CN112329997A - Power demand load prediction method and system, electronic device, and storage medium - Google Patents

Power demand load prediction method and system, electronic device, and storage medium Download PDF

Info

Publication number
CN112329997A
CN112329997A CN202011156740.6A CN202011156740A CN112329997A CN 112329997 A CN112329997 A CN 112329997A CN 202011156740 A CN202011156740 A CN 202011156740A CN 112329997 A CN112329997 A CN 112329997A
Authority
CN
China
Prior art keywords
load prediction
cnn
prediction
load
lstm
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011156740.6A
Other languages
Chinese (zh)
Inventor
钟成
李津
郭少勇
张正文
路鹏程
亢松
阮琳娜
邵苏杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xiongan New Area Power Supply Company State Grid Hebei Electric Power Co
State Grid Corp of China SGCC
Beijing University of Posts and Telecommunications
Original Assignee
Xiongan New Area Power Supply Company State Grid Hebei Electric Power Co
State Grid Corp of China SGCC
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xiongan New Area Power Supply Company State Grid Hebei Electric Power Co, State Grid Corp of China SGCC, Beijing University of Posts and Telecommunications filed Critical Xiongan New Area Power Supply Company State Grid Hebei Electric Power Co
Priority to CN202011156740.6A priority Critical patent/CN112329997A/en
Publication of CN112329997A publication Critical patent/CN112329997A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06315—Needs-based resource requirements planning or analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Marketing (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Tourism & Hospitality (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Educational Administration (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The embodiment of the invention provides a power demand load prediction method and system, electronic equipment and a storage medium, wherein the method comprises the following steps: collecting historical power load data in a certain time span of a target area as a prediction data set; inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area; wherein the load prediction model is trained based on historical power load data of a target area. According to the embodiment of the invention, the load prediction model is constructed, the historical power load data of the target area is collected to train the model, and the influence of the demand of the user in the target area on the power energy consumption is determined, so that the participation degree of the user in the prediction process is improved, and the accuracy of load prediction is further improved, so that the waste of power resources can be effectively reduced when power generation is carried out according to the load prediction result, the power resource management is optimized, and the influence on the environment is reduced.

