CN114611757A - Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network - Google Patents
Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network Download PDFInfo
- Publication number
- CN114611757A CN114611757A CN202210129326.9A CN202210129326A CN114611757A CN 114611757 A CN114611757 A CN 114611757A CN 202210129326 A CN202210129326 A CN 202210129326A CN 114611757 A CN114611757 A CN 114611757A
- Authority
- CN
- China
- Prior art keywords
- power system
- residual error
- network
- error network
- genetic algorithm
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
A short-term load prediction method of a power system based on a genetic algorithm and an improved depth residual error network comprises the following steps: acquiring power system load data of a certain power grid company within a week, considering the influence of weather characteristics of the current day on the power system load data, and performing data visualization on the weather characteristics of the current day; secondly, carrying out structural improvement on the deep residual error network based on the deep residual error network, determining a power system load prediction network topological structure, and adding an attention mechanism; and thirdly, adjusting the network topology structure determined in the second step by using a genetic algorithm. The short-term load prediction method of the power system based on the genetic algorithm and the improved deep residual error network can accurately predict the short-term load of the power system, is not easy to fall into a local optimal solution in a model training process, and has higher training efficiency and prediction accuracy compared with an original deep residual error network.
Description
Technical Field
The invention relates to the technical field of electric power, in particular to a short-term load prediction method of an electric power system based on a genetic algorithm and an improved deep residual error network.
Background
Accurate power load provides corresponding reference for the operation mode of the power grid. The power load prediction can be divided into long-term, medium-term and short-term predictions according to the length of time, and the design is developed around the short-term power load prediction. The short-term power load prediction is more significant than the medium-term and long-term power load prediction, and the accurate short-term power load prediction can reasonably arrange the maintenance of various devices in the power system. Meanwhile, the method is beneficial to saving energy sources such as coal, oil and the like. This is in line with the now advocated building of energy-saving society. Short-term load forecasting provides a reference for relevant power departments to establish electricity prices. The price of electricity is set up in relation to the future load of electricity. Therefore, accurate power load prediction can make lower electricity price for power departments, and the competitiveness of enterprises is improved.
Deep learning can be divided into a generation framework, a discrimination framework and a hybrid framework according to a learning framework. The generation architecture model mainly comprises: a limited Boltzmann machine, a self-encoder, a deep belief network, etc. The discriminant architecture model mainly comprises: deep feedforward networks, convolutional neural networks, and the like. The hybrid architecture model is then a collection of these two architectures. Deep learning can be classified into unsupervised learning and supervised learning according to whether data has a label. The unsupervised learning method mainly comprises the following steps: limited Boltzmann machine, automatic encoder, deep belief network, deep Boltzmann machine, etc. The supervised learning method mainly comprises the following steps: deep sensors, deep feed-forward networks, convolutional neural networks, deep stacked networks, recurrent neural networks, and the like. A number of experimental studies have shown that there is no clear boundary between supervised and unsupervised learning, such as: the deep belief network uses both a supervised learning method and an unsupervised learning method in the training process.
The existing prediction method based on deep learning researches can accurately predict the short-term load of the power system, but the problem of model overfitting in the training process cannot be solved well, and the influence of the initial threshold value and the weight of the prediction model on the final prediction accuracy is less considered.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a power system short-term load prediction method based on a genetic algorithm and an improved deep residual error network, the power system short-term load is predicted by using a mode of combining the genetic algorithm and the improved deep residual error network, and compared with other methods, the method not only can more accurately predict the load value, but also has better real-time performance.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows: a short-term load prediction method of a power system based on a genetic algorithm and an improved depth residual error network comprises the following specific steps:
acquiring power system load data of a certain power grid company within a week, considering the influence of weather characteristics of the current day on the power system load data, and performing data visualization on the weather characteristics of the current day;
secondly, carrying out structure improvement on the deep residual error network based on the deep residual error network, determining a power system load prediction network topological structure, and adding an attention mechanism;
thirdly, adjusting the network topology structure determined in the second step by using a genetic algorithm;
fourthly, dividing the data of one week obtained in the first step into a training set and a testing set, training and testing the network in the third step, and predicting a final load data value of the power system;
preferably, the time interval of the acquired load data of the power system is two hours, and the weather and meteorological feature data are classified into three types, including: the highest air temperature and the lowest air temperature of the time period of each datum, and the weather characteristic value of the day;
preferably, the weather characteristic value is 0 in sunny days, 0.5 in cloudy days and 1 in rainy days;
preferably, the depth residual network in the second step is ResNet50, the module for improvement belongs to the inclusion network, and the added attention mechanism is a channel attention mechanism and a spatial attention mechanism.
