CN111241755A - Power load prediction method - Google Patents

Power load prediction method Download PDF

Info

Publication number
CN111241755A
CN111241755A CN202010113676.7A CN202010113676A CN111241755A CN 111241755 A CN111241755 A CN 111241755A CN 202010113676 A CN202010113676 A CN 202010113676A CN 111241755 A CN111241755 A CN 111241755A
Authority
CN
China
Prior art keywords
data
model
time
output
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
CN202010113676.7A
Other languages
Chinese (zh)
Inventor
王林钰
郭磊
楚天舒
陈浩
黄晓霖
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.)
State Grid Suzhou Urban Energy Research Institute Co ltd
State Grid Jiangsu Electric Power Co Ltd
Original Assignee
State Grid Suzhou Urban Energy Research Institute Co ltd
State Grid Jiangsu Electric Power Co Ltd
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 State Grid Suzhou Urban Energy Research Institute Co ltd, State Grid Jiangsu Electric Power Co Ltd filed Critical State Grid Suzhou Urban Energy Research Institute Co ltd
Priority to CN202010113676.7A priority Critical patent/CN111241755A/en
Publication of CN111241755A publication Critical patent/CN111241755A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/044Recurrent networks, e.g. Hopfield networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06Energy or water supply
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04SSYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

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

Abstract

The invention provides a power load prediction method based on a long-short term memory neural network and external auxiliary information, which comprises the steps of firstly, acquiring and summarizing historical energy consumption data, meteorological data and holiday information according to a determined research object to form an initial sequence data set; then, cleaning the summarized data set, eliminating time sequence data strips at missing moments, interpolating small-range missing data, and normalizing the data to form a time sequence data pool; and then constructing a long-short term memory neural network model, and training the model based on the time sequence data pool. The method inspects and combines the coupling relation of the power energy consumption and the weather and holiday information on the basis of the long-term and short-term memory neural network, has the characteristics of high model precision and flexible prediction period, and is beneficial to the application of dispatching, analyzing and the like of the power grid.

