CN115859792A - Medium-term power load prediction method and system based on attention mechanism - Google Patents
Medium-term power load prediction method and system based on attention mechanism Download PDFInfo
- Publication number
- CN115859792A CN115859792A CN202211477802.2A CN202211477802A CN115859792A CN 115859792 A CN115859792 A CN 115859792A CN 202211477802 A CN202211477802 A CN 202211477802A CN 115859792 A CN115859792 A CN 115859792A
- Authority
- CN
- China
- Prior art keywords
- data
- power load
- attention
- module
- medium
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a mid-term power load prediction method based on an attention mechanism, which comprises the steps of obtaining historical load data and processing the historical load data to obtain a training data set; constructing a medium-term power load prediction initial model; adopting a training data set to predict an initial model of the power load in the training middle period to obtain a power load prediction model in the middle period; and adopting a medium-term power load prediction model to predict the actual medium-term power load of the power system. The invention also discloses a system for realizing the attention mechanism-based medium-term power load prediction method. The invention not only realizes the middle-term power load prediction based on the attention mechanism through the innovative model structure design, but also is particularly suitable for multivariable time sequence data with unknown coupling relation, such as power system load data, and has high reliability, good accuracy and objective science.
Description
Technical Field
The invention belongs to the field of electrical automation, and particularly relates to a mid-term power load prediction method and system based on an attention mechanism.
Background
With the development of economic technology and the improvement of living standard of people, electric energy becomes essential secondary energy in production and life of people, and brings endless convenience to production and life of people. Therefore, ensuring stable and reliable supply of electric energy is one of the most important tasks of the power system.
Load prediction of an electric power system is one of important tasks of the electric power system. The method has the advantages of accurate and reliable load prediction results, and capability of effectively helping the power system to make short-term, medium-term and long-term operation plans, capacity-increasing and extension plans of the power system and the like. Therefore, load prediction of the power system has been one of the research focuses of researchers.
Load prediction of a power system is generally classified into ultra-short-term prediction, medium-term prediction, and long-term prediction. The medium-term load prediction of the power system plays an important role in ensuring the daily operation of the power system and optimizing a power generation plan and a scheduling plan. At present, the medium-term load prediction method of the power system mainly comprises a traditional statistical method, a machine learning method and a deep learning method. Conventional statistical methods include various statistical-based linear time series prediction methods; however, when the methods are used for establishing the model, rich experience is needed, high requirements on data stability are required, and the method is very sensitive to disturbance. Compared with the traditional statistical method, the method based on machine learning uses an artificial neural network, a support vector machine, principal component analysis and the like to predict the power load; however, since the power load time series data is a relatively unstable and random time series, the machine learning method has poor processing capability for the time series data, and thus cannot obtain optimal prediction performance. At present, various deep learning networks are available for predicting the power load, such as a cyclic neural network, a long-short term memory and a gated cyclic unit, etc. to capture the time characteristics of a multivariate time sequence, a convolutional neural network to fit the space characteristics among multivariate, etc.; although the above deep learning algorithm can be used to extract the time characteristics of multivariate time series data, and the method also considers the relationship between variables, the limitation of the method is that: the input to the scheme must be standard 2D or 3D mesh data; however, since the coupling relationship between the power load data and other variable data is unknown, and there may be a relationship between any two variable pairs, such a complex dependency relationship cannot be described by euclidean space data (grid data); therefore, the current deep learning algorithm is not suitable for the multivariate time series with unknown coupling relation of the power system data; therefore, the existing deep learning method still has the defects of low reliability and poor accuracy when processing the load data of the power system.
Disclosure of Invention
The invention aims to provide a middle-term power load prediction method based on an attention mechanism, which has high reliability and good accuracy and is objective and scientific.
The invention also aims to provide a system for realizing the attention-based medium-term power load prediction method.
The invention provides a middle-term power load prediction method based on an attention mechanism, which comprises the following steps:
s1, acquiring historical load data of a power system, and processing the historical load data to obtain a training data set;
s2, constructing a medium-term power load prediction initial model based on time convolution and space convolution;
s3, training the intermediate-term power load prediction initial model constructed in the step S2 by adopting the training data set obtained in the step S1 to obtain an intermediate-term power load prediction model;
and S4, adopting the medium-term power load prediction model obtained in the step S3 to predict the actual medium-term power load of the power system.