Preferably, the third step is adjusted to an initial weight and a threshold length.
Preferably, in the data set division in the fourth step, the load data of the first six days in the week is a training set, and the load data of the last day is a test set.
The short-term load prediction method of the power system based on the genetic algorithm and the improved deep residual error network can accurately predict the short-term load of the power system, is not easy to fall into a local optimal solution in a model training process, and has higher training efficiency and prediction accuracy compared with an original deep residual error network.
Drawings
The invention is further illustrated by the following examples in conjunction with the accompanying drawings:
FIG. 1 is a schematic flow chart of a method for short term load forecasting of a power system according to the present invention;
FIG. 2 is a schematic diagram of a network topology constructed in an embodiment of the present invention;
FIG. 3 is a schematic diagram of a residual structure applied in the embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a spatial attention and channel attention model applied in an embodiment of the present invention;
FIG. 5 is a flow chart of a genetic algorithm used in an embodiment of the present invention;
FIG. 6 is a comparison graph of the short-term load value and the real value of the power system predicted by applying the method of the present invention in the embodiment of the present invention.
Detailed Description
As shown in fig. 1, the method of the present invention elaborates the construction of the short-term load database of the power system and the construction of the prediction network structure in detail, and comprises the following steps:
the method comprises the steps of firstly, acquiring power system load data of a certain power grid company within a week, considering the influence of weather characteristics on the power system load data on the power grid company, and carrying out data visualization on the weather characteristics on the day.
In the invention, the real data set is from a network company of a certain country and comprises short-term load data and weather data of a week, the data selection rule is to take one conforming data every two hours, so that 12 short-term load data of the power system are obtained every day, the weather characteristic data formulation rule is that the highest temperature and the lowest temperature of the day, the sunny day is 0, the cloudy day is 0.5, and the rainy day is 1.
And secondly, carrying out structural improvement on the deep residual error network based on the deep residual error network, determining a power system load prediction network topological structure, and adding an attention mechanism. Fig. 2 shows a diagram of the finally determined network topology.
The invention uses an increment v3 structure to replace a convolution and pooling layer in the initial stage of a ResNet50 network on the basis of a ResNet50 network, and simultaneously modifies the number of output nodes. The improved main model structure comprises an inclusion structure and a residual structure. As shown in fig. 2, the inclusion structure is composed of three parallel convolution and pooling layers, the residual structure is composed of four residual network structures, each residual structure is composed of several residual blocks, each residual layer is connected to an attention mechanism structure, 4 attention mechanism structures are added, and finally, a full-connection layer is connected, a support vector machine SVM is used for classifying and outputting prediction results.
The residual blocks are shown in fig. 3. The residual module consists of 4 convolution kernels, the input N passes through the first convolution layer
Because the parameter of the network model is large, firstly, a convolution kernel with the size of 1 × 1 is used to reduce the dimension of N, the dimensions of the output f (N) of the residual error network and the input data N may not keep consistency to a certain extent, therefore, when the dimensions of f (N) and N are the same, an identity mapping residual error module is used to process the two, and the formula is as follows:
M(N)＝ReLU(N+F(N))
where m (n) is the output result of the residual module, f (n) is the residual mapping function, and ReLU () is the activation function.
And when the two dimensions are different, the following formula is used for processing:
M(N)＝ReLU(λ(N)+F(N))
where λ is the linear projection.
