicity and fluctuation of the electricity consumption to accurately predict the electricity consumption is needed.

Disclosure of Invention

The application provides a prediction method, a prediction device, a computer readable storage medium, an electronic device and an electronic device for predicting electricity consumption, which at least solve the problem of inaccurate electricity consumption prediction in the prior art.

In order to achieve the above object, according to one aspect of the present application, there is provided a method for predicting an amount of electricity consumption, comprising: acquiring historical electricity consumption data and environment information corresponding to the historical electricity consumption data to obtain first historical electricity consumption data, decomposing the first historical electricity consumption data by utilizing an aggregate empirical mode decomposition method to at least obtain multi-order electricity consumption data components, wherein the time change periods of any two-order electricity consumption data components are different, and the environment information at least comprises weather conditions, electricity consumption areas and electricity consumption time; generating the power utilization data components of each order to an input data set, and inputting the input data set into an LSTM neural network corresponding to the power utilization data components of each order to process the input data set; and obtaining each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, at least adding a plurality of power consumption prediction results to obtain a total power consumption prediction result in a future preset time period, and adjusting the power supply quantity of a power supply system according to the total power consumption prediction result in the future preset time period, wherein the future preset time period is a time period which is after and adjacent to a historical time period corresponding to the historical power consumption data.

Optionally, decomposing the first historical electricity consumption data by using an aggregate empirical mode decomposition method to obtain at least a multi-order electricity consumption data component, including: according to the formulaDecomposing the first historical electricity data by using the aggregate empirical mode decomposition method to obtain a multi-order electricity data component and a residual component, wherein the residual component is a difference value of the sum of the first historical electricity data and the multi-order electricity data component, J is the total order of the electricity data component, x (t) is the first historical electricity data, C _j (t) is the electricity data component of the j th order, r _J And (t) is the residual component.

Optionally, after obtaining the multi-order power consumption data component and the residual component, the method includes: inputting the residual components into a corresponding LSTM neural network to process the residual components; and obtaining a prediction result output by the LSTM neural network corresponding to the residual component, and obtaining a residual prediction result.

Optionally, before inputting the input data set into the LSTM neural network corresponding to each order of the power-using data component, the method further includes: acquiring electricity consumption data before the time corresponding to the first historical electricity consumption data and environment information corresponding to the electricity consumption data, and obtaining second historical electricity consumption data; and training the initial LSTM neural network by taking the second historical electricity utilization data as input data of the initial LSTM neural network and taking the first historical electricity utilization data as output data of the initial LSTM neural network, so as to obtain the LSTM neural network after training is completed.

Optionally, inputting the input data set into the LSTM neural network corresponding to each order of the power consumption data component, so as to process the input data set, including: taking each stage of the power consumption data component as one data node of the LSTM neural network, and calculating the correlation between two adjacent data nodes in the data nodes to obtain a plurality of first correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data component; fitting a plurality of first correlation parameters to obtain first periodic parameters of the first historical electricity utilization data, wherein the first periodic parameters represent the time change period of the first historical electricity utilization data after the superposition of the two-order electricity utilization data components.

Optionally, inputting the input data set into the LSTM neural network corresponding to each order of the power consumption data component, so as to process the input data set, and further including: taking each order of the power consumption data components as one data node of the LSTM neural network, taking at least two adjacent data nodes as node subgroups, and calculating correlation between the two adjacent node subgroups to obtain a plurality of second correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data components; fitting a plurality of second correlation parameters to obtain second periodic parameters of the first historical electricity utilization data, wherein the second periodic parameters represent the time change period of the first historical electricity utilization data after the fourth-order electricity utilization data components are overlapped.

Optionally, at least adding the plurality of electricity consumption prediction results to obtain a total electricity consumption prediction result of a predetermined future time period, including: and adding the plurality of electricity consumption prediction results and the residual prediction result to obtain the total electricity consumption prediction result in a preset future time period.

Optionally, before decomposing the first historical electricity consumption data by using the ensemble empirical mode decomposition method, the method further includes: normalizing the first historical electricity consumption data; and carrying out noise removal processing on the normalized first historical electricity utilization data.

According to another aspect of the present application, there is provided a power consumption prediction apparatus including: the system comprises a decomposition unit, a first power consumption unit and a second power consumption unit, wherein the decomposition unit is used for acquiring historical power consumption data and environment information corresponding to the historical power consumption data to obtain first historical power consumption data, decomposing the first historical power consumption data by utilizing an integrated empirical mode decomposition method to at least obtain multi-order power consumption data components, the time change periods of the multi-order power consumption data components are different, and the environment information at least comprises weather conditions, power consumption areas and power consumption time; the input unit is used for generating the power utilization data components of each order into an input data set, and inputting the input data set into an LSTM neural network corresponding to the power utilization data components of each order so as to process the input data set; and the adjusting unit is used for acquiring each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, adding at least a plurality of power consumption prediction results to obtain a power consumption prediction result in a preset time period in the future, and adjusting the power supply quantity of the power supply system according to the power consumption prediction result in the preset time period in the future.

According to still another aspect of the present application, there is provided a computer readable storage medium including a stored program, wherein the program, when executed, controls a device in which the computer readable storage medium is located to perform any one of the above-described prediction methods.

According to a further aspect of the application there is provided an electronic device comprising a memory having a computer program stored therein and a processor arranged to execute any one of the prediction methods by means of the computer program.

According to still another aspect of the present application, there is provided an electronic apparatus including: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any one of the prediction methods.