Description

Power demand load prediction method and system, electronic device, and storage medium
Technical Field
The invention relates to the technical field of computers, in particular to a power demand load prediction method and system, electronic equipment and a storage medium.
Background
Global warming has raised a great deal of concern to the environment and energy of various societies. Potential viruses may be released by glacier melting, which brings about a huge disaster to human beings. Reducing energy usage and increasing energy efficiency are two main ways to solve this problem. However, balancing the relationship between reduced energy consumption and a surge in demand is a great challenge.
Demand Response (DR) is considered as a potential solution to this challenge in power systems, and power management can be optimized by reducing unnecessary power or adjusting non-emergency power demand. The supply and demand side two-way communication proposed along with the energy internet has facilitated a better DR model.
For a supplier, the on-demand power generation can be realized by accurately acquiring the user requirements, so that the power waste is avoided; for the user, optimization can be achieved by reducing or shifting peak hour power demand. However, the promotion of the DR mode is not expected to be so smooth as to be known by the low participation of the user. One recognized obstacle is that the physical infrastructure and information technology that is currently being deployed does not support its vision. Load prediction is a key link of the DR process, and the lack of accuracy and real-time response capability affects interaction accuracy and user satisfaction.
At present, some related researches are carried out on load prediction, but research evaluation indexes are mainly prediction accuracy, and a traditional prediction model does not carry out deeper analysis on historical load data, so that the influence of user demands on loads cannot be accurately reflected, and the relevance of the load demands on a time dimension is difficult to embody. Secondly, the influence of the complexity of the prediction algorithm on the response delay is discussed only rarely, and the method is not applicable to services with strict delay requirements.
Therefore, how to provide a power demand load prediction method and system, an electronic device, and a storage medium, how to optimize a prediction algorithm, further improve the user participation in the demand response process when predicting the load, optimize power resource management, and address environmental issues, is a problem to be solved urgently.
Disclosure of Invention
In order to overcome the defects in the prior art, embodiments of the present invention provide a power demand load prediction method and system, an electronic device, and a storage medium.
In a first aspect, an embodiment of the present invention provides a power demand load prediction method, including:
collecting historical power load data in a certain time span of a target area as a prediction data set;
inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area;
wherein the load prediction model is trained based on historical power load data of a target area.
Optionally, in the power demand load prediction method, before the step of inputting the prediction data set into the load prediction model and determining the power demand load prediction result in the target area to-be-measured time period, the method further includes:
acquiring historical power load data of a target area;
constructing a CNN-LSTM neural network model;
and training the CNN-LSTM neural network model by using the historical power load data of the target area to obtain the load prediction model.
Optionally, in the power demand load prediction method, the training of the CNN-LSTM neural network model using the historical power load data of the target area to obtain the load prediction model specifically includes:
in the training process, determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model by using an intelligent optimization algorithm;
and determining the load prediction model according to the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model.
Optionally, in the power demand load prediction method, in the training process, the determining the number of layers and the number of units of the CNN and LSTM in the CNN-LSTM neural network model by using an intelligent optimization algorithm specifically includes:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using an evolutionary algorithm.
Optionally, in the power demand load prediction method, the determining, by using an evolutionary algorithm, the number of layers and the number of units of the CNN and LSTM in the CNN-LSTM neural network model in a training process specifically includes:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a differential evolution algorithm.
Optionally, in the power demand load prediction method, the determining the number of layers and the number of units of the CNN and LSTM in the CNN-LSTM neural network model by using a differential evolution algorithm in a training process specifically includes:
determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a self-adaptive differential evolution algorithm;
and the scaling factor of the self-adaptive differential evolution algorithm is continuously adjusted in the model training process.
Optionally, in the power demand load prediction method, before the step of inputting the prediction data set into the load prediction model and determining the power demand load prediction result in the target area to-be-measured time period, the method further includes:
based on the prediction data set and the load prediction model, time complexity analysis is carried out, and load prediction response time delay under given computing capacity is estimated;
if the load prediction response time delay is judged to be larger than a preset threshold value;
and optimizing load prediction computing resource allocation.
In a second aspect, an embodiment of the present invention provides a power demand load prediction system, including:
the acquisition module is used for acquiring historical power load data in a certain time span of a target area as a prediction data set;
the prediction module is used for inputting the prediction data set into the load prediction model and determining a power demand load prediction result in a time period to be measured in the target area;
wherein the load prediction model is trained based on historical power load data of a target area.
In a third aspect, an embodiment of the present invention provides an electronic device, including a memory and a processor, where the processor and the memory complete communication with each other through a bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the various steps of the power demand load prediction method described above.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the power demand load prediction method.
According to the power demand load prediction method and system, the electronic device and the storage medium provided by the embodiment of the invention, the load prediction model is constructed, the historical power load data of the target area are collected to train the model, and the influence of the demand of the user in the target area on the power energy consumption is determined.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a flow chart of a method for predicting a power demand load according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a CNN-LSTM neural network model structure provided in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a CNN-LSTM neural network model according to another embodiment of the present invention;
fig. 4 is a flowchart of an implementation of the load prediction algorithm based on DE × CNN-LSTM according to an embodiment of the present invention;
FIG. 5 is a flowchart of a complexity analysis guidance computing resource allocation of a load prediction algorithm according to an embodiment of the present invention;
FIG. 6 is a network topology diagram of an energy Internet environment provided by an embodiment of the invention;
fig. 7 is a schematic diagram of parameter settings of the DE x CNN-LSTM load prediction algorithm according to an embodiment of the present invention;
FIG. 8 is a line graph illustrating the prediction accuracy of three load prediction models provided by embodiments of the present invention;
fig. 9 is a time delay line graph corresponding to load prediction according to an embodiment of the present invention;
FIG. 10 is a schematic structural diagram of a power demand load forecasting system according to an embodiment of the present invention;
fig. 11 is a schematic physical structure diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The prior art mainly includes the following three methods for load prediction:
the method comprises the following steps: the power load prediction method based on the multi-time scale long-time memory neural network comprises the following steps:
receiving historical power load data and area characteristic factors in a region needing to be predicted, which are input by a user, through an input unit;
constructing a regional power load prediction model according to the historical power load data and regional characteristic factors;