The attention mechanism added by the invention is divided into a space attention mechanism and a channel attention mechanism, and a specific structure diagram of the attention mechanism is shown in fig. 4. The channel attention mechanism adopts two pooling modes of global average pooling and maximum pooling, is used for better highlighting key characteristic information of an input matrix, eliminates the problem of redundant information accumulation caused by important information omission due to the fact that an original network only uses average pooling, and can improve the network prediction accuracy, and the calculation formula is as follows:
wherein γ represents a sigmod function; the MLP represents a shared network and consists of a hidden layer and a plurality of layers of perceptrons; AvgPool represents mean pooling, MaxPool represents maximum pooling;
representing a global average pooling of channel attention mechanisms,
representing the maximum average pooling calculation.
The calculation formula of the channel attention mechanism is as follows:
wherein, g7×7Representing a convolution calculation of size 7 x 7,
representing a global average pooling of spatial attention mechanisms,
representing the maximum average pooling calculation.
And thirdly, adjusting the network topology structure determined in the second step by using a genetic algorithm. Because the initial weight and the threshold of the network constructed by the invention have high randomness, the network may be trapped in a local minimum value in the training process. The formula of the encoding operation is as follows:
wherein (bi1, bi 2.. bil), i-th segment of a certain individual is, each segment is l in length, each bik is 0 or 1, and Ti and Ri are two end points of the domain of the Xi-th segment component.
The genetic algorithm is combined with the network topology model, the natural selection process is simulated through the genetic algorithm, and the initial solution generated by the network is subjected to reproduction iteration through the genetic theory to generate the global optimal solution. And inputting the optimal initial value into a network model for data training to generate a most reasonable power system load prediction model based on a genetic algorithm so as to obtain an optimal predicted value. The genetic algorithm is mainly used for optimizing the initial weight and the threshold value of the network determined by the invention. Firstly, carrying out optimized searching selection on data through relevant operations such as selection, crossing, mutation and the like; then, importing the data into a network for training; and finally, outputting the optimal result of the power system load prediction model, wherein the flow is shown in a figure 5.
And fourthly, dividing the data of one week obtained in the first step into a training set and a testing set, training and testing the network in the third step, and predicting the final load data value of the power system.
The data selected by the invention are divided into seven groups, which respectively represent data of seven days, each group comprises 15 data, the first 12 data are actual load values of the power system selected every 2 hours, the first two data are the lowest air temperature and the highest air temperature of the day, the third data are the weather condition of the day, 0 represents a clear day, 0.5 represents a cloudy day, 1 represents a rainy day, and the specific data are shown in table 1:
table 1 data presentation
The data of the previous 6 days are taken as a training set, input into the network structure constructed by the invention, trained and finally output a prediction result. The specific implementation steps are shown as specific embodiments.
Specifically, the power load was measured every two hours. Thus, there are 12 sets of load data per day. Considering that the power load prediction is greatly affected by factors such as environment and weather, the maximum air temperature, the minimum air temperature, and the weather characteristic value (0 indicates a clear day, 0.5 indicates a cloudy day, and 1 indicates a rainy day) are added to the sample. Here, the power load of the day before the load prediction and the weather characteristic value of the day are used as input variables of the network. An input vector of unit length 15 is thus obtained.
The power load measured every 2 hours on the day of the forecast day is taken as the target vector of the network. An output vector of unit length 12 is thus obtained.