The processing of step S1 specifically includes the following steps:
normalization is performed using the following equation:
in the formula X i ' is the normalized ith variable; x i Is the ith variable before normalization; min (X) i ) Is a variable X i A minimum value in time sequence; max (X) i ) Is a variable X i The maximum value in the time sequence; i is equal to i =1, 2.
S2, constructing a medium-term power load prediction initial model based on the time convolution and the space convolution, and specifically comprising the following steps:
the medium-term power load prediction initial model comprises three space-time diagram attention modules and a data reading module; three time-space diagram attention modules and a data reading module are sequentially connected in series;
the space-time diagram attention module is used for capturing the correlation among the timestamps in a self-adaptive mode, focusing attention on the change of local context, and calculating the attention of different subspaces to obtain prediction data;
and the data reading module is used for carrying out inverse normalization processing on the obtained prediction data to obtain final medium-term power load prediction data.
The structure of the attention module of the space-time diagram is as follows:
the space-time diagram attention module comprises a time convolution layer and a space convolution layer;
the time convolution layer comprises a gated attention unit and a convolution attention unit; the gated attention unit adopts a gated linear unit so as to adaptively capture the correlation between the timestamps; the convolution attention unit is used for calculating a variable query vector Q, a key K and a value V by adopting a one-dimensional convolution kernel with the length larger than 1, so that attention is focused on the change of local context, and more features are matched; the formula for calculating the time convolution layer is as follows:
in the formula h GCAU (X t ) Is the output of the time convolution layer at time t; x t Is the input of the time convolution layer at time t; w T Is a first parameter to be learned; c. C T Is a second parameter to be learned; u shape T Is a third parameter to be learned; b T Is a fourth parameter to be learned;
is the Hadamard product; σ () is a sigmoid activation function; a. The CA Is according to X t Calculating the resulting convolution attention value, an
softmax (. Cndot.) is a normalized exponential function, H CA For the number of heads of attention, relu (. Cndot.) is the activation function, F 1 (. Cndot.) and F 2 (. Are affine transformation functions;
the space convolution layer adopts a multi-head self-attention unit; the prediction of the medium-term power load is completed by adopting a space convolution layer, and the calculation formula is
Wherein->
Is node v at layer k +1 i+1 Is a characteristic variable, | | is a matrix splicing operation, N (v) i ) Is a node v i A () is an attention function and
att ij is an intermediate variable and alpha tt ij ＝Leaky relu(e T (W G X i,t ||W G X j,t ) Leaky relu (-) is an activation function, e is a unit vector, W G For the parameter to be learned, X i,t For node v at time t i Time series data of representation, X j,t For node v at time t j Representative timing data.
The data reading module specifically comprises:
the data reading module performs inverse normalization processing by adopting the following formula:
in the formula
Predicting data for the denormalized power load; y is gat Noting the output data of the module for the last space-time diagram; and Y is power load prediction data.
The invention also discloses a system for realizing the method for predicting the middle-term power load based on the attention mechanism, which specifically comprises a data acquisition module, a model construction module, a model training module and a load prediction module; the data acquisition module, the model construction module, the model training module and the load prediction module are sequentially connected in series; the data acquisition module is used for acquiring and processing historical load data of the power system to obtain a training data set and uploading the data to the model construction module; the model construction module is used for constructing a medium-term power load prediction initial model based on time convolution and space convolution according to the acquired data and uploading the data to the model training module; the model training module is used for training the medium-term power load prediction initial model by adopting training data according to the acquired data to obtain a medium-term power load prediction model and uploading the data to the load prediction module; and the load prediction module is used for predicting the actual medium-term power load of the power system by adopting a medium-term power load prediction model according to the acquired data.
The method and the system for predicting the middle-term power load based on the attention mechanism not only realize the middle-term power load prediction based on the attention mechanism through the innovative model structure design, but also are particularly suitable for multivariable time series data with unknown coupling relation, such as power system load data, and have the advantages of high reliability, good accuracy and objective science.