training and modeling the regional power load prediction model by adopting a long-time memory neural network (LSTM) to generate a multi-time scale long-time memory neural (MT-LSTM) network, wherein the MT-LSTM network comprises a plurality of LSTM units, the LSTM units of the MT-LSTM network are divided into three groups of { G1, G2 and G3}, and the LSTM units of each group Gk (k is more than or equal to 1 and less than or equal to 3) capture historical power load data and regional characteristic factors of different time scales in different time periods;
predicting the power load in the region to be predicted by utilizing a multi-time scale long-time memory neural network generated by training and generating a power load prediction result in the region;
and outputting the power load prediction result in the required prediction area through an output unit.
The method only adopts LSTM to predict the load, and compared with CNN-LSTM, the method is weaker in data feature abstraction capability. The main body of the scheme is to predict the power load of multiple time scales, but the application scenes and services of long-term prediction and short-term prediction are not explicitly described, and whether different scales are combined for application is possible. Besides, the scheme does not consider the complexity of the algorithm, so that the application range of the scheme is further limited.
The second method comprises the following steps: the power load prediction system based on edge calculation:
the system includes an edge computing server and one or more embedded energy management devices. The edge computing server in the system is mainly responsible for acquiring external characteristic factor data, processing the data, constructing a long-time memory neural network-based power load prediction model, and transmitting the constructed power load prediction model and configuration to the embedded energy management equipment;
the embedded energy management equipment is mainly responsible for collecting equipment data and predicting power load and managing energy according to the model.
The power load prediction system can reduce the configuration requirement of the power load prediction system on the computing capacity of the embedded energy management equipment based on the edge computing architecture, and can exchange data in real time and update a power load prediction model through the data interaction modules at the two ends, so that the critical requirements in the aspects of real-time business, data optimization, application intelligence, prediction precision and the like are met.
The system includes an edge computing server and one or more embedded energy management devices. The edge computing server in the system is mainly responsible for acquiring external characteristic factor data, processing the data, constructing a long-time memory neural network-based power load prediction model, and transmitting the constructed power load prediction model and the configuration thereof to the embedded energy management equipment; the embedded energy management equipment is mainly responsible for collecting equipment data and predicting power load and managing energy according to the model. Wherein the edge computing server comprises: the system comprises a data interaction module, a data acquisition module, a data preprocessing module and a model establishing module; the embedded energy management equipment comprises a data interaction module, a power load prediction module, a data acquisition module and an energy management module. However, the invention only describes the system composition and the information interaction flow, and does not explain the processing capability of each module and what processing algorithm is adopted. The given system is relatively generalized, and further detailed and targeted discussion is needed for specific scenes and services.
The third method comprises the following steps: the medium-long term power load prediction method comprises the following steps:
acquiring power load data and influence factors thereof in a preset time period;
dividing the power load data and the influence factors into a plurality of first time scale data according to a first time scale, and dividing the influence factors into a plurality of second time scale data according to a second time scale;
constructing a power load prediction model according to the plurality of first time scale data and the plurality of second time scale data;
training the power load prediction model by adopting a long-short term memory neural network to generate a stacked long-short term memory network model;
and predicting the power load by stacking the long-term and short-term memory network models, and generating power load prediction data.
The method can solve the problem of data dependence of different time scales and improve the prediction precision of medium and long-term power loads.
The method is based on multi-time scale data and a long-short term memory neural network for modeling, can solve the problem of dependence of different time scale data, and improves the prediction precision of medium and long term power loads. In addition, embodiments of the invention have some additional features: the first time scale data includes electrical load data and climate data, and the second time scale data includes economic data and climate data. The preset time period is years, the first time scale is months, and the second time scale is years. However, the invention also does not consider the time complexity problem of the prediction method, and limits the application range of the prediction method.
Fig. 1 is a flowchart of a power demand load prediction method according to an embodiment of the present invention, and as shown in fig. 1, the method includes:
step S1, collecting historical power load data in a certain time span of a target area as a prediction data set;
step S2, inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area;
wherein the load prediction model is trained based on historical power load data of a target area.
Specifically, the energy internet can be understood as a network that comprehensively uses advanced power electronic technology, information technology and intelligent management technology to interconnect a large number of energy nodes such as a novel power network, an oil network, a natural gas network and the like, which are composed of distributed energy collection devices, distributed energy storage devices and various types of loads, so as to realize energy peer-to-peer exchange and sharing of energy in bidirectional flow.
In the energy internet, demand forecasting is the basis for power price formulation; the demand reduction situation determines the degree of adjustment of the electricity price. Therefore, in the process of predicting the power demand load, the participation degree of users needs to be improved, so that the accuracy of load prediction is further improved.
In step S1, historical power load data in a certain time span of the target area is collected as a prediction data set.
For example: in the energy internet, demand forecasting is the basis for power price formulation; the increase or decrease of the demand determines the adjustment degree of the electricity price. Each cell is assumed to have an aggregator, which is responsible for information interaction between users and suppliers, and the information mainly includes power consumption of the users, willingness to participate in demand response, price of electricity issued by the suppliers, and the like. The information exchange is performed once every hour, and since the demand response implementation that we mainly consider is load shedding during peak hours, the prediction process execution period is 9:00-23: 00.
If the power demand load of the cell a (target area) is predicted between 9:00 and 10:00 in 8 am of 10/8 of 2020, historical power load data of the cell a before 9:00 in 8 am of 5/8 of 2020 is collected as a prediction data set, for example, historical power load data of the cell a in 8:00 am of 5/8 of 2020 to 8:00 am of 10/8 of 2020 is collected as a prediction data set.
Since the amount of data in the prediction data set is related to the calculation resources to be used and the prediction time lag to be generated at the time of prediction, the amount of data must be considered at the time of selection of the prediction data set, and therefore, too small a size affects the accuracy of prediction, and too large a size wastes too much calculation resources.
In addition, the data in the forecast data set may be continuous or filtered, for example, the power demand load between 5/8/2020 and 10/7/2020 and between 9:00 and 10: 00/am each day is selected as the forecast data set. Selecting a data set for the prediction time period can further improve the accuracy of the prediction.
In the actual application process, the specific data size of the historical data set, the selected time period, and the specific composition of the data may be adjusted according to the actual situation, which is not limited in this embodiment. The historical data used to train the load prediction model is not limited as such.
In step S2, the prediction data set acquired in step S1 is input into a load prediction model trained in advance using the historical power load data of the target area, and a power demand load prediction result in the target area measurement time period is determined.
The power generation is performed according to the load prediction result, and the waste of power resources can be effectively reduced.