The convolution algorithm can only process the normalized data, and the designed sample data is normalized. The purpose of normalization is to process the sample data into data between [0,1 ]. The normalization formula is as follows:
next, the network is trained. In the training process, values such as a learning rate, a gradient threshold value and training times are determined. The learning rate is used as an important neural network parameter, and determines the convergence rate of the whole network in a certain sense, and if the learning rate is larger than that of the current neural network system, the whole neural network system can not be converged finally. However, the learning rate is low and poor, so that the training times are increased, and the system convergence time is greatly prolonged. From the combination of these conditions, the learning rate needs to be considered from the combination of sample data, and the specific rule is to select from a larger learning rate and a smaller learning rate until the final convergence of the whole neural network. Of course, if the convergence rate becomes too slow, it should be tried to get closer to a larger value appropriately, and finally the learning rate most suitable for the current network is obtained. In the case of past literature experience, the learning rate was generally selected between 0 and 0.2. The selection of the training times also affects the overall convergence time of the convolutional neural network, if the selected sample data is very large, the calculation amount may be too large to increase the time cost, and if the selected training times are too small, the final training result may not be converged, so it is very important to select an appropriate training time. The selection of the training target is set according to the error range allowed in actual production, and whether the error is within an acceptable range cannot be judged according to the standard of the training target. After training to the range acceptable in actual production, the training target can be used as the training target of the user. The training parameters of the present invention are shown in table 2:
TABLE 2 training parameters
After the training is completed, the load value of the power system is predicted, and the prediction result is shown in fig. 6, wherein the blue line is the actual load value, and the red line is the predicted load value of the model of the invention. And using the mean absolute error MAE:
mean absolute percent error MAPE
And Root Mean Square Error (RMSE) e (RMSE):
where actual (t) is the t-th real value, and forecast (t) is the t-th predicted value.
The prediction results of the present invention were evaluated and the results are shown in table 3:
TABLE 3 prediction result evaluation Table
Claims (6)
1. A short-term load prediction method of a power system based on a genetic algorithm and an improved depth residual error network is characterized by comprising the following specific steps:
acquiring power system load data of a certain power grid company within a week, considering the influence of weather characteristics of the current day on the power system load data, and performing data visualization on the weather characteristics of the current day;
secondly, carrying out structure improvement on the deep residual error network based on the deep residual error network, determining a power system load prediction network topological structure, and adding an attention mechanism;
thirdly, adjusting the network topology structure determined in the second step by using a genetic algorithm;
and fourthly, dividing the data of one week obtained in the first step into a training set and a testing set, training and testing the network in the third step, and predicting the final load data value of the power system.
2. The method for predicting the short-term load of the power system based on the genetic algorithm and the improved deep residual error network as claimed in claim 1, wherein the time interval of the acquired load data of the power system is two hours, and the weather and meteorological feature data are classified into three categories, which comprises: the highest air temperature and the lowest air temperature of the time period in which each datum is located, and the weather characteristic value of the day.
3. The power system short-term load prediction method based on the genetic algorithm and the improved deep residual error network as claimed in claim 2, wherein the weather characteristic value is 0 in a sunny day, 0.5 in a cloudy day, and 1 in a rainy day.
4. The power system short-term load prediction method based on genetic algorithm and improved depth residual error network as claimed in claim 1, wherein the depth residual error network in the second step is ResNet50, the module for improvement belongs to an inclusion network, and the added attention mechanism is a channel attention mechanism and a space attention mechanism.
5. The method for predicting short-term load of power system based on genetic algorithm and improved deep residual error network as claimed in claim 1, wherein the adjustment in the third step is to initial weight and threshold length.