According to the technical scheme, first historical electricity utilization data are decomposed into multiple-order electricity utilization data components with different time change periods, each-order electricity utilization data component is used as an input data set to be input into an LSTM neural network corresponding to each-order electricity utilization data component trained in advance, the LSTM neural network is used for processing each-order electricity utilization data component, the LSTM neural network outputs electricity utilization prediction results corresponding to each-order electricity utilization data component, and then the electricity utilization prediction results corresponding to each electricity utilization data component are added to obtain total electricity utilization prediction results in a preset time period in the future. Compared with the method for predicting the electricity consumption in the future preset time period directly according to the historical electricity consumption data in the prior art, the method provided by the application can decompose the first historical electricity consumption data into the multi-order electricity consumption data components, analyze the periodic variation characteristics and the fluctuation variation characteristics of the first historical electricity consumption data through the multi-order electricity consumption data components, predict the electricity consumption in each corresponding future preset time period according to each order of electricity consumption data components, superimpose a plurality of electricity consumption prediction results to obtain the total electricity consumption prediction result in the future preset time period, and adjust the power supply quantity of the power supply system according to the total electricity consumption prediction result. Therefore, the problem of inaccurate electricity consumption prediction results in the prior art can be solved, and the effect of accurately predicting the electricity consumption is achieved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

fig. 1 is a block diagram showing a hardware configuration of a mobile terminal for performing a power consumption prediction method according to an embodiment of the present application;

fig. 2 is a schematic flow chart of a method for predicting electricity consumption according to an embodiment of the present application;

FIG. 3 is a schematic diagram of an LSTM neural network model in a method for predicting power consumption according to an embodiment of the present application;

FIG. 4 is a block diagram showing a prediction flow of an LSTM neural network model in a method for predicting power consumption according to an embodiment of the present application;

fig. 5 shows a block diagram of a power consumption prediction apparatus according to an embodiment of the present application.

Wherein the above figures include the following reference numerals:

102. a processor; 104. a memory; 106. a transmission device; 108. and an input/output device.

Detailed Description

It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other. The application will be described in detail below with reference to the drawings in connection with embodiments.

In order that those skilled in the art will better understand the present application, a technical solution in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, shall fall within the scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures 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 in order to describe the embodiments of the application 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.

For convenience of description, the following will describe some terms or terminology involved in the embodiments of the present application:

EEMD: the method is characterized by integrating empirical mode decomposition (Ensemble Empirical Mode Decomposition), a time-frequency analysis method is used for carrying out signal decomposition according to time scale features of signals, decomposing an original signal into a plurality of eigenmode function (IMF) components, and each IMF component represents a local feature signal of the original signal in different time scales and is suitable for time-frequency analysis of nonlinear and non-stationary signal sequences.

LSTM: the long-term and short-term memory network is a time-circulating neural network and is proposed for solving the long-term dependence problem existing in the common RNN (circulating neural network).

As described in the background art, in the prior art, the power consumption data has larger fluctuation, and the inaccurate prediction result of the power consumption leads to the reduced power supply quality, so as to solve the problem of inaccurate power consumption prediction.

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application.

The method embodiments provided in the embodiments of the present application may be performed in a mobile terminal, a computer terminal or similar computing device. Taking the operation on a mobile terminal as an example, fig. 1 is a block diagram of a hardware structure of a mobile terminal according to an embodiment of the present application. As shown in fig. 1, a mobile terminal may include one or more (only one is shown in fig. 1) processors 102 (the processor 102 may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 for communication functions and an input-output device 108. It will be appreciated by those skilled in the art that the structure shown in fig. 1 is merely illustrative and not limiting of the structure of the mobile terminal described above. For example, the mobile terminal may also include more or fewer components than shown in fig. 1, or have a different configuration than shown in fig. 1.

The memory 104 may be used to store a computer program, for example, a software program of application software and a module, such as a computer program corresponding to a method for predicting electricity consumption in an embodiment of the present application, and the processor 102 executes the computer program stored in the memory 104 to perform various functional applications and data processing, that is, implement the above-mentioned method. Memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located relative to the processor 102, which may be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, simply referred to as NIC) that can connect to other network devices through a base station to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is configured to communicate with the internet wirelessly.

In the present embodiment, a method for predicting the amount of electricity used to run on a mobile terminal, a computer terminal, or a similar computing device is provided, it should be noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowchart, in some cases, the steps illustrated or described may be performed in an order different from that herein.

Fig. 2 is a flowchart of a method for predicting power consumption according to an embodiment of the present application. As shown in fig. 2, the method comprises the steps of:

step S201, acquiring historical electricity consumption data and environment information corresponding to the historical electricity consumption data to obtain first historical electricity consumption data, decomposing the first historical electricity consumption data by utilizing an aggregate empirical mode decomposition method to at least obtain multi-order electricity consumption data components, wherein the time change periods of any two orders of the electricity consumption data components are different, and the environment information at least comprises weather conditions, electricity consumption areas and electricity consumption time;

specifically, under different industries, different electricity utilization areas and different weather conditions, the electricity utilization data of the user are different, and have different time change periods. For example: for industrial electricity, although different areas have different electricity consumption, the industrial electricity in a certain area is generally provided with a certain periodicity by taking the circumference as a period, namely the electricity consumption in a working day is relatively stable, the electricity consumption in a weekend is obviously reduced, and certain periodicity characteristics are presented; for domestic electricity, the periodicity is relatively unobvious, and the electricity consumption is related to holidays, weather conditions and the like, and generally presents larger fluctuation; therefore, the total power consumption in the area has a certain periodicity and a certain volatility after the power consumption conditions of industrial power consumption, residential power consumption, commercial power consumption, government institutions, institutions and the like are superimposed. Namely, the time sequence characteristics of the electricity consumption data are related to factors such as weather conditions, electricity consumption areas and electricity consumption time, so that the application firstly obtains the electricity consumption data, the weather conditions, the electricity consumption areas and the electricity consumption time corresponding to the electricity consumption data and takes the electricity consumption data as first historical electricity consumption data. Since the conventional machine learning method does not analyze volatility while analyzing periodicity, the present application decomposes the first historical electricity usage data using an ensemble empirical mode decomposition method (EEMD) to decompose the first historical electricity usage data into a plurality of electricity usage data components (IMFs) having different variation periods in time, thereby analyzing the volatility of the electricity usage data using the high frequency IMF components and analyzing the periodicity of the electricity usage data using the low frequency IMF components. For example: the electricity consumption of the industrial area is greatly influenced by working days and non-working days, so the week is taken as a time change period, the electricity consumption of the residential area is greatly influenced by weather conditions, and particularly, the electricity consumption of an air conditioner is large due to the change of temperatures in different seasons (the high temperature in summer), so the year is taken as a time change period. Of course, the above illustrated case is a longer time scale example, and after the historical electricity data is decomposed by the EEMD, a minute time variation period can be captured.

Step S202, generating the electricity consumption data components of each order into an input data set, and inputting the input data set into an LSTM neural network corresponding to the electricity consumption data components of each order so as to process the input data set;

specifically, after the first historical electricity utilization data is decomposed into multiple-order electricity utilization data components, each-order electricity utilization data component is input into a pre-trained LSTM neural network, the pre-trained LSTM neural network predicts each-order electricity utilization data component, and a corresponding prediction result is output.