Description

Power load prediction method
Technical Field
The invention relates to the technical field of power information, in particular to a power load prediction method based on a long-short term memory neural network and external auxiliary information.
Background
The electric power energy is a pillar energy of the modern society, accurately and effectively predicts the load of the power grid, and is of great importance to the safe and stable operation of the power grid and the economy and high efficiency of electric power production, so that the load prediction is always a research hotspot in the field of electric power information. For an energy-consuming object, the energy consumption of the energy-consuming object is influenced by a plurality of factors such as self characteristics, external environment, time period and the like, so that the external appearance of load data is very random, effective analysis and prediction are difficult to perform based on a physical mechanism, and traditional models such as linear regression and the like cannot meet the requirements in the actual load prediction of a complex energy-consuming object.
With the construction and investment of the smart grid and the vigorous development of information technology, on one hand, energy consumption data and external environment data of the energy utilization object can be recorded in real time or regularly, and massive historical data of the electric energy load are generated; on the other hand, data-based information processing methods are emerging continuously, and the method surpasses the traditional mechanism analysis method in the aspect of describing the potential characteristics and randomness of the object. Currently, common power load prediction methods include Artificial Neural Networks (ANN), Support Vector Machines (SVMs), Autoregressive Moving Average models (ARIMA), and the like.
The traditional technology has the following technical problems:
however, in practical applications, the method usually cannot compromise the periodicity of the energy consumption data and the external randomness, so that the final prediction accuracy is limited.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a power load prediction method based on a long-short term memory neural network and external auxiliary information, the method can effectively analyze the self energy consumption characteristics of energy utilization objects and the coupling relation with external factors based on historical energy consumption data and external environment data, fully excavates the periodic characteristics of power energy consumption, and can accurately predict the power load at the future moment based on actual historical data after model training is completed. The method can be used for the problems in the field of power information such as power system analysis, power grid prediction and scheduling and the like.
In order to solve the technical problem, the invention provides a power load prediction method based on a long-short term memory neural network and external auxiliary information, which comprises the following specific steps:
first step, historical data summarization
Collecting and summarizing the power energy consumption data of the selected energy utilization object according to the selected energy utilization object, and recording external environment data at corresponding moments;
secondly, preprocessing the data to obtain a time sequence data pool
Preprocessing the summarized historical data, eliminating abnormal values, then performing normalization processing, and taking the obtained data as a time sequence data pool required by model training;
thirdly, LSTM network modeling and model training
In the time series data pool, historical time series data (x) with the length of nt-n+1,...,xt) As model input, energy consumption data (y) for m successive time points in the futuret+1,...,yt+m) And constructing an LSTM network model as model output, and training the model by adopting data in the time sequence data pool.
And fourthly, predicting the power load at the future moment according to the actual historical data and the trained LSTM network model.
In one embodiment, in the first step, the external environment data to be recorded includes temperature, humidity, wind power, precipitation and other available meteorological data and holiday information corresponding to the time.
In one embodiment, in the second step, the abnormal energy consumption data values are eliminated, specifically:
calculating the mean and standard deviation of all the acquired historical data:
Figure BDA0002390832770000031
Figure BDA0002390832770000032
and eliminating the energy consumption data outside the (mu-3 sigma, mu +3 sigma) interval.
In one embodiment, in the second step, linear padding is performed on a small range of missing data, specifically:
Figure BDA0002390832770000033
and m-n is less than 3, and the corresponding missing time sequence data strips which do not belong to the condition are subjected to integral elimination.
In one embodiment, in the second step, to eliminate the influence of different dimensions, the cleaned and filled energy consumption and meteorological data are normalized, specifically:
Figure BDA0002390832770000034
in one embodiment, in the second step, the holiday information is binarized, specifically, if the current day is a working day, x is 1, otherwise x is 0.
In one embodiment, in the third step, the input information of the constructed LSTM network includes: the electric energy load of the first 24 hours, the temperature, the humidity, the air pressure, the wind direction and the holiday information of the predicted time total 29 groups of data.
In one embodiment, in the third step, the constructed LSTM network contains two hidden layers, the first hidden layer includes 20 LSTM units, and the second hidden layer includes 10 LSTM units.
In one embodiment, in the third step, each LSTM unit in the constructed model includes two states of a unit state and a hidden state, and three thresholds of a forgetting gate, an input gate, and an output gate, specifically:
the cell state represents the attribute of the current cell itself, and is marked as C at time tt
The hidden state represents the attribute of the current neuron output externally, and is recorded as h at the time tt
The forgetting gate selects the unit state at the previous moment according to the output at the current moment and the hidden state at the previous moment,
ft=σ(Wf·[ht-1,xt]+bf)