6. The method for predicting the short-term load of the power system based on the genetic algorithm and the improved deep residual error network as claimed in claim 1, wherein the data set is divided in four steps, the load data of the first six days of the week is a training set, and the load data of the last day is a testing set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129326.9A CN114611757A (en) | 2022-02-11 | 2022-02-11 | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129326.9A CN114611757A (en) | 2022-02-11 | 2022-02-11 | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114611757A true CN114611757A (en) | 2022-06-10 |
Family
ID=81858360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210129326.9A Pending CN114611757A (en) | 2022-02-11 | 2022-02-11 | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114611757A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070676A (en) * | 2023-03-28 | 2023-05-05 | 南京气象科技创新研究院 | Expressway road surface temperature forecasting method based on attention mechanism and self-encoder |
| CN117200208A (en) * | 2023-09-12 | 2023-12-08 | 河海大学 | User-level short-term load prediction method and system based on multi-scale component feature learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110854891A (en) * | 2019-11-08 | 2020-02-28 | 中国农业大学 | Power distribution network pre-disaster resource allocation method and system |
| CN113160115A (en) * | 2021-02-01 | 2021-07-23 | 浙江农林大学 | Crop disease identification method and system based on improved depth residual error network |
| CN113449610A (en) * | 2021-06-08 | 2021-09-28 | 杭州格像科技有限公司 | Gesture recognition method and system based on knowledge distillation and attention mechanism |
| CN113468531A (en) * | 2021-07-15 | 2021-10-01 | 杭州电子科技大学 | Malicious code classification method based on deep residual error network and mixed attention mechanism |
-
2022
- 2022-02-11 CN CN202210129326.9A patent/CN114611757A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110854891A (en) * | 2019-11-08 | 2020-02-28 | 中国农业大学 | Power distribution network pre-disaster resource allocation method and system |
| CN113160115A (en) * | 2021-02-01 | 2021-07-23 | 浙江农林大学 | Crop disease identification method and system based on improved depth residual error network |
| CN113449610A (en) * | 2021-06-08 | 2021-09-28 | 杭州格像科技有限公司 | Gesture recognition method and system based on knowledge distillation and attention mechanism |
| CN113468531A (en) * | 2021-07-15 | 2021-10-01 | 杭州电子科技大学 | Malicious code classification method based on deep residual error network and mixed attention mechanism |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116070676A (en) * | 2023-03-28 | 2023-05-05 | 南京气象科技创新研究院 | Expressway road surface temperature forecasting method based on attention mechanism and self-encoder |
| CN116070676B (en) * | 2023-03-28 | 2023-06-30 | 南京气象科技创新研究院 | Expressway road surface temperature forecasting method based on attention mechanism and self-encoder |
| CN117200208A (en) * | 2023-09-12 | 2023-12-08 | 河海大学 | User-level short-term load prediction method and system based on multi-scale component feature learning |
| CN117200208B (en) * | 2023-09-12 | 2024-04-19 | 河海大学 | User-level short-term load prediction method and system based on multi-scale component feature learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN110705743B (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
| CN114611757A (en) | Electric power system short-term load prediction method based on genetic algorithm and improved depth residual error network | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN111861013A (en) | Power load prediction method and device | |
| CN113762387B (en) | Multi-element load prediction method for data center station based on hybrid model prediction | |
| Chen et al. | MultiCycleNet: multiple cycles self-boosted neural network for short-term electric household load forecasting | |
| CN114021848A (en) | Generating capacity demand prediction method based on LSTM deep learning | |
| CN113011680A (en) | Power load prediction method and system | |
| CN115796004A (en) | Photovoltaic power station ultra-short term power intelligent prediction method based on SLSTM and MLSTNet models | |
| Jønler et al. | Probabilistic forecasts of global horizontal irradiance for solar systems | |
| CN115481788B (en) | Phase change energy storage system load prediction method and system | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| Xu et al. | Short-term electricity consumption forecasting method for residential users based on cluster classification and backpropagation neural network | |
| CN116894504A (en) | Wind power cluster power ultra-short-term prediction model establishment method | |
| CN110659775A (en) | LSTM-based improved electric power short-time load prediction algorithm | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| Yong et al. | Short-term building load forecasting based on similar day selection and LSTM network | |
| Zhao et al. | Short-term photovoltaic power prediction based on DE-GWO-LSTM | |
| CN114707422A (en) | Intelligent power check method based on load prediction | |
| Prashanthi et al. | A comparative study of the performance of machine learning based load forecasting methods | |
| Liu et al. | Ultra-short-term wind power forecasting based on stacking model | |
| Tian et al. | Wind Speed Prediction based on Long Short-Term Memory network and Policy Suggestions for Wind Power Industry Development | |
| Wang et al. | The Research of Short-term Electric Load Forecasting based on Machine Learning Algorithm | |
| CN116523148B (en) | Distribution network distribution transformer overload early warning method, device and equipment |
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 |