Step S203, obtaining each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, adding at least a plurality of the power consumption prediction results to obtain a total power consumption prediction result in a future predetermined period, and adjusting a power supply amount of a power supply system according to the total power consumption prediction result in the future predetermined period, where the future predetermined period is a period that is after and adjacent to a historical period corresponding to the historical power consumption data.

Specifically, after the LSTM neural network outputs each power consumption prediction result corresponding to each power consumption data component, since each power consumption data component is obtained by decomposing the first historical power consumption data, a plurality of power consumption prediction results are added to obtain a total power consumption prediction result. The purpose of predicting the electricity consumption data of the future predetermined period of time from the historical electricity consumption data can be achieved, for example: to predict the power consumption of twenty-four hours in the open day, the weather conditions, the power consumption area, and the power consumption time corresponding to the power consumption data within one month before today and the power consumption data within one month before today may be obtained as the first historical power consumption data, and the weather conditions, the power consumption area, and the power consumption time corresponding to the power consumption data within one month before the present week may be obtained as the first historical power consumption data, so that the historical time period and the future predetermined time period may be any reasonable time period, and the present application does not specifically limit the relationship between the range of the above-described historical time period, the range of the future predetermined time period, and the range of the historical time period and the range of the future time period.

According to the embodiment, first historical electricity utilization data are decomposed into multiple-order electricity utilization data components with different time change periods, each-order electricity utilization data component is used as an input data set to be input into an LSTM neural network corresponding to each-order electricity utilization data component trained in advance, the LSTM neural network is used for processing each-order electricity utilization data component, the LSTM neural network outputs electricity utilization prediction results corresponding to each-order electricity utilization data component, and then the electricity utilization prediction results corresponding to each electricity utilization data component are added to obtain total electricity utilization prediction results in a preset time period in the future. Compared with the method for predicting the electricity consumption in the future preset time period directly according to the historical electricity consumption data in the prior art, the method provided by the application can decompose the first historical electricity consumption data into the multi-order electricity consumption data components, analyze the periodic variation characteristics and the fluctuation variation characteristics of the first historical electricity consumption data through the multi-order electricity consumption data components, predict the electricity consumption in each corresponding future preset time period according to each order of electricity consumption data components, superimpose a plurality of electricity consumption prediction results to obtain the total electricity consumption prediction result in the future preset time period, and adjust the power supply quantity of the power supply system according to the total electricity consumption prediction result. Therefore, the problem of inaccurate electricity consumption prediction results in the prior art can be solved, and the effect of accurately predicting the electricity consumption is achieved.

In a specific implementation process, the step S201 may be implemented by the following steps: according to the formulaDecomposing the first historical electricity data by using the aggregate empirical mode decomposition method to obtain a multi-order electricity data component and a residual component, wherein the residual component is a difference value of the sum of the first historical electricity data and the multi-order electricity data component, J is the total order of the electricity data component, x (t) is the first historical electricity data, and C _j (t) is the j-th order of the above-mentioned electricity consumption data component, r _J And (t) is the residual component described above. According to the method, the first historical electricity utilization data is decomposed into the multi-order electricity utilization data components and residual components, so that the time change period of each-order electricity utilization data component can be analyzed, the time change rule of the first historical electricity utilization data is comprehensively analyzed, and the electricity utilization quantity is better predicted.

Specifically, since the integrated empirical mode decomposition method can decompose the signal into components of different frequencies, in the above-mentioned power consumption decomposition process, the first historical power consumption data is decomposed into power consumption data components of different frequencies, i.e., different time variation periods, and each stage of power consumption data components uses C _j And (t) expressing that after the decomposition is completed, subtracting the sum of the multi-order power consumption data components from the first historical power consumption data, and obtaining a result which is a residual component.

In order to comprehensively analyze the time-varying period of the first historical electricity consumption data, the step S201 of the present application further includes the steps of: inputting the residual components into a corresponding LSTM neural network to process the residual components; the method processes the residual components through the LSTM neural network and obtains the prediction results of the residual components, so that the electricity consumption of a preset time period in the future can be comprehensively predicted according to the first historical electricity consumption data, and inaccurate electricity consumption prediction caused by omission of the first historical electricity consumption data is avoided.

Specifically, the first historical electricity consumption data is decomposed into multiple-order electricity consumption data components by the integrated empirical mode decomposition method, and after the first historical electricity consumption data is decomposed, the data in the remaining first historical electricity consumption data cannot be decomposed any more, which is called a residual component, so that the residual component does not have an obvious time change period, and the residual component is input into an LSTM neural network corresponding to the residual component, so that an electricity consumption prediction result corresponding to the residual component is obtained.

Before the step S202, the method further includes: acquiring electricity consumption data before the time corresponding to the first historical electricity consumption data and environment information corresponding to the electricity consumption data, and obtaining second historical electricity consumption data; and training the initial LSTM neural network by taking the second historical electricity utilization data as input data of the initial LSTM neural network and taking the first historical electricity utilization data as output data of the initial LSTM neural network, so as to obtain the LSTM neural network after training is completed. According to the method, the initial LSTM neural network is trained according to the second historical electricity consumption data, so that parameters of the LSTM neural network can be obtained, and the prediction of electricity consumption is realized.

Specifically, in order to train the LSTM neural network model, first, the electricity data before the time corresponding to the first historical electricity data and the environmental information corresponding to the electricity data are obtained, and are used as the second historical electricity data, that is, the input data of the initial LSTM neural network, it is to be noted that the second historical electricity data is decomposed by the method of decomposing the aggregate empirical mode to obtain multiple-order electricity data components, the multiple-order electricity data components corresponding to the second historical electricity data are used as the input of the initial LSTM neural network model, the first historical electricity data is used as the output of the initial LSTM neural network model, the initial LSTM neural network model is trained, the parameter values, such as weight and bias, of the neural network model are obtained, and the prediction of the electricity consumption is realized based on the LSTM neural network after the training is completed.