According to the power demand load forecasting method provided by the embodiment of the invention, the load forecasting model is constructed, the historical power load data of the target area is collected to train the model, and the influence of the demand of the user in the target area on the power energy consumption is determined.
Based on the foregoing embodiment, optionally, in the power demand load prediction method, before the step of inputting the prediction data set into the load prediction model and determining the power demand load prediction result in the target area time period to be measured, the method further includes:
acquiring historical power load data of a target area;
constructing a CNN-LSTM neural network model;
and training the CNN-LSTM neural network model by using the historical power load data of the target area to obtain the load prediction model.
Specifically, before the load prediction model is used for power demand load prediction, a neural network model structure needs to be constructed, and the model needs to be trained to obtain a trained model.
Fig. 2 is a schematic structural diagram of a CNN-LSTM neural network model provided in an embodiment of the present invention, and as shown in fig. 2, the model is formed by a CNN (convolutional neural network) and an LSTM (long short term memory network). The CNN part is suitable for extracting data features, the LSTM is widely applied to processing time series problems, and the CNN-LSTM can perform feature representation from the two dimensions and is suitable for load prediction.
Obtaining historical power load data of a target area, training the constructed CNN-LSTM neural network model, and when the accuracy reaches preset precision, determining that the model is successfully trained to obtain a load prediction model.
It should be noted that, in this embodiment, the condition for stopping training the model may set the maximum iteration number as the condition for stopping training the model, in addition to the accuracy of the model reaching the preset precision. The method can be specifically adjusted according to actual conditions, and this embodiment does not limit this.
On the basis of the embodiment, the load prediction model is jointly constructed by using the CNN and the LSTM, so that the historical load data can be analyzed to a greater extent, the influence of user demands on the load can be effectively reflected, the relevance of the load demands on a time dimension is reflected, and the accuracy of load prediction is improved.
Based on the foregoing embodiment, optionally, in the power demand load prediction method, the training the CNN-LSTM neural network model using the historical power load data of the target area to obtain the load prediction model specifically includes:
in the training process, determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model by using an intelligent optimization algorithm;
and determining the load prediction model according to the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model.
Specifically, in the CNN-LSTM neural network model, the number of layers and units of CNN and LSTM is not preset, but is further determined by a joint intelligent optimization algorithm in the training process.
In the model training process, the intelligent optimization algorithm is used for determining the number of layers and the number of units of CNN and LSTM in the CNN-LSTM neural network model, and parameter configuration of the CNN-LSTM neural network model under the historical power load data training sample of the fixed data volume is obtained.
And determining the number of layers and units of the CNN and the LSTM in the CNN-LSTM neural network model according to the determined parameter configuration, and further determining the load prediction model.
It should be noted that the intelligent optimization algorithm includes a genetic algorithm, a differential evolution algorithm, an immune algorithm, an ant colony algorithm, a particle swarm algorithm, a tabu search algorithm, and the like, and may be selected and improved according to actual requirements in a specific application process, which is not limited in this embodiment.
On the basis of the embodiment, the structure of the CNN-LSTM neural network model is optimized by using an intelligent optimization algorithm in the training process, so that the model has better convergence performance under high accuracy, and the accuracy of prediction by using the model is improved.
Based on the foregoing embodiment, optionally, in the power demand load prediction method, in the training process, determining the number of layers and the number of units of the CNN and LSTM in the CNN-LSTM neural network model by using an intelligent optimization algorithm specifically includes:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using an evolutionary algorithm.
Specifically, evolutionary computation is a search algorithm based on biological evolution mechanisms such as natural selection and natural inheritance, and in the process of searching for the optimal solution, the optimal solution is generally improved from one group of original problems to another group of better solutions, and then further improved from the improved solutions.
The essence of the evolutionary algorithm is a clustering algorithm, which searches a strategy space through clustering and ergodic mode. The problem of local minima in the non-linearity problem is better overcome than the gradient descent. The network architecture can be effectively improved by using an evolutionary algorithm.
In the embodiment of the invention, the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model are determined in the training process by using an evolutionary algorithm, so that the optimal network framework and parameters within certain prediction precision are obtained.
It should be noted that the evolutionary algorithm includes a genetic algorithm, a differential evolutionary algorithm, a particle swarm algorithm, a shuffled frog leaping algorithm, and the like, and may be selected according to actual requirements in a specific application process, and further, the evolutionary algorithm may be improved, and a dynamic adaptive technique (such as a fuzzy adaptive method) is adopted to automatically adjust algorithm control parameters and encoding precision in the evolutionary process, which is not limited in this embodiment.
Based on the foregoing embodiment, optionally, fig. 3 is a schematic structural diagram of a CNN-LSTM neural network model provided in another embodiment of the present invention, as shown in fig. 3, in the power demand load prediction method, the determining, by using an evolutionary algorithm, the number of layers and the number of units of CNN and LSTM in the CNN-LSTM neural network model in a training process specifically includes:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a differential evolution algorithm.
Specifically, the ConvD, MaxPooling2D, Flatten and density parts in the figure represent CNN parts, including convolutional layers, pooling layers and compression layers, suitable for extracting data features. The LSTM and RpeatVector parts in the graph represent LSTM, the LSTM is widely applied to processing time series problems, and a CNN-LSTM neural network model is optimized by using a differential evolution algorithm.
The Differential Evolution Algorithm (DE) is commonly used to solve the global optimization problem, and compared with other genetic algorithms, the Differential Evolution Algorithm is used to optimize the CNN-LSTM neural network model, so that the method has better convergence.
Based on the above embodiments, optionally, fig. 4 is a flowchart of an implementation of the load prediction algorithm based on DE × CNN-LSTM according to an embodiment of the present invention, as shown in fig. 4,
in the power demand load prediction method, the determining the number of layers and the number of units of the CNN and LSTM in the CNN-LSTM neural network model by using a differential evolution algorithm in a training process specifically includes:
determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a self-adaptive differential evolution algorithm;
and the scaling factor of the self-adaptive differential evolution algorithm is continuously adjusted in the model training process.
Specifically, the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model are determined in the training process by using the adaptive differential evolution algorithm DE (' indicates that the scaling factor of DE is adjustable), and the scaling factor of the adaptive differential evolution algorithm is continuously adjusted in the model training process.
The specific algorithm for load prediction based on DE-CNN-LSTM is as follows:
Figure BDA0002743019440000131
Figure BDA0002743019440000141
the core idea of the algorithm is as follows: 1) an initial parameter set of DE is obtained.
2) The CNN-LSTM training parameters are used.
3) And judging whether the fitness function is larger than the error requirement or whether the iteration number is smaller than the upper limit.
4) If yes, performing mutation, crossing and selection of DE, and then returning to the step 2; otherwise, jumping out of the loop and obtaining the optimal parameter setting, namely obtaining the architecture of the load prediction algorithm based on DE-CNN-LSTM.
It should be noted that the above detailed adaptive differential evolution algorithm flow is only used as a specific example to explain the present invention, and in a specific application process, the adaptive differential evolution algorithm may be optimized and adjusted according to an actual situation, which is not limited in this embodiment.