where σ (-) is a sigmoid function, WfAnd bfRespectively are the weight and the deviation item of the forgetting gate;
the input gate generates candidate cell states based on the input and the hidden state of the previous time, and performs screening based on the threshold condition,
Figure BDA0002390832770000041
it=σ(Wi·[ht-1,xt]+bi)
similarly, where tanh (. cndot.) is a hyperbolic tangent function, WCAnd bCTo generate weights and bias terms for candidate cell states, WiAnd biRespectively inputting the weight and the bias item of the threshold;
then updating the unit state;
Figure BDA0002390832770000042
the output gate determines the value of the hidden state according to the unit state at the current moment, and performs screening according to the input at the current moment and the hidden state at the previous moment,
ot=σ(Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
wherein, WoAnd boThe weight and the bias term of the output threshold are respectively, and finally the hidden state h is generatedtThe overall output as an LSTM unit is passed to the next layer and to the next time.
In one embodiment, in the third step, the constructed LSTM network output layer is a linear layer, specifically,
ypred=Wh·ht+bh
wherein, WhAnd bhRespectively, weights and bias terms for the output layer.
In one embodiment, in the third step, the evaluation index of the constructed LSTM network model is the root mean square error, specifically,
Figure BDA0002390832770000051
in one embodiment, in the third step, the constructed LSTM network model is trained using the ADAM algorithm, the learning rate is set to 0.001, and the training duration is 1500 epochs.
Based on the same inventive concept, the present application also provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of any of the methods when executing the program.
Based on the same inventive concept, the present application also provides a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of any of the methods.
Based on the same inventive concept, the present application further provides a processor for executing a program, wherein the program executes to perform any one of the methods.
The invention has the beneficial effects that:
the invention discloses a power load prediction method based on a long-term and short-term memory neural network and external auxiliary information, which abstracts power load prediction into a time sequence analysis problem and adopts an artificial intelligence method to solve the problem. Firstly, effectively preprocessing historical data, and ensuring the integrity and the availability of the data through cleaning, linear filling and normalization; an LSTM network is adopted as a reference model for energy consumption modeling, the time sequence characteristic of energy consumption information is fully considered, and the problem of gradient disappearance existing in a common recurrent neural network is avoided; meanwhile, external factors related to electric energy consumption are analyzed, and combined modeling is carried out; the training method has better robustness for models under different data. From experimental verification conditions, the method can effectively improve the accuracy of the load prediction of the energy-using object in application.
Drawings
FIG. 1 is a flow chart of a power load prediction method based on a long-short term memory neural network and external auxiliary information according to the present invention.
FIG. 2 is a structural diagram of an LSTM unit in the method for predicting a power load based on a long-short term memory neural network and external auxiliary information according to the present invention.
FIG. 3 is a 1500epochs training effect diagram in the power load prediction method based on the long-short term memory neural network and the external auxiliary information.
FIG. 4 is a diagram of the model prediction effect in the power load prediction method based on the long-short term memory neural network and the external auxiliary information.
Detailed Description
The present invention is further described below in conjunction with the following figures and specific examples so that those skilled in the art may better understand the present invention and practice it, but the examples are not intended to limit the present invention.
The invention discloses a power load prediction method based on a long-short term memory neural network and external auxiliary information, which is processed according to the following detailed steps as shown in figure 1:
historical data collection summary
According to the selected prediction objects, historical energy consumption data of the selected prediction objects per hour are collected and summarized, and as can be seen from an actual load curve in fig. 4, besides a certain periodicity, the electric energy load also has volatility and contingency at the same time, which are influenced by external factors. According to investigation and analysis, the relationship between external meteorological factors and energy consumption is close, so that meteorological data in 4 parts of temperature, humidity, air pressure and wind power are selected and added into an energy consumption model according to the acquirable data condition; in addition, holidays are also critical to energy consumption and should therefore also be taken into account in the energy consumption model. The collected data sample is shown in table 1.
TABLE 1
Time of day Load(s) Temperature of Humidity Air pressure Wind power
01/01/2011 00:59 970 -4.17 0.74 1013.907 5.42
01/01/2011 01:59 932 -4.43 0.731739 1013.8 5.56
01/01/2011 02:59 884 -4.76 0.720385 1013.8 5.01
01/01/2011 03:59 839 -5.13 0.73 1012.88 4.59
01/01/2011 04:59 813 -5.50 0.73 1012.8 4.30
01/01/2011 05:59 819 -5.99 0.73 1012.8 4.69
01/01/2011 06:59 853 -5.92 0.744615 1012.8 4.76
01/01/2011 07:59 891 -5.74 0.7684 1011.788 5.38
Data pre-processing
And preprocessing the collected and aggregated data to facilitate subsequent model processing. Firstly, eliminating abnormal loads and corresponding data strips, specifically, calculating the average value and variance of all acquired historical energy consumption data:
Figure BDA0002390832770000071
Figure BDA0002390832770000072
and eliminating the energy consumption data outside the (mu-3 sigma, mu +3 sigma) interval.
Second, a small range of missing data is linearly padded, specifically,
Figure BDA0002390832770000073
and m-n is less than 3, and the corresponding missing time sequence data strips which do not belong to the condition are subjected to integral elimination.