In order to analyze the characteristic that the time period of the first historical electricity consumption data is longer or has no obvious time period, that is, the fluctuation characteristic, in some embodiments, the step S202 may be specifically implemented by the following steps: taking the electricity consumption data component of each step as one data node of the LSTM neural network, and calculating the correlation between two adjacent data nodes in the data nodes to obtain a plurality of first correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the electricity consumption data component; fitting a plurality of the first correlation parameters to obtain a first periodic parameter of the first historical electricity utilization data, wherein the first periodic parameter represents the time change period of the first historical electricity utilization data after the superposition of the two-order electricity utilization data components.

Specifically, the 1-scale layer of the application runs on each hidden node of the LSTM neural network, extracts the characteristics of adjacent nodes, and obtains a plurality of first correlation parameters for extracting the characteristics of high-order IMF components with obvious periodicity. The correlation parameters are extracted by the neural network and represented by the output of the neural network in the form of activated features, in particular, assuming that the input part is x, the 1-scale layer parameter mapping process is represented in the form of lstm_1 () function, wherein the parameters of lstm_1 () i.e. the connection weights of the neural network, the process of computing from the input part to the features takes the form of full connection, the size span of which is 1, whereby the first correlation parameter is represented in the form of y=lstm_1 (x).

In some embodiments, the step S202 may be specifically implemented by the following steps: taking each stage of the electricity consumption data component as a data node of the LSTM neural network, taking at least two adjacent data nodes as node subgroups, and calculating the correlation between the two adjacent node subgroups to obtain a plurality of second correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the electricity consumption data component; fitting a plurality of second correlation parameters to obtain a second periodic parameter of the first historical electricity utilization data, wherein the second periodic parameter represents the time change period of the first historical electricity utilization data after the fourth-order electricity utilization data components are overlapped. The method calculates the correlation between two adjacent node groups, so that the characteristic that the time change period in the first historical electricity utilization data is longer or has no obvious time change period, namely the fluctuation characteristic, can be further analyzed.

In particular, for the low order IMF component, its ripple and approximate periodicity can be regarded as a superposition of signals having different periodicity. Thus, the present invention considers multiple nodes simultaneously and calculates the correlation between them by feature extraction to capture periodicity on different scales, and the 2-scale MLSTM calculates features by considering every two neighboring nodes. For prediction of low-order IMF components, the correlation parameters are also extracted by the neural network, the output of the neural network is represented as the form of activated features, unlike the 1-scale layer, for 2-scale MLSTM, the input part is assumed to be x, the 2-scale layer parameter mapping process is represented as lstm_2 () function form, wherein the parameters of lstm_2 () are the connection weights of the neural network, and the process from the input part to the feature calculation adopts the form of cross-scale connection, the size span of which is 2, i.e. two nodes crossing the input are calculated each time. Thus, the second correlation parameter is expressed in the form y=lstm_2 (x).

In order to obtain an accurate electricity consumption prediction result, in some embodiments, the step S203 may be specifically implemented by the following steps: and adding the plurality of electricity consumption prediction results and the residual prediction result to obtain the total electricity consumption prediction result in a preset future time period. The method adds the multiple power consumption prediction results obtained according to the multi-order power consumption data component prediction and the residual prediction results, so that the integrity of the power consumption prediction results can be ensured.

Specifically, because the first historical electricity consumption data is decomposed into the multi-order electricity consumption data component and the residual error component, the total electricity consumption prediction result needs to be added with the electricity consumption prediction result obtained according to the multi-order electricity consumption data component and the residual error prediction result obtained according to the residual error component prediction, so that the total electricity consumption prediction result corresponding to the first historical electricity consumption data is obtained, and the integrity of the electricity consumption prediction result is ensured.

In some embodiments, the step S201 further includes the steps of: normalizing the first historical electricity utilization data before decomposing the first historical electricity utilization data by using an aggregate empirical mode decomposition method; and carrying out noise removal processing on the normalized first historical electricity consumption data. The method is used for preprocessing the data, so that the data with larger errors in the data can be removed, and inaccurate prediction of the power consumption is avoided.

Specifically, in the practical application process, the first historical electricity consumption data can be regularized by using a normal function to extract features by using an LSTM neural network, then the first historical electricity consumption data is divided into data sequences by taking 7 days as a unit, simple preprocessing is performed on the data sequences, obvious noise data are screened out and removed so as to learn time distribution features of the first historical electricity consumption data, the rest data are partially marked, and the weather condition, the electricity consumption area and the electricity consumption time are spliced to the historical electricity consumption data in a vector form to form the first historical electricity consumption data.

In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the power consumption prediction method of the present application will be described in detail below with reference to specific embodiments.

The embodiment relates to a specific electricity consumption prediction method, as shown in fig. 3 and fig. 4, including the following steps:

step S1: acquiring historical electricity consumption data and environment information corresponding to the historical electricity consumption data, obtaining first historical electricity consumption data, and preprocessing the first historical electricity consumption data: normalizing the first historical power consumption data by using a normal function, screening out obvious noise data and removing the noise data;

step S2: adopting a set empirical mode decomposition method for the first historical electricity utilization dataThe decomposition is carried out by the method (EEMD) as follows:wherein the residual component is the difference between the first historical electricity consumption data and the sum of the multiple orders of the electricity consumption data components, J is the total order of the electricity consumption data components, x (t) is the first historical electricity consumption data, and C _j (t) is the j-th order of the above-mentioned electricity consumption data component, r _J (t) is the residual component;

step S3: fig. 3 is a schematic diagram of a multi-scale LSTM neural network model in a specific method for predicting electricity consumption according to the present application, and as shown in fig. 3, the training process of the initial LSTM neural network model is as follows: acquiring second historical electricity utilization data, decomposing the second historical electricity utilization data according to the step S1 and the step S2, taking each-order electricity utilization data component as one node of the LSTM neural network, calculating the correlation between two adjacent data nodes, extracting the characteristics of the adjacent nodes, wherein the characteristic extraction layer of the 1-scale characteristic extraction can be expressed as h ₀ 、h ₁ 、h ₂ 、h ₃ 、h ₄ 、…、h _N A plurality of first correlation parameters are denoted as x ₁ 、x ₂ 、x ₃ 、x ₄ 、…、x _N According to the formulaFeature extraction of feature extraction layer for 1-scale feature extraction,/a>Represents x ₁ 、x ₂ 、x ₃ 、x ₄ 、…、x _N ；