Drawings
FIG. 1 is a schematic process flow diagram of the process of the present invention.
Fig. 2 is a schematic diagram of training loss on a training set after 1000 model trainings according to an embodiment of the method of the present invention.
FIG. 3 is a comparison of predicted results for embodiments of the method of the present invention.
FIG. 4 is a functional block diagram of the system of the present invention.
Detailed Description
FIG. 1 is a schematic flow chart of the method of the present invention: the invention provides a middle-term power load prediction method based on an attention mechanism, which comprises the following steps:
s1, acquiring historical load data of a power system, and processing the historical load data to obtain a training data set;
in specific implementation, the following steps are adopted for processing:
normalization is performed using the following equation:
in the formula X i ' is the normalized ith variable; x i Is the ith variable before normalization; min (X) i ) Is a variable X i A minimum value in time sequence; max (X) i ) Is a variable X i A maximum value in time sequence; the value of i is i =1,2, N is the number of the prediction variables;
s2, constructing a medium-term power load prediction initial model based on time convolution and space convolution; the method specifically comprises the following steps:
the medium-term power load prediction initial model comprises three space-time diagram attention modules and a data reading module; three time-space diagram attention modules and a data reading module are sequentially connected in series;
the space-time diagram attention module is used for capturing the correlation among the timestamps in a self-adaptive mode, focusing attention on the change of local context, and calculating the attention of different subspaces to obtain prediction data;
the data reading module is used for carrying out reverse normalization processing on the obtained prediction data to obtain final medium-term power load prediction data;
in specific implementation, the structure of the space-time diagram attention module is as follows:
the space-time diagram attention module comprises a time convolution layer and a space convolution layer;
the time convolution layer comprises a gated attention unit and a convolution attention unit; the gated attention unit adopts a gated linear unit so as to adaptively capture the correlation between the timestamps; the convolution attention unit is used for calculating a variable query vector Q, a key K and a value V by adopting a one-dimensional convolution kernel with the length larger than 1, so that attention is focused on the change of local context, and more features are matched; the formula for calculating the time convolution layer is as follows:
in the formula h GCAU (X t ) Is the output of the time convolution layer at time t; x t Input for the time convolution layer at time t; w T Is a first parameter to be learned; c. C T Is a second parameter to be learned; u shape T Is a third parameter to be learned; b is a mixture of T Is a fourth parameter to be learned;
is the Hadamard product; sigma () is a sigmoid activation function; a. The CA Is according to X t Calculating the resulting convolution attention value, an
softmax (. Cndot.) is a normalized exponential function, H CA For the number of heads of attention, relu (. Circle.) is the activation function, F 1 (. And F) 2 (. Are affine transformation functions;
the space convolution layer adopts a multi-head self-attention unit; the prediction of the medium-term power load is completed by adopting a space convolution layer, and the calculation formula is
(the equation shows that for the same node, it is necessary to calculate H attention separately and merge the H attention matrices in a concatenation manner), where ^ H>
Is node v at layer k +1 i+1 Is a characteristic variable, | | is a matrix splicing operation, N (v) i ) Is a node v i A () is an attention function and
att ij is an intermediate variable and alpha tt ij ＝Leaky relu(e T (W G X i,t ||W G X j,t ) Leaky relu (-) is an activation function, e is a unit vector, W G For the parameter to be learned, X i,t Node v at time t i Time series data of representation, X j,t For node v at time t j Representative time series data; />
The data reading module specifically comprises:
the data reading module adopts the following formula to perform inverse normalization processing:
in the formula
Predicting data for the power load after the denormalization; y is gat Noting the output data of the module for the last space-time diagram; y is power load prediction data;
s3, training the intermediate-term power load prediction initial model constructed in the step S2 by adopting the training data set obtained in the step S1 to obtain an intermediate-term power load prediction model;
and S4, adopting the medium-term power load prediction model obtained in the step S3 to predict the actual medium-term power load of the power system.