Then, in order to eliminate the influence brought by different dimensions, the energy consumption and meteorological data after cleaning and filling are normalized, specifically:
Figure BDA0002390832770000074
finally, the holiday information is subjected to binarization processing, specifically, if the current day is a working day, x is 1, otherwise x is 0.
Model building and training
In the load prediction task, it is necessary to predict the load at a target time from the history data. In this case, the invention uses the energy consumption data of the first 24 hours and external auxiliary information of the prediction time to model and predict the load without losing generality. Specifically, the model input comprises the electric energy load of the previous 24 hours, and the temperature, humidity, air pressure, wind direction and holiday information at the predicted moment, and 29 groups of data are calculated; the model output is the load at the predicted time.
In this regard, the invention builds an LSTM network model based on the PyTorch deep learning framework, specifically, the model includes two hidden layers and one output layer. The first hidden layer contains 20 LSTM units and the second hidden layer contains 10 LSTM units. Each LSTM unit comprises a unit state, a hidden state, a forgetting gate, an input gate and an output gate, and specifically comprises the following three thresholds:
the cell state represents the attribute of the current cell itself, and is marked as C at time tt
The hidden state represents the attribute of the current neuron output externally, and is recorded as h at the time tt
The forgetting gate selects the unit state at the previous moment according to the output at the current moment and the hidden state at the previous moment,
ft=σ(Wf·[ht-1,xt]+bf)
where σ (-) is a sigmoid function, WfAnd bfRespectively are the weight and the deviation item of the forgetting gate;
the input gate generates candidate cell states based on the input and the hidden state of the previous time, and performs screening based on the threshold condition,
Figure BDA0002390832770000081
it=σ(Wi·[ht-1,xt]+bi)
similarly, where tanh (. cndot.) is a hyperbolic tangent function, WCAnd bCTo generate weights and bias terms for candidate cell states, WiAnd biRespectively inputting the weight and the bias item of the threshold;
then, updating the unit state;
Figure BDA0002390832770000082
the output gate determines the value of the hidden state according to the unit state at the current moment, and performs screening according to the input at the current moment and the hidden state at the previous moment,
ot=σ(Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
wherein, WoAnd boThe weight and the bias term of the output threshold are respectively, and finally the hidden state h is generatedtThe overall output as an LSTM unit is passed to the next layer and to the next time.
The output layer of the model is a linear layer, specifically,
ypred=Wh·ht+bh
wherein, WhAnd bhRespectively, weights and bias terms for the output layer.
After the model construction is completed, the model loss function is set to the root mean square error, specifically,
Figure BDA0002390832770000091
based on the obtained and processed data, the model is trained by adopting an ADAM algorithm, the learning rate is 0.001, and the training time is 1500 epochs. The model training effect is shown in fig. 3.
Actual load prediction
After the model training is finished, the load of the future time is predicted based on actual historical data and external auxiliary information of the prediction time, and the prediction is used for assisting the work of analysis, operation and maintenance, scheduling and the like of the reference power grid. Since the training data is normalized, after the model obtains a predicted value, inverse transformation is required to obtain a final predicted load, specifically.
Figure BDA0002390832770000092
In order to verify the practical effect of the invention, the invention adopts the real load data and the meteorological data of a place of Rou Bu Er ya of 2011 all the year to carry out experimental analysis, wherein the holiday only considers two cases of weekday and weekend. The experimental result shows that the constructed model can accurately predict the load change of the energy consumption object, and can play a great positive role in the analysis and early warning of the power grid.
Specifically, the first 90% of data is used as a training set for guiding model training, and the second 10% of data is used as a verification set for verifying the prediction effect of the model. After the model is trained by 1500epochs, the loss function is reduced to a lower level, at this time, the training is finished, the model is adopted to predict and verify the concentrated load, and the prediction result is shown in fig. 4, wherein the ordinate is the load, the abscissa is the time, the blue curve is the real load, and the orange curve is the predicted load. It can be seen that the real load has strong periodicity and certain randomness. The final predicted value obtained by the LSTM network model based on the historical data and the external auxiliary data can accurately fit the trend of the real load, and only a small amount of errors exist in the daily peaks and valleys.
The invention only utilizes the historical energy consumption data of the energy object and external auxiliary information which is easy to obtain to effectively predict the load at the future moment, does not relate to the specific energy object mechanism research, does not need corresponding prior knowledge, and completes the load prediction task in a mode of constructing a data model through an LSTM network. Based on PyTorch framework development, the whole method is simple, convenient and reliable in flow. Meanwhile, when the model is applied, the time granularity and prediction of data can be flexibly adjusted according to actual conditions, and the model is more flexible than a traditional method and is beneficial to the accuracy and the real-time performance of an actual load prediction task.
The above-mentioned embodiments are merely preferred embodiments for fully illustrating the present invention, and the scope of the present invention is not limited thereto. The equivalent substitution or change made by the technical personnel in the technical field on the basis of the invention is all within the protection scope of the invention. The protection scope of the invention is subject to the claims.

