A short-term power load forecasting method combining weather parameters

Abstract

This invention discloses a short-term power load forecasting method incorporating weather parameters, comprising: acquiring power load data and corresponding time-related weather parameters; preprocessing the data; using the EEMD method to decompose the power load data into high-frequency nonlinear components and low-frequency linear components; training a CNN-LSTM model based on multi-scale feature extraction with an attention mechanism, incorporating weather parameters, and outputting predicted values ​​for the high-frequency nonlinear components; for the low-frequency linear components, employing the SARIMA time series model, and outputting predicted values ​​for the low-frequency linear components; adding the predicted values ​​for the high-frequency nonlinear components and the low-frequency linear components to obtain a complete load forecast result, performing error analysis, and iteratively optimizing the model. This invention improves the forecasting capability for complex load fluctuations and the accuracy and reliability of short-term load forecasting by introducing weather data, extracting high-frequency nonlinear components through sequence decomposition, and using weather parameters as input to construct a deep learning forecasting model.

Description

Technical Field

[0001] This invention belongs to the field of power load forecasting, and relates to short-term power load forecasting technology, specifically to a short-term power load forecasting method that combines weather parameters. Background Technology

[0002] With the continuous development of science and technology and the social economy, my country's power industry system has also experienced tremendous growth. The continuous development of my country's power system and the increasing modernization and complexity of power grid management have drawn increasing attention to the research on short-term power system load forecasting. The power system has accumulated a vast amount of electricity consumption data. Data mining of this massive amount of data can yield valuable information, providing data support for diversified power grid services. Forecasting electricity consumption data is an important application of time series forecasting.

[0003] Short-term electricity load is related to many factors, including weather, economy, day type, and society. Since people's electricity consumption behavior changes with their perception of environmental comfort (e.g., turning on air conditioning for cooling on hot days and heating on cold days), weather parameters have a significant impact on electricity load. Statistics show a clear positive correlation between summer temperature and load, and temperature has a cumulative effect on load; several consecutive days of high temperatures will alter electricity load characteristics. Therefore, short-term electricity load forecasting methods incorporating weather parameters are worth considering. The relationship between weather parameters and electricity load is complex and nonlinear. Traditional methods have failed to fully exploit this nonlinear relationship, limiting forecast accuracy. Current load forecasting methods often employ time series analysis, statistical models, or direct modeling based on deep learning. These methods often fall short when dealing with complex nonlinear fluctuations and high-frequency changes in load, making it difficult to accurately capture the dynamic characteristics of the load. Summary of the Invention

[0004] Purpose of the invention: To address the shortcomings of existing technologies in handling complex nonlinear fluctuations and high-frequency changes in load, which often make it difficult to accurately capture the dynamic characteristics of load, this invention provides a short-term power load forecasting method that incorporates weather parameters. By extracting high-frequency nonlinear components through sequence decomposition and using weather parameters as input to construct a deep learning forecasting model, this invention improves the forecasting capability for complex load fluctuations and the accuracy and reliability of short-term load forecasting.

[0005] Technical Solution: To achieve the above objectives, this invention provides a short-term power load forecasting method that incorporates weather parameters, comprising the following steps:

[0006] S1: Obtain power load data and corresponding weather parameters within a preset time period, including temperature, humidity, rainfall, wind speed, and weather type (sunny, rainy, snowy, etc.);

[0007] S2: Preprocess the data to ensure data quality and consistency;

[0008] S3: The integrated empirical mode decomposition (EEMD) method is used to decompose the power load data into high-frequency nonlinear components and low-frequency linear components.

[0009] S4: Based on high-frequency nonlinear components and combined with weather parameters, train a CNN-LSTM model based on multi-scale feature extraction with an attention mechanism, and output the predicted value of high-frequency nonlinear components for the time period to be predicted.

[0010] S5: For low-frequency linear components, the time series model SARIMA is used to capture the long-term trend of low-frequency linear components and output the predicted value of low-frequency linear components.

[0011] S6: Add the predicted values ​​of the high-frequency nonlinear components to the predicted values ​​of the low-frequency linear components to obtain the complete load prediction result. Perform error analysis on the prediction results and the actual values, and iteratively optimize the model.

[0012] S7: Output the power load prediction results through the optimized model.

[0013] Furthermore, in step S1, the power load data obtained from historical data includes historical temperature, humidity, and rainfall data obtained from the meteorological service platform, and the timestamps are aligned with the load data.

[0014] Furthermore, the preprocessing in step S2 includes cleaning, completion, noise reduction, and standardization of the data.

[0015] Furthermore, the method for decomposing the power load data in step S3 is as follows:

[0016] A1: EEMD (Integrated Empirical Mode Decomposition) improves the robustness and stability of EMD (Empirical Mode Decomposition) by adding random Gaussian white noise.