The effect of the process according to the invention is illustrated below with reference to specific examples:
performing time sequence prediction on Australian power load data; the application object of the example is Australian power load data set, and the experimental data is composed of the following variables [ dry bulb temperature, dew point temperature, wet bulb temperature, humidity, electricity price, power load [ ]]Therefore, the network input is history data X = [ X ] with length h before time t gtem ，X ltem ，X stem ，X h ，X pri ，X ele ]Predicted power load Y of length p after time t as output of network gat Power load Y = [ X ] having real value and length p after time t ele ](ii) a The experiment mainly comprises the following steps:
data preprocessing:
the australian power load data set contains the total australian power load from 1/2006 to 1/2011, with a length of 87649, a sampling frequency of once an hour, according to 7:1:2, dividing the training data, the verification data and the test data to obtain a training set, a verification set and a test set with lengths of 61354, 8765 and 17530 respectively;
firstly, carrying out normalization processing on a training set, and storing the maximum and minimum values of each variable of the training set; normalizing the verification data set by using the maximum and minimum values of all variables of the training set, and finally normalizing the input data X of the test data set by using the maximum and minimum values of all variables of the training set, so as to keep the real power load Y of the test data set without normalization;
model training:
this method was implemented in model testing using the pytorech version 1.12.0 of CUDA 11.3. Inputting normalized historical data X with the time sequence length of 30 multiplied by 24 multiplied by 7 by the model, wherein the size of a sliding window is 24 multiplied by 7, the step length is 24, and the number of heads of attention of the graph is 7; in addition, a Dropout strategy is applied, the Dropout rate is set to be 0.3, the learning rate during training is set to be 1e-3, the number of training iterations is set to be 1000, and the batch size is set to be 64; acquiring the gradient of the network error for each weight parameter by using an Adam optimization algorithm, and acquiring a new weight through a parameter updating process; normalized prediction data Y with model output timing length of 30 × 24 × 7 gat The loss function for model training is defined as
And after each round of training of the data set, using the verification set for verification, and storing the model at the moment when the model obtains the optimal result on the verification set. And when the training times are reached or the prediction result of the verification set is converged to be smaller than a preset convergence threshold value, finishing the training. The present invention plots the training loss obtained after each training for 5 epochs, as shown in FIG. 2;
model prediction:
inputting the test data set into a trained time-space diagram attention module to obtain output data Y gat Use trainingCollecting maximum and minimum value pairs Y of each variable gat Inverse normalization is carried out to obtain final prediction data
Can be picked up by a pair>
The processing of (a) results in an electrical load per hour, per day and per week. The present invention plots the predicted and true values of the power load simultaneously in figure 3. Since the length of data per hour is as long as 17530, it is inconvenient to draw all the data, and therefore only the data of the first 2000 are cut out to draw.
The performance of the prediction effect of the present invention is evaluated using four evaluation indexes, including an average absolute error (MAE), a Root Mean Square Error (RMSE), an average absolute percentage error (MAPE), and a Correlation Coefficient (CC). The lower the values of RMSE, MAE and MAPE, the better the prediction of the model. The higher the value of CC, the better the prediction effect of the model.
The four indices are calculated as:
the results of the final calculation are shown in table 1:
TABLE 1 Experimental results schematic table
As can be seen from the data in Table 1 and the curve in FIG. 3, the fitting performance of the prediction data and the actual data of the method of the present invention is very good, which shows that the prediction effect of the method of the present invention is good.
FIG. 4 is a schematic diagram of functional modules of the system of the present invention: the system for realizing the attention-based medium-term power load prediction method specifically comprises a data acquisition module, a model construction module, a model training module and a load prediction module; the data acquisition module, the model construction module, the model training module and the load prediction module are sequentially connected in series; the data acquisition module is used for acquiring and processing historical load data of the power system to obtain a training data set and uploading the data to the model construction module; the model construction module is used for constructing a medium-term power load prediction initial model based on time convolution and space convolution according to the acquired data and uploading the data to the model training module; the model training module is used for training the medium-term power load prediction initial model by adopting training data according to the acquired data to obtain a medium-term power load prediction model and uploading the data to the load prediction module; and the load prediction module is used for predicting the actual medium-term power load of the power system by adopting a medium-term power load prediction model according to the acquired data.