On the basis of the embodiment, the embodiment of the invention can further improve the convergence of model training, shorten the model training time and reduce unnecessary waste of computing resources on the basis of ensuring the prediction precision by continuously adjusting the scaling factor of the self-adaptive differential evolution algorithm in the model training process.
Based on the foregoing embodiment, optionally, fig. 5 is a flowchart for guiding the calculation of resource allocation by complexity analysis of a load prediction algorithm provided in the embodiment of the present invention, and as shown in fig. 5, in the power demand load prediction method, before the step of inputting a prediction data set into a load prediction model and determining a power demand load prediction result in a time period to be measured in a target area, the method further includes:
based on the prediction data set and the load prediction model, time complexity analysis is carried out, and load prediction response time delay under given computing capacity is estimated;
if the load prediction response time delay is judged to be larger than a preset threshold value;
and optimizing load prediction computing resource allocation.
Specifically, the data volume of the data set and the corresponding time delay of the load prediction influenced by the computing resources are predicted.
And performing time complexity analysis based on the prediction data set and the load prediction model, and predicting the load prediction response time delay under the given computing capacity.
The temporal complexity of CNN can be expressed as:
Figure BDA0002743019440000161
p is the side length of each convolution kernel output feature map. Q is the number of layers of the convolutional layer. l represents the first convolutional layer. Cl-1And ClRepresenting the corresponding channel. The temporal complexity of LSTM is:
Figure BDA0002743019440000162
where m represents the input size and n represents the hidden layer size. The total temporal complexity of CNN-LSTM is:
Figure BDA0002743019440000163
and comparing the calculated time delay (total time complexity) with a preset threshold (maximum acceptable time delay), and judging whether the load prediction response time delay is greater than the preset threshold.
And if the load prediction response time delay is judged to be larger than the preset threshold, optimizing the distribution of the load prediction computing resources (determining the quantity of computing resources to be distributed by utilizing the ratio relation).
For example, selecting an on-peak demand period, 13:00-14:00 a day, with an input prediction data set size of 12600, one can obtain an expected response delay at the current computing resource r of:
Figure BDA0002743019440000164
the time delay requirement is set as
Figure BDA0002743019440000165
If it is
Figure BDA0002743019440000166
The currently allocated computing resources can meet the demand, otherwise, the additionally allocated resource amount is
Figure BDA0002743019440000167
It should be noted that, in the embodiment of the present invention, load prediction is performed by using an edge node, if the predicted response delay obtained by calculation is greater than a preset value, the calculation resource allocation is preferentially adjusted at the edge node, and if the response delay calculation is performed again after allocation, it is found that the service requirement cannot be still met, the calculation task is considered to be unloaded to another node or uploaded to the cloud, and load prediction is performed at the cloud. The specific resource allocation arrangement condition may be adjusted according to the actual situation, which is not limited in this embodiment.
On the basis of the above embodiment, the embodiment of the present invention determines whether the current computing resource can meet the predicted delay requirement by performing the load prediction corresponding delay evaluation before the load prediction, so that the scheme can be applied to the service with high delay requirement, and the waste of the prediction time and the computing resource is reduced.
The load prediction algorithm and the advantages of the algorithm provided by the present invention are described below with reference to specific examples:
fig. 6 is a network topology diagram of an energy internet environment according to an embodiment of the present invention, and as shown in fig. 6, an energy internet environment is simulated by using iFogSim. Compared with other simulation tools, the iFogSim is customized for edge calculation, and can better simulate edge nodes and IoT devices. It includes three basic components: physical components, logical components, and management components.
An energy internet environment was simulated using iFogSim. Compared with other simulation tools, the iFogSim is customized for edge calculation, and can better simulate edge nodes and IoT devices. It includes three basic components: physical components, logical components, and management components.
The invention is mainly characterized by power requirement, and in the embodiment, the number of units and the number of layers of the LSTM and the Dense are optimized. We used PSO-CNN-LSTM prediction, DE-CNN-LSTM prediction (scale factor not adjustable) and DE-CNN-LSTM prediction (scale factor adjustable) as comparison algorithms.
Fig. 7 is a schematic diagram of parameter settings of the DE x CNN-LSTM load prediction algorithm according to an embodiment of the present invention, where the parameter settings of DE x CNN-LSTM prediction (scaling factor adjustable) are shown in fig. 7.
Fig. 8 is a line graph of the prediction accuracy of the three load prediction models provided by the embodiment of the present invention, as shown in fig. 8, the three algorithms are modeled respectively, and in the process of multiple experiments, it is found that the prediction accuracy can reach the target higher than 99% within 30 iterations.
Wherein, PSO-CNN-LSTM (Particle swarm optimization, PSO for short) has the shortest time for completing genetic evolution, but the optimal accuracy rate is 93.18%; DE-CNN-LSTM takes 843.37s and has an accuracy of 99.01%, and the optimal number of cells for LSTM and Dense is 34 and 2, respectively. For DE-CNN-LSTM, the time is 1100.71s, the accuracy can reach 99.28%, and the optimal number of cells for LSTM and Dense is 6 and 1, respectively.
It is clear that both algorithms DE-CNN-LSTM and DE-CNN-LSTM meet the accuracy requirements, which is less time consuming than DE-CNN-LSTM. In the same way we have found that a layer of LSTM and density is sufficient to meet our accuracy requirements. The architecture of DE-CNN-LSTM is now determined, its complexity calculated as follows:
Timeo～O(nm+n2+n)～O(m)
to evaluate the relationship between complexity and response time, 4 sets of training data sets were set and 2 sets of delays were tested to obtain the gain ratio. Their input data set sizes are 8400 and 17640, respectively, with corresponding delays of 14.0204s and 22.2327s, respectively. The delays of other volume data sets can be predicted by:
Figure BDA0002743019440000181
fig. 9 is a time delay line graph corresponding to load prediction, as shown in fig. 9, eight groups of data with different complexity are tested, and a test result shows that an average prediction accuracy probability reaches 99.0186%, which proves that a response time delay is in positive correlation with time complexity.
In addition, we tested the response delay introduced by the cloud processing, as shown by the red dots. For these eight different volume data set inputs, processing with edge nodes reduces response delay by 79% -85%.
Therefore, the problem that due to the fact that the data volume is large, the response delay is insufficient and does not meet the service requirement can be effectively solved, and the load prediction result can meet the response requirement.
Fig. 10 is a flowchart of a power demand load prediction method according to an embodiment of the present invention, and as shown in fig. 10, the system includes:
the acquisition module 101 is configured to acquire historical power load data of a target area within a certain time span as a prediction data set;
the prediction module 102 is configured to input the prediction data set into a load prediction model, and determine a power demand load prediction result in a time period to be measured in a target area;
wherein the load prediction model is trained based on historical power load data of a target area.
Specifically, the energy internet can be understood as a network that comprehensively uses advanced power electronic technology, information technology and intelligent management technology to interconnect a large number of energy nodes such as a novel power network, an oil network, a natural gas network and the like, which are composed of distributed energy collection devices, distributed energy storage devices and various types of loads, so as to realize energy peer-to-peer exchange and sharing of energy in bidirectional flow.
In the energy internet, demand forecasting is the basis for power price formulation; the demand reduction situation determines the degree of adjustment of the electricity price. Therefore, in the process of predicting the power demand load, the participation degree of users needs to be improved, so that the accuracy of load prediction is further improved.