Step S4: using two adjacent data nodes as node subgroups, calculating the correlation between the two adjacent node subgroups, extracting the characteristics of the two adjacent nodes, and expressing the characteristic extraction process of 2-scale characteristic extraction as h ₀ 、（h ₁ 、h ₂ ）、（h ₃ 、h ₄ ）、…、h _N According to the formulaFeature extraction of feature extraction layer for 2-scale feature extraction, < >>Represents x ₁ 、x ₂ 、x ₃ 、x ₄ 、…、x _N Training the LSTM neural network, wherein the LSTM neural network after training is used for predicting the subsequent power consumption;

step S5: after decomposing the first historical electricity consumption data into multiple-order electricity consumption data components, inputting the LSTM neural network after the training is completed and outputting a prediction result of electricity consumption, wherein a prediction flow chart is shown in fig. 4, and the electricity consumption data (the first historical electricity consumption data) is decomposed by EEMD to obtain J-order IMF components which are respectively expressed as 1 ^th IMF-order component (first-order electrical data component), 2 ^th IMF-order component (second-order electrical data component), …, J ^th -inputting the IMF-order components (J-th-order power consumption data components) and residual terms (residual components) and additional information (weather conditions, power consumption areas and power consumption time corresponding to historical power consumption data) into corresponding MLSTM models (LSTM neural network models) to obtain power consumption prediction results corresponding to the power consumption data components of each order, expressed as 1 ^st -prediction value of IMF of order, 2 ^st Predicted value of IMF of order, …, J ^st -predicting values of IMF of order (power consumption predicting result) and predicting values of residuals (residual predicting result), adding the J predicting values and the residual predicting values to obtain a total power consumption predicting result of a predetermined time period in the future, and adjusting the power supply amount of the power supply system according to the total power consumption predicting result.

The embodiment of the application also provides a device for predicting the electricity consumption, and the device for predicting the electricity consumption can be used for executing the method for predicting the electricity consumption provided by the embodiment of the application. The device is used for realizing the above embodiments and preferred embodiments, and is not described in detail. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. While the means described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.

The following describes a device for predicting electricity consumption provided by an embodiment of the present application.

Fig. 5 is a schematic diagram of a prediction apparatus for electricity consumption according to an embodiment of the present application. As shown in fig. 5, the apparatus includes:

A decomposition unit 10, configured to obtain historical electricity consumption data and environmental information corresponding to the historical electricity consumption data, obtain first historical electricity consumption data, decompose the first historical electricity consumption data by using an aggregate empirical mode decomposition device, and at least obtain multiple-order electricity consumption data components, where time variation periods of the multiple-order electricity consumption data components are different, and the environmental information at least includes weather conditions, electricity consumption areas and electricity consumption time;

specifically, under different industries, different electricity utilization areas and different weather conditions, the electricity utilization data of the user are different, and have different time change periods. For example: for industrial electricity, although different areas have different electricity consumption, the industrial electricity in a certain area is generally provided with a certain periodicity by taking the circumference as a period, namely the electricity consumption in a working day is relatively stable, the electricity consumption in a weekend is obviously reduced, and certain periodicity characteristics are presented; for domestic electricity, the periodicity is relatively unobvious, and the electricity consumption is related to holidays, weather conditions and the like, and generally presents larger fluctuation; therefore, the total power consumption in the area has a certain periodicity and a certain volatility after the power consumption conditions of industrial power consumption, residential power consumption, commercial power consumption, government institutions, institutions and the like are superimposed. Namely, the time sequence characteristics of the electricity consumption data are related to factors such as weather conditions, electricity consumption areas and electricity consumption time, so that the application firstly obtains the electricity consumption data, the weather conditions, the electricity consumption areas and the electricity consumption time corresponding to the electricity consumption data and takes the electricity consumption data as first historical electricity consumption data. Since the conventional machine learning apparatus does not analyze the volatility while analyzing the periodicity, the present application decomposes the first historical electricity usage data using an ensemble empirical mode decomposition apparatus (EEMD) to decompose the first historical electricity usage data into a plurality of electricity usage data components (IMFs) having different variation periods in time, thereby analyzing the volatility of the electricity usage data using the high frequency IMF components and analyzing the periodicity of the electricity usage data using the low frequency IMF components. For example: the electricity consumption of the industrial area is greatly influenced by working days and non-working days, so the week is taken as a time change period, the electricity consumption of the residential area is greatly influenced by weather conditions, and particularly, the electricity consumption of an air conditioner is large due to the change of temperatures in different seasons (the high temperature in summer), so the year is taken as a time change period. Of course, the above illustrated case is a longer time scale example, and after the historical electricity data is decomposed by the EEMD, a minute time variation period can be captured.

An input unit 20 for generating the power consumption data component of each order into an input data set, and inputting the input data set into the LSTM neural network corresponding to the power consumption data component of each order to process the input data set;

And the adjusting unit 30 is configured to obtain each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, add at least a plurality of the power consumption prediction results to obtain a power consumption prediction result in a predetermined future time period, and adjust the power supply amount of the power supply system according to the power consumption prediction result in the predetermined future time period.

According to the embodiment, first historical electricity utilization data are decomposed into multiple-order electricity utilization data components with different time change periods, each-order electricity utilization data component is used as an input data set to be input into an LSTM neural network corresponding to each-order electricity utilization data component trained in advance, the LSTM neural network is used for processing each-order electricity utilization data component, the LSTM neural network outputs electricity utilization prediction results corresponding to each-order electricity utilization data component, and then the electricity utilization prediction results corresponding to each electricity utilization data component are added to obtain total electricity utilization prediction results in a preset time period in the future. Compared with the device for predicting the electricity consumption in the future preset time period directly according to the historical electricity consumption data in the prior art, the device provided by the application can decompose the first historical electricity consumption data into the multi-order electricity consumption data components, analyze the periodic variation characteristics and the fluctuation variation characteristics of the first historical electricity consumption data through the multi-order electricity consumption data components, predict the electricity consumption in each corresponding future preset time period according to each order of electricity consumption data components, superimpose a plurality of electricity consumption prediction results to obtain the total electricity consumption prediction result in the future preset time period, and adjust the power supply quantity of the power supply system according to the total electricity consumption prediction result. Therefore, the problem of inaccurate electricity consumption prediction results in the prior art can be solved, and the effect of accurately predicting the electricity consumption is achieved.