[0017] Let the original signal be X(t), and the signal after adding noise be:

[0018] X i (t)=X(t)+∈ i (t)

[0019] Where, ∈ i (t) is the white noise superimposed for the i-th time;

[0020] Each signal was decomposed using EMD to obtain a set of Intrinsic Mode Functions (IMFS). The decomposition results are as follows:

[0021]

[0022] Among them, IMF ij (t) is the j-th IMF component of the i-th signal, R i (t) is the residual term;

[0023] By superimposing different types of white noise and repeating the above steps multiple times, the IMFs of all noise versions are averaged to eliminate the influence of noise on the results; the averaged IMFs are:

[0024]

[0025] Where M is the number of noise superpositions;

[0026] A2: After obtaining the IMF (Information Families), zero-crossing rate detection is used to analyze the frequency characteristics of each IMF, especially the high-frequency components. The zero-crossing rate can be calculated using the following formula:

[0027]

[0028] Where N is the number of samples, and x[n] is the sample value in the IMF.

[0029] The EEMD method is used to decompose the power load sequence, and the expression is as follows:

[0030] L(t)=L inear (t)+L nonlinear (t)

[0031] Where: L inear (t) is the low-frequency linear component, representing the long-term trend; L nonlinear (t) is a high-frequency nonlinear component, representing the rapid fluctuation characteristics of the load.

[0032] Furthermore, the modeling and prediction steps for the high-frequency nonlinear components in step S4 are as follows:

[0033] B1: The high-frequency nonlinear component L nonlinear (t) and the corresponding weather parameters for the time period are used as model inputs; the weather parameters are organized into a matrix according to the time steps:

[0034] X weather ={X weather,t ,X weather,t+1 ,…,X weather,T}

[0035] Each X weather,t It is a vector of weather parameters;

[0036] The high-frequency nonlinear components are high-frequency components extracted from the power load time series using EEMD:

[0037] X high ={Xhigh,t ,X high,t+1 ,…,X high,T}

[0038] Then the weather parameters and high-frequency components are normalized to [0,1];

[0039] B2: The CNN-LSTM model based on multi-scale feature extraction consists of three main modules: CNN, LSTM, and Attention mechanism. Based on the weather parameters input in step B1, CNN extracts the local features of the weather parameters and identifies the correlation between weather and load fluctuations. Based on the high-frequency nonlinear components input in step B1, LSTM captures the time series characteristics of the high-frequency components and models their dynamic changes. The attention mechanism is used to improve the model's ability to handle key time steps and important weather parameters.

[0040] B3: Use historical power load data to construct training and validation sets, use mean squared error as the loss function, and dynamically adjust the learning rate through the Adam optimization algorithm to improve training efficiency and obtain a well-trained CNN-LSTM model.

[0041] Furthermore, in step B2, the CNN-LSTM model includes convolutional layers and pooling layers. The convolutional layers extract spatial local patterns of weather parameters, such as the coupling relationship between temperature and humidity. The pooling layers reduce feature dimensionality while retaining key patterns. The CNN extracts local features F of the weather parameters. weather It can reflect the short-term impact of temperature changes on the load, as well as the interaction of parameters such as wind speed and humidity;

[0042] F weather ={f1,f2,…,fd}

[0043] Among them, f i Let i be the i-th weather feature;

[0044] score(f i ) = W weather *f i +b weather

[0045] Among them, W weather and b weather These are the learning parameters.

[0046] Furthermore, in step B2, the CNN-LSTM model models the high-frequency load components and captures dynamic changes over time. The LSTM network model structure includes a forget gate, an input gate, and an output gate. The features H extracted by LSTM... high This represents the sequence of hidden states that generate time steps, including periodic fluctuations and random characteristics or outlier changes;

[0047] H high ={h1,h2,…,hT}

[0048] score(h t ) = O t tanh(C t )

[0049] Among them, O t It is the output gate result, C t This is the result of the unit update status.

[0050] Furthermore, the operation of the attention mechanism in the CNN-LSTM model in step B2 includes:

[0051] The importance of using fully connected networks to compute features:

[0052]

[0053] Generate attention-weighted weather features:

[0054]

[0055] Calculate the attention score using contextual information and hidden states:

[0056]

[0057] Generate attention-weighted time series features

[0058]

[0059] Features after incorporating the attention mechanism are then concatenated:

[0060] F combined =Concat(F weather,attn H high,attn )

[0061] Fully connected output:

[0062] y t =W out *F combined +b out .

[0063] The purpose of introducing an attention mechanism is to assign weights to different time steps or feature components, enabling the model to focus more on important input information. Dynamic weights are assigned to local features of weather parameters to focus on important meteorological factors, such as sudden changes in temperature or humidity. Time steps of high-frequency load components are weighted to highlight the fluctuation characteristics of key time periods.