In the embodiment of the present invention, the obtaining module 101 is configured to collect historical power load data of a target area within a certain time span as a prediction data set.
For example: in order to predict the power demand load of the cell a (target area) between 9:00 and 10:00 in the morning of 10/8/2020, historical power load data of the cell a before 9:00 in the morning of 5/8/2020 is collected as a prediction data set, for example, historical power load data of the cell a in the time range of 8:00 in the morning of 5/8/2020 to 8:00 in the morning of 10/8/2020 is collected as the prediction data set.
Since the amount of data in the prediction data set is related to the calculation resources to be used and the prediction time lag to be generated at the time of prediction, the amount of data must be considered at the time of selection of the prediction data set, and therefore, too small a size affects the accuracy of prediction, and too large a size wastes too much calculation resources.
In addition, the data in the forecast data set may be continuous or filtered, for example, the power demand load between 5/8/2020 and 10/7/2020 and between 9:00 and 10: 00/am each day is selected as the forecast data set. Selecting a data set for the prediction time period can further improve the accuracy of the prediction.
In the actual application process, the specific data size of the historical data set, the selected time period, and the specific composition of the data may be adjusted according to the actual situation, which is not limited in this embodiment. The historical data used to train the load prediction model is not limited as such.
The prediction module is configured to input the prediction data set acquired by the acquisition module 101 into a load prediction model trained in advance by using historical power load data of the target area, and determine a power demand load prediction result in the time period to be measured in the target area.
The power generation is performed according to the load prediction result, and the waste of power resources can be effectively reduced.
According to the power demand load forecasting system provided by the embodiment of the invention, the load forecasting model is constructed, the historical power load data of the target area is collected to train the model, and the influence of the demand of the user in the target area on the power energy consumption is determined.
It should be noted that, the power demand load prediction system provided in the embodiment of the present invention is used for executing the power demand load prediction method, and a specific implementation manner thereof is consistent with the method implementation manner, and is not described herein again.
Fig. 11 is a schematic entity structure diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 11, the electronic device may include: a processor (processor)111, a communication interface (communication interface)112, a memory (memory)113 and a communication bus (bus)114, wherein the processor 111, the communication interface 112 and the memory 113 complete communication with each other through the communication bus 114. The processor 111 may call logic instructions in the memory 113 to execute the above power demand load prediction method, including: collecting historical power load data in a certain time span of a target area as a prediction data set; inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area; wherein the load prediction model is trained based on historical power load data of a target area.
In addition, the logic instructions in the memory 113 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
In another aspect, an embodiment of the present invention further provides a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute the power demand load prediction method provided by the above-mentioned method embodiments, including: reading the currently stored electronic tag information of all tools to obtain the currently stored data information of all tools; the tool comprises: collecting historical power load data in a certain time span of a target area as a prediction data set; inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area; wherein the load prediction model is trained based on historical power load data of a target area.
In another aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented by a processor to execute the methods provided in the foregoing embodiments to perform power demand load prediction, where the method includes: collecting historical power load data in a certain time span of a target area as a prediction data set; inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area; wherein the load prediction model is trained based on historical power load data of a target area.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A power demand load prediction method comprising:
collecting historical power load data in a certain time span of a target area as a prediction data set;
inputting the prediction data set into a load prediction model, and determining a power demand load prediction result in a time period to be measured in a target area;
wherein the load prediction model is trained based on historical power load data of a target area.
2. The power demand load prediction method according to claim 1, wherein before the step of inputting the prediction data set into the load prediction model and determining the power demand load prediction result in the time period in which the target area is to be measured, the method further comprises:
acquiring historical power load data of a target area;
constructing a CNN-LSTM neural network model;
and training the CNN-LSTM neural network model by using the historical power load data of the target area to obtain the load prediction model.
3. The power demand load prediction method according to claim 2, wherein the training of the CNN-LSTM neural network model using the historical power load data of the target area to obtain the load prediction model specifically comprises:
in the training process, determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model by using an intelligent optimization algorithm;
and determining the load prediction model according to the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model.
4. The power demand load forecasting method of claim 3, wherein during the training process, the number of layers and the number of units of CNN and LSTM in the CNN-LSTM neural network model are determined by using an intelligent optimization algorithm, and specifically comprises:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using an evolutionary algorithm.
5. The power demand load prediction method according to claim 4, wherein the determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model by using an evolutionary algorithm in a training process specifically comprises:
and determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a differential evolution algorithm.
6. The power demand load prediction method according to claim 5,
the determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model by using a differential evolution algorithm in a training process specifically comprises the following steps:
determining the number of layers and the number of units of the CNN and the LSTM in the CNN-LSTM neural network model in a training process by using a self-adaptive differential evolution algorithm;
and the scaling factor of the self-adaptive differential evolution algorithm is continuously adjusted in the model training process.
7. The power demand load prediction method according to any one of claims 1 to 6, further comprising, before the step of inputting the prediction data set into the load prediction model and determining the power demand load prediction result in the time period in which the target area is to be measured, the steps of:
based on the prediction data set and the load prediction model, time complexity analysis is carried out, and load prediction response time delay under given computing capacity is estimated;
if the load prediction response time delay is judged to be larger than a preset threshold value;
and optimizing load prediction computing resource allocation.
8. A power demand load prediction system, comprising:
the acquisition module is used for acquiring historical power load data in a certain time span of a target area as a prediction data set;
the prediction module is used for inputting the prediction data set into the load prediction model and determining a power demand load prediction result in a time period to be measured in the target area;
wherein the load prediction model is trained based on historical power load data of a target area.
9. An electronic device, comprising a memory and a processor, wherein the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the power demand load prediction method of any one of claims 1 to 7.
10. A non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the power demand load prediction method according to any one of claims 1 to 7.
CN202011156740.6A 2020-10-26 2020-10-26 Power demand load prediction method and system, electronic device, and storage medium Pending CN112329997A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011156740.6A CN112329997A (en) 2020-10-26 2020-10-26 Power demand load prediction method and system, electronic device, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011156740.6A CN112329997A (en) 2020-10-26 2020-10-26 Power demand load prediction method and system, electronic device, and storage medium