Claims (6)
1. An attention mechanism-based medium-term power load prediction method comprises the following steps:
s1, acquiring historical load data of a power system, and processing the historical load data to obtain a training data set;
s2, constructing a medium-term power load prediction initial model based on time convolution and space convolution;
s3, training the intermediate-term power load prediction initial model constructed in the step S2 by adopting the training data set obtained in the step S1 to obtain an intermediate-term power load prediction model;
and S4, adopting the medium-term power load prediction model obtained in the step S3 to predict the actual medium-term power load of the power system.
2. The method of claim 1, wherein the step S1 of processing comprises the following steps:
normalization is performed using the following equation:
in the formula X i ' is the normalized ith variable; x i Is the ith variable before normalization; min (X) i ) Is a variable X i A minimum value in time sequence; max (X) i ) Is a variable X i The maximum value in the time sequence; the value of i is i =1, 2.. N, N is the number of the prediction variables.
3. The attention mechanism-based medium term power load prediction method according to claim 2, wherein the step S2 of constructing a medium term power load prediction initial model based on the time convolution and the space convolution specifically comprises the following steps:
the medium-term power load prediction initial model comprises three space-time diagram attention modules and a data reading module; three time-space diagram attention modules and a data reading module are sequentially connected in series;
the space-time diagram attention module is used for capturing the correlation among the timestamps in a self-adaptive mode, focusing attention on the change of local context, and calculating the attention of different subspaces to obtain prediction data;
and the data reading module is used for carrying out inverse normalization processing on the obtained prediction data to obtain final medium-term power load prediction data.
4. The attention mechanism-based medium term power load forecasting method of claim 3, wherein the space-time diagram attention module is configured as follows:
the space-time diagram attention module comprises a time convolution layer and a space convolution layer;
the time convolution layer comprises a gated attention unit and a convolution attention unit; the gated attention unit adopts a gated linear unit so as to adaptively capture the correlation between the timestamps; the convolution attention unit is used for calculating a variable query vector Q, a key K and a value V by adopting a one-dimensional convolution kernel with the length larger than 1, so that attention is focused on the change of local context, and more features are matched; the formula for the time convolution layer is:
in the formula h GCAU (X t ) Output of the time convolution layer at time t; x t Is the input of the time convolution layer at time t; w T Is a first parameter to be learned; c. C T Is a second parameter to be learned; u shape T Is a third parameter to be learned; b T Is the fourth parameter to be learned;
is the Hadamard product; sigma () is a sigmoid activation function; a. The CA Is according to X t Calculating the resulting convolution attention value, an
softmax (. Cndot.) is a normalized exponential function, H CA For the number of heads of attention, relu (. Circle.) is the activation function, F 1 (. And F) 2 (. Cndot.) is an affine transformation function; />
The space convolution layer adopts a multi-head self-attention unit; the prediction of the medium-term power load is completed by adopting a space convolution layer, and the calculation formula is
Wherein->
Is node v at layer k +1 i+1 Is the characteristic variable of (1), i is the matrix splicing operation, N (v) i ) Is a node v i A () is an attention function and
att ij is an intermediate variable and alpha tt ij ＝Leaky relu(e T (W G X i,t ||W G X j,t ) Leaky relu (-) is an activation function, e is a unit vector, W G For the parameter to be learned, X i,t Node v at time t i Time series data of representation, X j,t For node v at time t j Representative timing data.
5. The attention mechanism-based medium term power load forecasting method of claim 4, wherein the data reading module specifically comprises:
the data reading module adopts the following formula to perform inverse normalization processing:
in the formula
Predicting data for the power load after the denormalization; y is gat Noting the output data of the module for the last space-time diagram; and Y is power load prediction data.