In a specific implementation process, the decomposition unit comprises a decomposition module for generating a formulaDecomposing the first historical electricity data by using the integrated empirical mode decomposition apparatus to obtain a multi-order electricity data component and a residual component, wherein the residual component is a difference value of a sum of the first historical electricity data and the multi-order electricity data component, J is a total order of the electricity data component, x (t) is the first historical electricity data, and C _j (t) is the j-th order of the above-mentioned electricity consumption data component, r _J And (t) is the residual component described above. The device decomposes the first historical electricity consumption data into a multi-order electricity consumption data component and a residual error component, so that the time change period of each order of electricity consumption data component can be analyzed, the time change rule of the first historical electricity consumption data is comprehensively analyzed, and the electricity consumption is better predicted.

Specifically, since the integrated empirical mode decomposition apparatus is capable of decomposing the signal into components of different frequencies, in the above-described power consumption decomposing process, the first historical power consumption data is decomposed into power consumption data components having different frequencies, i.e., different time-varying periods, and each stage of power consumption data components is divided into power consumption data components using C _j And (t) expressing that after the decomposition is completed, subtracting the sum of the multi-order power consumption data components from the first historical power consumption data, and obtaining a result which is a residual component.

In order to comprehensively analyze the time variation period of the first historical electricity utilization data, the decomposition unit further comprises an input module and an acquisition module, wherein the input module is used for inputting the residual components into the corresponding LSTM neural network so as to process the residual components; the device processes the residual components through the LSTM neural network and obtains the prediction results of the residual components, so that the electricity consumption of a preset time period in the future can be comprehensively predicted according to the first historical electricity consumption data, and inaccurate electricity consumption prediction caused by omission of the first historical electricity consumption data is avoided.

Specifically, the first historical electricity consumption data is decomposed into multiple-order electricity consumption data components by the integrated empirical mode decomposition device, and after the first historical electricity consumption data is decomposed, the data in the remaining first historical electricity consumption data cannot be decomposed any more and is called a residual component, so that the residual component does not have an obvious time change period, and the residual component is input into an LSTM neural network corresponding to the residual component to obtain an electricity consumption prediction result corresponding to the residual component.

The device further comprises an acquisition unit and a training unit, wherein the acquisition unit is used for acquiring electricity consumption data before the time corresponding to the first historical electricity consumption data and environment information corresponding to the electricity consumption data to obtain second historical electricity consumption data; the training unit is used for taking the second historical electricity consumption data as input data of an initial LSTM neural network, taking the first historical electricity consumption data as output data of the initial LSTM neural network, training the initial LSTM neural network, and obtaining the LSTM neural network after training is completed. The device trains the initial LSTM neural network according to the second historical electricity consumption data, so that parameters of the LSTM neural network can be obtained to realize the prediction of electricity consumption.

Specifically, in order to train the LSTM neural network model, first, the electricity data before the time corresponding to the first historical electricity data and the environmental information corresponding to the electricity data are obtained, and are used as the second historical electricity data, that is, the input data of the initial LSTM neural network, it is to be noted that the second historical electricity data is decomposed by the integrated empirical mode decomposition device to obtain multiple-order electricity data components, the multiple-order electricity data components corresponding to the second historical electricity data are used as the input of the initial LSTM neural network model, the first historical electricity data is used as the output of the initial LSTM neural network model, the initial LSTM neural network model is trained, the parameter values, such as weights, biases, and the like, of the neural network model are obtained, and the prediction of the electricity consumption is realized based on the LSTM neural network after the training is completed.

In order to analyze the characteristic that the time change period is longer or has no obvious time change period, namely, the fluctuation characteristic in the first historical electricity consumption data, in some embodiments, the input unit comprises a calculation module and a fitting module, wherein the calculation module is used for taking each order of the electricity consumption data component as one data node of the LSTM neural network, calculating the correlation between two adjacent data nodes in the plurality of data nodes to obtain a plurality of first correlation parameters, and the number of the data nodes in the LSTM neural network is the same as the order of the electricity consumption data component; the fitting module is used for fitting a plurality of the first correlation parameters to obtain first periodic parameters of the first historical electricity utilization data, wherein the first periodic parameters represent the time change period of the first historical electricity utilization data after the two-order electricity utilization data components are overlapped.

Specifically, the 1-scale layer of the application runs on each hidden node of the LSTM neural network, extracts the characteristics of adjacent nodes, and obtains a plurality of first correlation parameters for extracting the characteristics of high-order IMF components with obvious periodicity.

In some embodiments, the input unit further includes a calculation module and a fitting module, where the calculation module is configured to use each order of the electrical data component as a data node of the LSTM neural network, use at least two adjacent data nodes as node groups, calculate correlations between the two adjacent node groups, and obtain a plurality of second correlation parameters, where the number of data nodes in the LSTM neural network is the same as the order of the electrical data component; the fitting module is used for fitting a plurality of second correlation parameters to obtain second periodic parameters of the first historical electricity utilization data, wherein the second periodic parameters represent the time change period after the fourth-order electricity utilization data components are overlapped in the first historical electricity utilization data. The device calculates the correlation between two adjacent node groups, so that the characteristic that the time change period in the first historical electricity utilization data is longer or has no obvious time change period, namely the fluctuation characteristic, can be further analyzed.

In particular, for the low order IMF component, its ripple and approximate periodicity can be regarded as a superposition of signals having different periodicity. Thus, the present invention considers multiple nodes simultaneously and calculates the correlation between them by feature extraction to capture periodicity on different scales, and the 2-scale MLSTM calculates features by considering every two neighboring nodes.

In order to obtain accurate electricity consumption prediction results, in some embodiments, the adjustment unit includes an execution module configured to add a plurality of the electricity consumption prediction results and the residual prediction results to obtain the total electricity consumption prediction result within a predetermined time period in the future. The device adds a plurality of electricity consumption prediction results obtained according to the multi-order electricity consumption data component prediction and the residual prediction results, so that the integrity of the electricity consumption prediction results can be ensured.

In some embodiments, the decomposition unit further includes a first processing module and a second processing module, where the first processing module is configured to normalize the first historical electricity consumption data before decomposing the first historical electricity consumption data by using the integrated empirical mode decomposition device; and the second processing module is used for carrying out noise removal processing on the normalized first historical electricity utilization data. The device carries out pretreatment on the data, so that the data with larger errors in the data can be removed, and inaccurate prediction of the power consumption is avoided.