Publications (1)

Publication Number Publication Date
CN112329997A true CN112329997A (en) 2021-02-05

Family

ID=74311703

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011156740.6A Pending CN112329997A (en) 2020-10-26 2020-10-26 Power demand load prediction method and system, electronic device, and storage medium

Country Status (1)

Country Link
CN (1) CN112329997A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113256114A (en) * 2021-05-25 2021-08-13 国网宁夏电力有限公司电力科学研究院 Day-ahead and time-ahead two-stage demand response method based on distributed power generation consumption
CN113255973A (en) * 2021-05-10 2021-08-13 曙光信息产业(北京)有限公司 Power load prediction method, power load prediction device, computer equipment and storage medium
CN113283659A (en) * 2021-06-03 2021-08-20 上海分未信息科技有限公司 Power load response task allocation method for virtual power plant
CN114862177A (en) * 2022-04-29 2022-08-05 国网江苏省电力有限公司南通供电分公司 Energy interconnection energy storage and distribution method and system

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104865830A (en) * 2015-04-29 2015-08-26 华北电力大学（保定） Dual-intelligent-optimization control method for unit load
CN107622329A (en) * 2017-09-22 2018-01-23 深圳市景程信息科技有限公司 The Methods of electric load forecasting of Memory Neural Networks in short-term is grown based on Multiple Time Scales
CN107947738A (en) * 2017-12-11 2018-04-20 南京航空航天大学 A kind of Forecasting Methodology of solar energy unmanned plane cell plate voltage
CN109376012A (en) * 2018-10-10 2019-02-22 电子科技大学 A kind of self-adapting task scheduling method based on Spark for isomerous environment
CN109840512A (en) * 2019-02-28 2019-06-04 北京科技大学 A kind of Facial action unit recognition methods and identification device
CN109991542A (en) * 2019-03-27 2019-07-09 东北大学 Lithium ion battery residual life prediction technique based on WDE optimization LSTM network
CN110532154A (en) * 2018-05-23 2019-12-03 中国移动通信集团浙江有限公司 Application system expansion method, device and equipment
CN111027775A (en) * 2019-12-12 2020-04-17 中国长江三峡集团有限公司 Step hydropower station generating capacity prediction method based on long-term and short-term memory network
CN111080472A (en) * 2019-11-12 2020-04-28 国网山西省电力公司晋中供电公司 Load prediction and analysis method for power system
CN111091247A (en) * 2019-12-24 2020-05-01 天津相和电气科技有限公司 Power load prediction method and device based on deep neural network model fusion
CN111105029A (en) * 2018-10-29 2020-05-05 北京地平线机器人技术研发有限公司 Neural network generation method and device and electronic equipment
CN111144643A (en) * 2019-12-24 2020-05-12 天津相和电气科技有限公司 Day-ahead power load prediction method and device based on double-end automatic coding
CN111241957A (en) * 2020-01-04 2020-06-05 圣点世纪科技股份有限公司 Finger vein in-vivo detection method based on multi-feature fusion and DE-ELM
CN111310684A (en) * 2020-02-24 2020-06-19 东声(苏州)智能科技有限公司 Model training method and device, electronic equipment and storage medium