Claims (10)

1. A power load prediction method based on a long-short term memory neural network and external auxiliary information is characterized by comprising the following steps:
and collecting and summarizing historical power energy consumption data and external environment data at corresponding moments according to the selected energy utilization objects.
Preprocessing the summarized data, eliminating abnormal values, then performing normalization processing, and taking the obtained data as a time sequence data pool required by model training;
in the time series data pool, historical time series data (x) with the length of nt-n+1,...,xt) As model input, energy consumption data (y) of m continuous moments in the futuret+1,...,yt+m) Constructing an LSTM network model as model output, and training the model by adopting data in a time sequence data pool;
and predicting the power load at the future moment according to the actual historical data and the trained LSTM network model.
2. The method according to claim 1, wherein the external environmental data includes temperature, humidity, wind, precipitation and holiday information corresponding to the time.
3. The method of claim 1 for power load prediction based on long-short term memory neural network and external auxiliary information, wherein the constructed input information of the LSTM network comprises: the electric energy load of the first 24 hours, the temperature, the humidity, the air pressure, the wind direction and the holiday information of the predicted time total 29 groups of data.
4. The method as claimed in claim 1, wherein the constructed LSTM network comprises two hidden layers, the first hidden layer comprises 20 LSTM units, and the second hidden layer comprises 10 LSTM units.
5. The method for predicting the power load based on the long-short term memory neural network and the external auxiliary information as claimed in claim 1, wherein each LSTM unit in the constructed model comprises two states of a unit state and a hidden state and three thresholds of a forgetting gate, an input gate and an output gate, and specifically:
the cell state represents the attribute of the current cell itself, and is marked as C at time tt
The hidden state represents the attribute of the current neuron output externally, and is recorded as h at the time tt
The forgetting gate selects the unit state at the previous moment according to the output at the current moment and the hidden state at the previous moment,
ft=σ(Wf·[ht-1,xt]+bf)
where σ (-) is a sigmoid function, WfAnd bfRespectively are the weight and the deviation item of the forgetting gate;
the input gate generates candidate cell states based on the input and the hidden state of the previous time, and performs screening based on the threshold condition,
Figure FDA0002390832760000021
it=σ(Wi·[ht-1,xt]+bi)
wherein tanh (. cndot.) is a hyperbolic tangent function, WCAnd bCTo generate weights and bias terms for candidate cell states, WiAnd biRespectively inputting the weight and the bias item of the threshold;
then updating the unit state;
Figure FDA0002390832760000022
the output gate determines the value of the hidden state according to the unit state at the current moment, and performs screening according to the input at the current moment and the hidden state at the previous moment,
ot=σ(Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
wherein, WoAnd boThe weight and the bias term of the output threshold are respectively, and finally the hidden state h is generatedtThe overall output as an LSTM unit is passed to the next layer and to the next time.
6. The power load prediction method based on long-short term memory neural network and external auxiliary information as claimed in claim 1, wherein in the third step, the constructed LSTM network output layer is a linear layer, specifically,
ypred=Wh·ht+bh
wherein, WhAnd bhRespectively, weights and bias terms for the output layer.
7. The method according to claim 1, wherein the constructed LSTM network model is trained by using ADAM algorithm, learning rate is set to 0.001, and training duration is 1500 epochs.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the program is executed by the processor.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
10. A processor, characterized in that the processor is configured to run a program, wherein the program when running performs the method of any of claims 1 to 7.
CN202010113676.7A 2020-02-24 2020-02-24 Power load prediction method Pending CN111241755A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010113676.7A CN111241755A (en) 2020-02-24 2020-02-24 Power load prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010113676.7A CN111241755A (en) 2020-02-24 2020-02-24 Power load prediction method