The power consumption prediction method device comprises a processor and a memory, wherein the decomposition unit, the input unit, the adjustment unit and the like are all stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions. The modules are all located in the same processor; alternatively, the above modules may be located in different processors in any combination.

The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel can be provided with one or more than one, and the prediction of the electricity consumption is realized by adjusting the kernel parameters.

The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.

The embodiment of the application provides a computer readable storage medium, which comprises a stored program, wherein the program is used for controlling a device where the computer readable storage medium is located to execute the method for predicting the electricity consumption.

Specifically, the method for predicting the electricity consumption comprises the following steps:

Optionally, decomposing the first historical electricity consumption data by using an aggregate empirical mode decomposition method to obtain at least a multi-order electricity consumption data component, including: according to the formulaDecomposing the first historical electricity utilization data by using the method for decomposing the aggregate empirical mode to obtainA plurality of orders of the electricity consumption data component and a residual component, wherein the residual component is a difference value of a sum of the first historical electricity consumption data and the plurality of orders of the electricity consumption data component, J is a total order of the electricity consumption data component, x (t) is the first historical electricity consumption data, C _j (t) is the j-th order of the above-mentioned electricity consumption data component, r _J And (t) is the residual component described above.

Optionally, after obtaining the multi-order above-mentioned power consumption data component and residual component, it includes: inputting the residual components into a corresponding LSTM neural network to process the residual components; and obtaining a prediction result output by the LSTM neural network corresponding to the residual component, and obtaining a residual prediction result.

Optionally, before inputting the input data set into the LSTM neural network corresponding to each order of the electricity-using data component, the method further includes: acquiring electricity consumption data before the time corresponding to the first historical electricity consumption data and environment information corresponding to the electricity consumption data, and obtaining second historical electricity consumption data; and training the initial LSTM neural network by taking the second historical electricity utilization data as input data of the initial LSTM neural network and taking the first historical electricity utilization data as output data of the initial LSTM neural network, so as to obtain the LSTM neural network after training is completed.

Optionally, inputting the input data set into the LSTM neural network corresponding to each order of the electricity data component, so as to process the input data set, including: taking the electricity consumption data component of each step as one data node of the LSTM neural network, and calculating the correlation between two adjacent data nodes in the data nodes to obtain a plurality of first correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the electricity consumption data component; fitting a plurality of the first correlation parameters to obtain a first periodic parameter of the first historical electricity utilization data, wherein the first periodic parameter represents the time change period of the first historical electricity utilization data after the superposition of the two-order electricity utilization data components.

Optionally, inputting the input data set into the LSTM neural network corresponding to each order of the electricity data component, so as to process the input data set, and further including: taking each stage of the electricity consumption data component as a data node of the LSTM neural network, taking at least two adjacent data nodes as node subgroups, and calculating the correlation between the two adjacent node subgroups to obtain a plurality of second correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the electricity consumption data component; fitting a plurality of second correlation parameters to obtain a second periodic parameter of the first historical electricity utilization data, wherein the second periodic parameter represents the time change period of the first historical electricity utilization data after the fourth-order electricity utilization data components are overlapped.

Optionally, before decomposing the first historical electricity consumption data by using the aggregate empirical mode decomposition method, the method further includes: normalizing the first historical electricity consumption data; and carrying out noise removal processing on the normalized first historical electricity consumption data.

The embodiment of the invention provides equipment, which comprises a processor, a memory and a program stored in the memory and capable of running on the processor, wherein the processor realizes at least the following steps when executing the program:

The device herein may be a server, PC, PAD, cell phone, etc.

The application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with at least the following method steps:

It will be appreciated by those skilled in the art that the modules or steps of the application described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may be implemented in program code executable by computing devices, so that they may be stored in a storage device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps of them may be fabricated into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.

From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

1) In the method for predicting the electricity consumption, first historical electricity consumption data are decomposed into multi-order electricity consumption data components with different time change periods, each order of electricity consumption data component is used as an input data set to be input into an LSTM neural network corresponding to each order of pre-trained electricity consumption data component, the LSTM neural network is used for processing each order of electricity consumption data component, the LSTM neural network outputs electricity consumption prediction results corresponding to each order of electricity consumption data component, and then the electricity consumption prediction results corresponding to each electricity consumption data component are added to obtain total electricity consumption prediction results in a preset time period in the future. Compared with the method for predicting the electricity consumption in the future preset time period directly according to the historical electricity consumption data in the prior art, the method provided by the application can decompose the first historical electricity consumption data into the multi-order electricity consumption data components, analyze the periodic variation characteristics and the fluctuation variation characteristics of the first historical electricity consumption data through the multi-order electricity consumption data components, predict the electricity consumption in each corresponding future preset time period according to each order of electricity consumption data components, superimpose a plurality of electricity consumption prediction results to obtain the total electricity consumption prediction result in the future preset time period, and adjust the power supply quantity of the power supply system according to the total electricity consumption prediction result. Therefore, the problem of inaccurate electricity consumption prediction results in the prior art can be solved, and the effect of accurately predicting the electricity consumption is achieved.

2) In the electricity consumption prediction device, first historical electricity consumption data are decomposed into multiple-order electricity consumption data components with different time change periods, each-order electricity consumption data component is used as an input data set to be input into an LSTM neural network corresponding to each-order electricity consumption data component trained in advance, the LSTM neural network is used for processing each-order electricity consumption data component, the LSTM neural network outputs electricity consumption prediction results corresponding to each-order electricity consumption data component, and then the electricity consumption prediction results corresponding to each electricity consumption data component are added to obtain total electricity consumption prediction results in a preset time period in the future. Compared with the device for predicting the electricity consumption in the future preset time period directly according to the historical electricity consumption data in the prior art, the device provided by the application can decompose the first historical electricity consumption data into the multi-order electricity consumption data components, analyze the periodic variation characteristics and the fluctuation variation characteristics of the first historical electricity consumption data through the multi-order electricity consumption data components, predict the electricity consumption in each corresponding future preset time period according to each order of electricity consumption data components, superimpose a plurality of electricity consumption prediction results to obtain the total electricity consumption prediction result in the future preset time period, and adjust the power supply quantity of the power supply system according to the total electricity consumption prediction result. Therefore, the problem of inaccurate electricity consumption prediction results in the prior art can be solved, and the effect of accurately predicting the electricity consumption is achieved.