6. A system for realizing the attention mechanism-based medium-term power load prediction method is characterized by specifically comprising a data acquisition module, a model construction module, a model training module and a load prediction module; the data acquisition module, the model construction module, the model training module and the load prediction module are sequentially connected in series; the data acquisition module is used for acquiring and processing historical load data of the power system to obtain a training data set and uploading the data to the model construction module; the model construction module is used for constructing a medium-term power load prediction initial model based on time convolution and space convolution according to the acquired data and uploading the data to the model training module; the model training module is used for training the medium-term power load prediction initial model by adopting training data according to the acquired data to obtain a medium-term power load prediction model and uploading the data to the load prediction module; and the load prediction module is used for predicting the actual medium-term power load of the power system by adopting a medium-term power load prediction model according to the acquired data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211477802.2A CN115859792A (en) | 2022-11-23 | 2022-11-23 | Medium-term power load prediction method and system based on attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211477802.2A CN115859792A (en) | 2022-11-23 | 2022-11-23 | Medium-term power load prediction method and system based on attention mechanism |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115859792A true CN115859792A (en) | 2023-03-28 |
Family
ID=85665590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211477802.2A Pending CN115859792A (en) | 2022-11-23 | 2022-11-23 | Medium-term power load prediction method and system based on attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115859792A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116757369A (en) * | 2023-08-22 | 2023-09-15 | 国网山东省电力公司营销服务中心(计量中心) | Attention mechanism-based carbon emission analysis method and system |
| CN116865261A (en) * | 2023-07-19 | 2023-10-10 | 王克佳 | Power load prediction method and system based on twin network |
-
2022
- 2022-11-23 CN CN202211477802.2A patent/CN115859792A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116865261A (en) * | 2023-07-19 | 2023-10-10 | 王克佳 | Power load prediction method and system based on twin network |
| CN116865261B (en) * | 2023-07-19 | 2024-03-15 | 梅州市嘉安电力设计有限公司 | Power load prediction method and system based on twin network |
| CN116757369A (en) * | 2023-08-22 | 2023-09-15 | 国网山东省电力公司营销服务中心(计量中心) | Attention mechanism-based carbon emission analysis method and system |
| CN116757369B (en) * | 2023-08-22 | 2023-11-24 | 国网山东省电力公司营销服务中心(计量中心) | Attention mechanism-based carbon emission analysis method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115859792A (en) | Medium-term power load prediction method and system based on attention mechanism | |
| CN112016734A (en) | Stack type self-coding multi-model load prediction method and system based on LSTM | |
| CN110443724B (en) | Electric power system rapid state estimation method based on deep learning | |
| de Guia et al. | Using stacked long short term memory with principal component analysis for short term prediction of solar irradiance based on weather patterns | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| AU2021106200A4 (en) | Wind power probability prediction method based on quantile regression | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN116774089A (en) | Convolutional neural network battery state of health estimation method and system based on feature fusion | |
| CN116402194A (en) | Multi-time scale load prediction method based on hybrid neural network | |
| CN114862032B (en) | XGBoost-LSTM-based power grid load prediction method and device | |
| CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
| CN113673846B (en) | Demand situation awareness method under electric power Internet of things architecture | |
| CN115860232A (en) | Steam load prediction method, system, electronic device and medium | |
| CN112232714B (en) | Deep learning-based risk assessment method for distribution network under incomplete structural parameters | |
| CN114372418A (en) | Wind power space-time situation description model establishing method | |
| CN113780644A (en) | Photovoltaic output prediction method based on online learning | |
| Li et al. | A Novel Short-term Load Forecasting Model by TCN-LSTM Structure with Attention Mechanism | |
| CN117833243B (en) | Method, system, equipment and medium for predicting short-term demand of electric power | |
| Wu et al. | Optimizing CNN-LSTM Model for Short-term PV Power Prediction using Northern Goshawk Optimization | |
| CN117407681B (en) | Time sequence data prediction model establishment method based on vector clustering | |
| Qian et al. | Improved Attention-based AE-LSTM for Short-term Power Load Forecasting | |
| CN116129212A (en) | Power load prediction method and system based on graph neural network | |
| CN117613849A (en) | Combined prediction method and device for regional load of power system and storage medium | |
| CN116467580A (en) | SOH estimation and RUL prediction method of lithium ion battery based on TTSNet | |
| Zhang et al. | Evaluation and Research of Power Forecasting Based on LSTM Time Series |
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 |