Publications (1)

Publication Number Publication Date
CN111241755A true CN111241755A (en) 2020-06-05

Family

ID=70873344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010113676.7A Pending CN111241755A (en) 2020-02-24 2020-02-24 Power load prediction method

Country Status (1)

Country Link
CN (1) CN111241755A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112733457A (en) * 2021-01-18 2021-04-30 武汉大学 Load electricity utilization characteristic modeling method for improving double-layer long-short term memory network
CN112990556A (en) * 2021-02-24 2021-06-18 江苏大学 User power consumption prediction method based on Prophet-LSTM model
CN113033766A (en) * 2020-10-23 2021-06-25 广州博纳信息技术有限公司 Energy efficiency trend prediction method based on LSTM neural network algorithm model
CN113077105A (en) * 2021-04-16 2021-07-06 国网安徽省电力有限公司 Long-holiday load prediction method and device
CN113239614A (en) * 2021-04-22 2021-08-10 西北工业大学 Atmospheric turbulence phase space-time prediction algorithm
CN113792490A (en) * 2021-09-16 2021-12-14 国网江苏省电力有限公司营销服务中心 Cement mixer energy consumption modeling method based on support vector regression
CN113962456A (en) * 2021-10-19 2022-01-21 江苏方天电力技术有限公司 Medium-and-long-term load prediction method considering industry relevance
CN114326391A (en) * 2021-12-13 2022-04-12 哈尔滨工程大学 Building energy consumption prediction method
CN114623569A (en) * 2021-11-04 2022-06-14 国网浙江省电力有限公司湖州供电公司 Cluster air conditioner load differentiation regulation and control method based on deep reinforcement learning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109034500A (en) * 2018-09-04 2018-12-18 湘潭大学 A kind of mid-term electric load forecasting method of multiple timings collaboration
CN109659933A (en) * 2018-12-20 2019-04-19 浙江工业大学 A kind of prediction technique of power quality containing distributed power distribution network based on deep learning model
CN110059844A (en) * 2019-02-01 2019-07-26 东华大学 Energy storage device control method based on set empirical mode decomposition and LSTM
CN110738344A (en) * 2018-07-20 2020-01-31 中国农业大学 Distributed reactive power optimization method and device for load prediction of power system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110738344A (en) * 2018-07-20 2020-01-31 中国农业大学 Distributed reactive power optimization method and device for load prediction of power system
CN109034500A (en) * 2018-09-04 2018-12-18 湘潭大学 A kind of mid-term electric load forecasting method of multiple timings collaboration
CN109659933A (en) * 2018-12-20 2019-04-19 浙江工业大学 A kind of prediction technique of power quality containing distributed power distribution network based on deep learning model
CN110059844A (en) * 2019-02-01 2019-07-26 东华大学 Energy storage device control method based on set empirical mode decomposition and LSTM