The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for predicting power consumption, comprising:

acquiring historical electricity consumption data and environment information corresponding to the historical electricity consumption data to obtain first historical electricity consumption data, decomposing the first historical electricity consumption data by utilizing an aggregate empirical mode decomposition method to at least obtain multi-order electricity consumption data components, wherein the time change periods of any two-order electricity consumption data components are different, and the environment information at least comprises weather conditions, electricity consumption areas and electricity consumption time;

generating the power utilization data components of each order to an input data set, and inputting the input data set into an LSTM neural network corresponding to the power utilization data components of each order to process the input data set;

obtaining each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, adding at least a plurality of power consumption prediction results to obtain a total power consumption prediction result in a future preset time period, and adjusting the power supply quantity of a power supply system according to the total power consumption prediction result in the future preset time period, wherein the future preset time period is a time period which is after and adjacent to a historical time period corresponding to the historical power consumption data;

Inputting the input data set into an LSTM neural network corresponding to each order of the power consumption data component to process the input data set, comprising:

taking each stage of the power consumption data component as one data node of the LSTM neural network, and calculating the correlation between two adjacent data nodes in the data nodes to obtain a plurality of first correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data component;

fitting a plurality of first correlation parameters to obtain first periodic parameters of the first historical electricity utilization data, wherein the first periodic parameters represent time change periods of the first historical electricity utilization data after superposition of two-order electricity utilization data components;

inputting the input data set into an LSTM neural network corresponding to each order of the power consumption data component to process the input data set, and further comprising:

taking each order of the power consumption data components as one data node of the LSTM neural network, taking at least two adjacent data nodes as node subgroups, and calculating correlation between the two adjacent node subgroups to obtain a plurality of second correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data components;

Fitting a plurality of second correlation parameters to obtain second periodic parameters of the first historical electricity utilization data, wherein the second periodic parameters represent time change periods of the first historical electricity utilization data after the fourth-order electricity utilization data components are overlapped.

2. The prediction method according to claim 1, wherein decomposing the first historical electricity consumption data by using an ensemble empirical mode decomposition method at least results in a multi-order electricity consumption data component, comprising:

according to the formulaDecomposing the first historical electricity data by using the aggregate empirical mode decomposition method to obtain a multi-order electricity data component and a residual component, wherein the residual component is a difference value of the sum of the first historical electricity data and the multi-order electricity data component, J is the total order of the electricity data component, x (t) is the first historical electricity data, C _j (t) is the electricity data component of the j th order, r _J And (t) is the residual component.

3. The prediction method according to claim 2, characterized by comprising, after obtaining the power consumption data component and the residual component in multiple steps:

inputting the residual components into a corresponding LSTM neural network to process the residual components;

And obtaining a prediction result output by the LSTM neural network corresponding to the residual component, and obtaining a residual prediction result.

4. The prediction method according to claim 1, further comprising, before inputting the input data set into the LSTM neural network corresponding to each order of the power-consuming data component to process the input data set:

acquiring electricity consumption data before the time corresponding to the first historical electricity consumption data and environment information corresponding to the electricity consumption data, and obtaining second historical electricity consumption data;

and training the initial LSTM neural network by taking the second historical electricity utilization data as input data of the initial LSTM neural network and taking the first historical electricity utilization data as output data of the initial LSTM neural network, so as to obtain the LSTM neural network after training is completed.

5. A prediction method according to claim 3, wherein adding at least a plurality of the power consumption prediction results to obtain a total power consumption prediction result for a predetermined period of time in the future comprises:

and adding the plurality of electricity consumption prediction results and the residual prediction result to obtain the total electricity consumption prediction result in a preset future time period.

6. The prediction method according to claim 1, further comprising, before decomposing the first historical electricity usage data using a collective empirical mode decomposition method:

normalizing the first historical electricity consumption data;

and carrying out noise removal processing on the normalized first historical electricity utilization data.

7. A power consumption prediction apparatus, comprising:

the system comprises a decomposition unit, a first power consumption unit and a second power consumption unit, wherein the decomposition unit is used for acquiring historical power consumption data and environment information corresponding to the historical power consumption data to obtain first historical power consumption data, decomposing the first historical power consumption data by utilizing an integrated empirical mode decomposition method to at least obtain multi-order power consumption data components, the time change periods of the multi-order power consumption data components are different, and the environment information at least comprises weather conditions, power consumption areas and power consumption time;

the input unit is used for generating the power utilization data components of each order into an input data set, and inputting the input data set into an LSTM neural network corresponding to the power utilization data components of each order so as to process the input data set;

the adjusting unit is used for acquiring each power consumption prediction result output by the LSTM neural network corresponding to each power consumption data component, adding at least a plurality of power consumption prediction results to obtain a power consumption prediction result in a preset future time period, and adjusting the power supply quantity of a power supply system according to the power consumption prediction result in the preset future time period;

The input unit includes:

the first calculation module is used for taking each order of the power consumption data component as one data node of the LSTM neural network, calculating the correlation between two adjacent data nodes in the plurality of data nodes, and obtaining a plurality of first correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data component;

the first fitting module is used for fitting a plurality of first correlation parameters to obtain first periodic parameters of the first historical electricity utilization data, wherein the first periodic parameters represent time change periods of the first historical electricity utilization data after the two-order electricity utilization data components are overlapped;

the input unit further includes:

the second calculation module is used for taking each order of the power consumption data component as one data node of the LSTM neural network, taking at least two adjacent data nodes as node subgroups, and calculating the correlation between the two adjacent node subgroups to obtain a plurality of second correlation parameters, wherein the number of the data nodes in the LSTM neural network is the same as the order of the power consumption data component;

And the second fitting module is used for fitting a plurality of second correlation parameters to obtain second periodic parameters of the first historical electricity utilization data, wherein the second periodic parameters represent time change periods of the first historical electricity utilization data after the fourth-order electricity utilization data components are overlapped.

8. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored program, wherein the program, when run, controls a device in which the computer-readable storage medium is located to perform the prediction method of any one of claims 1 to 6.

9. An electronic device comprising a memory and a processor, wherein the memory has stored therein a computer program, the processor being arranged to execute the prediction method of any of claims 1 to 6 by means of the computer program.

10. An electronic device, comprising: one or more processors, memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing the prediction method of any of claims 1-6.