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104865830A (en) * 2015-04-29 2015-08-26 华北电力大学（保定） Dual-intelligent-optimization control method for unit load
CN107622329A (en) * 2017-09-22 2018-01-23 深圳市景程信息科技有限公司 The Methods of electric load forecasting of Memory Neural Networks in short-term is grown based on Multiple Time Scales
CN107947738A (en) * 2017-12-11 2018-04-20 南京航空航天大学 A kind of Forecasting Methodology of solar energy unmanned plane cell plate voltage
CN110532154A (en) * 2018-05-23 2019-12-03 中国移动通信集团浙江有限公司 Application system expansion method, device and equipment
CN109376012A (en) * 2018-10-10 2019-02-22 电子科技大学 A kind of self-adapting task scheduling method based on Spark for isomerous environment
CN111105029A (en) * 2018-10-29 2020-05-05 北京地平线机器人技术研发有限公司 Neural network generation method and device and electronic equipment
CN109840512A (en) * 2019-02-28 2019-06-04 北京科技大学 A kind of Facial action unit recognition methods and identification device
CN109991542A (en) * 2019-03-27 2019-07-09 东北大学 Lithium ion battery residual life prediction technique based on WDE optimization LSTM network
CN111080472A (en) * 2019-11-12 2020-04-28 国网山西省电力公司晋中供电公司 Load prediction and analysis method for power system
CN111027775A (en) * 2019-12-12 2020-04-17 中国长江三峡集团有限公司 Step hydropower station generating capacity prediction method based on long-term and short-term memory network
CN111091247A (en) * 2019-12-24 2020-05-01 天津相和电气科技有限公司 Power load prediction method and device based on deep neural network model fusion
CN111144643A (en) * 2019-12-24 2020-05-12 天津相和电气科技有限公司 Day-ahead power load prediction method and device based on double-end automatic coding
CN111241957A (en) * 2020-01-04 2020-06-05 圣点世纪科技股份有限公司 Finger vein in-vivo detection method based on multi-feature fusion and DE-ELM
CN111310684A (en) * 2020-02-24 2020-06-19 东声(苏州)智能科技有限公司 Model training method and device, electronic equipment and storage medium

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
DAWEI GENG ETC.: "《Short-term Wind Speed Prediction Based on Principal Component Analysis and LSTM》", 《APPLIED SCIENCE》 *
王昆仑等: "《数据结构与算法》" *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113255973A (en) * 2021-05-10 2021-08-13 曙光信息产业(北京)有限公司 Power load prediction method, power load prediction device, computer equipment and storage medium
CN113256114A (en) * 2021-05-25 2021-08-13 国网宁夏电力有限公司电力科学研究院 Day-ahead and time-ahead two-stage demand response method based on distributed power generation consumption
CN113283659A (en) * 2021-06-03 2021-08-20 上海分未信息科技有限公司 Power load response task allocation method for virtual power plant
CN114862177A (en) * 2022-04-29 2022-08-05 国网江苏省电力有限公司南通供电分公司 Energy interconnection energy storage and distribution method and system

Similar Documents

Publication Publication Date Title
Kumar et al. A hybrid multi-agent based particle swarm optimization algorithm for economic power dispatch
CN112329997A (en) Power demand load prediction method and system, electronic device, and storage medium
Fu et al. Planning of distributed renewable energy systems under uncertainty based on statistical machine learning
Jalalkamali Using of hybrid fuzzy models to predict spatiotemporal groundwater quality parameters
Xie et al. An integrated Gaussian process modeling framework for residential load prediction
Capizzi et al. An advanced neural network based solution to enforce dispatch continuity in smart grids
CN110837915A (en) Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning
Jeyaraj et al. Computer‐assisted demand‐side energy management in residential smart grid employing novel pooling deep learning algorithm
Chiu et al. A hybrid CNN-GRU based probabilistic model for load forecasting from individual household to commercial building
Liu et al. Dynamic economic dispatch with wind power penetration based on non-parametric kernel density estimation
CN111445065A (en) Energy consumption optimization method and system for refrigeration group control of data center
Shi et al. Load Forecasting of Electric Vehicle Charging Stations: Attention Based Spatiotemporal Multi-Graph Convolutional Networks
Zhao et al. LandSys II: Agent-based land use–forecast model with artificial neural networks and multiagent model
CN113240219A (en) Land utilization simulation and prediction method
CN117350336A (en) Charging load prediction and model training method, device, equipment and storage medium
Wang et al. A study of situation awareness-based resource management scheme in cloud environment
CN117290090A (en) Node cooperative resource allocation method, device, equipment and storage medium
CN111784019A (en) Power load processing method and device
CN116415732A (en) User side power load data processing method based on improved ARNN
CN113537607B (en) Power failure prediction method
Wang et al. Charging load forecasting of electric vehicles based on sparrow search algorithm‐improved random forest regression model
CN114048886A (en) Effective berthage combination prediction method based on neural network, terminal and storage medium
Balaji et al. Deep Learning Based Energy Consumption Prediction on Internet of Things Environment.
Lijuan Research on Algorithm and Modification of Elman Network Based on Fuzzy Classification and Short-Term Electricity Consumption Prediction
Dewangan et al. Load Forecasting Models in Smart Grid Using Smart Meter Information: A Review. Energies 2023, 16, 1404

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20210205

RJ01 Rejection of invention patent application after publication