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113033766A (en) * 2020-10-23 2021-06-25 广州博纳信息技术有限公司 Energy efficiency trend prediction method based on LSTM neural network algorithm model
CN112733457A (en) * 2021-01-18 2021-04-30 武汉大学 Load electricity utilization characteristic modeling method for improving double-layer long-short term memory network
CN112733457B (en) * 2021-01-18 2022-03-15 武汉大学 Load electricity utilization characteristic modeling method for improving double-layer long-short term memory network
CN112990556A (en) * 2021-02-24 2021-06-18 江苏大学 User power consumption prediction method based on Prophet-LSTM model
CN113077105A (en) * 2021-04-16 2021-07-06 国网安徽省电力有限公司 Long-holiday load prediction method and device
CN113077105B (en) * 2021-04-16 2023-11-24 国网安徽省电力有限公司 Long-false-day load prediction method and device
CN113239614A (en) * 2021-04-22 2021-08-10 西北工业大学 Atmospheric turbulence phase space-time prediction algorithm
CN113792490A (en) * 2021-09-16 2021-12-14 国网江苏省电力有限公司营销服务中心 Cement mixer energy consumption modeling method based on support vector regression
CN113962456A (en) * 2021-10-19 2022-01-21 江苏方天电力技术有限公司 Medium-and-long-term load prediction method considering industry relevance
CN114623569A (en) * 2021-11-04 2022-06-14 国网浙江省电力有限公司湖州供电公司 Cluster air conditioner load differentiation regulation and control method based on deep reinforcement learning
CN114623569B (en) * 2021-11-04 2023-09-29 国网浙江省电力有限公司湖州供电公司 Cluster air conditioner load differential regulation and control method based on deep reinforcement learning
CN114326391A (en) * 2021-12-13 2022-04-12 哈尔滨工程大学 Building energy consumption prediction method

Similar Documents

Publication Publication Date Title
CN111241755A (en) Power load prediction method
CN113962364B (en) Multi-factor power load prediction method based on deep learning
Wu et al. A novel hybrid system based on multi-objective optimization for wind speed forecasting
CN112508275B (en) Power distribution network line load prediction method and equipment based on clustering and trend indexes
CN112949945B (en) Wind power ultra-short-term prediction method for improving bidirectional long-term and short-term memory network
CN112990556A (en) User power consumption prediction method based on Prophet-LSTM model
CN112686464A (en) Short-term wind power prediction method and device
CN113554466B (en) Short-term electricity consumption prediction model construction method, prediction method and device
CN116307652A (en) Artificial intelligent resource allocation method for intelligent power grid
CN113449919B (en) Power consumption prediction method and system based on feature and trend perception
CN114218872B (en) DBN-LSTM semi-supervised joint model-based residual service life prediction method
CN112329990A (en) User power load prediction method based on LSTM-BP neural network
Massaoudi et al. Short-term electric load forecasting based on data-driven deep learning techniques
CN112803398A (en) Load prediction method and system based on empirical mode decomposition and deep neural network
CN115860177A (en) Photovoltaic power generation power prediction method based on combined machine learning model and application thereof
CN115358437A (en) Power supply load prediction method based on convolutional neural network
CN115034485A (en) Wind power interval prediction method and device based on data space
CN116757465A (en) Line risk assessment method and device based on double training weight distribution model
CN116663746A (en) Power load prediction method and device, computer equipment and storage medium
CN115099135A (en) Improved artificial neural network multi-type operation power consumption prediction method
CN117113086A (en) Energy storage unit load prediction method, system, electronic equipment and medium
CN116470491A (en) Photovoltaic power probability prediction method and system based on copula function
CN114254828B (en) Power load prediction method based on mixed convolution feature extractor and GRU
CN114298413A (en) Hydroelectric generating set runout trend prediction method
CN113761023A (en) Photovoltaic power generation short-term power prediction method based on improved generalized neural network

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