[0064] Furthermore, in step S5, the low-frequency linear component L... linear (t) Independent predictions were made using the SARIMA model.

[0065] Furthermore, error analysis in step S6 is crucial, as it is not only a key step in evaluating model performance but also an important basis for identifying potential directions for model change. Error analysis methods include:

[0066] D1: Calculate the difference between the actual load value and the predicted load value;

[0067] D2: Commonly used evaluation metrics include mean squared error (MSE), mean absolute error (MAE), absolute percentage error (MAPE), and the absolute coefficient R, which measures the goodness of fit between the predicted and actual load values. 2 ;

[0068] D3: Using visualization tools, plot time series graphs of actual load values, predicted load values, and errors to observe the fluctuation trend of errors; use box plots to show the distribution characteristics of errors and identify the concentrated intervals of outliers.

[0069] Beneficial Effects: Compared with existing technologies, this invention combines EEMD sequence decomposition and zero-crossing rate detection methods to decompose the signal and analyze the frequency characteristics of each IMF, enabling a deeper understanding of signal changes. The extracted high-frequency nonlinear components are used as input to construct a deep learning combined prediction model. An attention mechanism is added to the CNN-LSTM model to handle the relationship between high-frequency nonlinear components and weather parameters, improving the prediction capability of complex load fluctuations and the accuracy and reliability of short-term load forecasting.

[0070] Attached Figure

[0071] Figure 1 This is a schematic flowchart of the method of the present invention;

[0072] Figure 2 This is a schematic diagram of the sequence decomposition process in this invention;

[0073] Figure 3 This is a schematic diagram of the CNN-LSTM model in this invention. Detailed Implementation

[0074] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0075] This invention provides a short-term power load forecasting method that incorporates weather parameters, such as... Figure 1 As shown, it includes the following steps:

[0076] S1: Obtain power load data and corresponding weather parameters within a preset time period, including temperature, humidity, rainfall, wind speed, and weather type (sunny, rainy, snowy, etc.);

[0077] The power load data is obtained from historical data with an hourly time granularity. Historical temperature, humidity, and rainfall data are obtained from the meteorological service platform and their timestamps are aligned with the load data.

[0078] S2: Clean, complete, denoise, and standardize the data to ensure data quality and consistency;

[0079] S3: The integrated empirical mode decomposition (EEMD) method is used to decompose the power load data into high-frequency nonlinear components and low-frequency linear components.

[0080] like Figure 2 As shown, the method for decomposing power load data is as follows:

[0081] A1: EEMD (Integrated Empirical Mode Decomposition) improves the robustness and stability of EMD (Empirical Mode Decomposition) by adding random Gaussian white noise.

[0082] Let the original signal be X(t), and the signal after adding noise be:

[0083] X i (t)=X(t)+∈ i (t)

[0084] Where, ∈ i (t) is the white noise superimposed for the i-th time;

[0085] Each signal was decomposed using EMD to obtain a set of Intrinsic Mode Functions (IMFS). The decomposition results are as follows:

[0086]

[0087] Among them, IMF ij (t) is the j-th IMF component of the i-th signal, R i (t) is the residual term;

[0088] By superimposing different types of white noise and repeating the above steps multiple times, the IMFs of all noise versions are averaged to eliminate the influence of noise on the results; the averaged IMFs are:

[0089]

[0090] Where M is the number of noise superpositions;

[0091] A2: After obtaining the IMF (Information Frequency Detection Factor), zero-crossing rate detection is used to analyze the frequency characteristics of each IMF, especially the high-frequency components. The zero-crossing rate is calculated using the following formula:

[0092]

[0093] Where N is the number of samples, and x[n] is the sample value in the IMF.

[0094] High-frequency IMFs typically exhibit higher zero-crossing rates, while low-frequency IMFs typically exhibit lower zero-crossing rates. The power load sequence can be decomposed to obtain the following expression:

[0095] L(t)=L linear (t)+L nonlinear (t)

[0096] Where: L linear (t) is the low-frequency linear component, representing the long-term trend; L nonlinear (t) is a high-frequency nonlinear component, representing the rapid fluctuation characteristics of the load.

[0097] S4: Based on high-frequency nonlinear components and combined with weather parameters, train a CNN-LSTM model based on multi-scale feature extraction with an attention mechanism, and output the predicted value of high-frequency nonlinear components for the time period to be predicted.

[0098] The steps for modeling and predicting high-frequency nonlinear components are as follows:

[0099] B1: The high-frequency nonlinear component L nonlinear (t) and the corresponding weather parameters for the time period are used as model inputs; the weather parameters are organized into a matrix according to the time steps:

[0100] X weather ={X weather,t ,X weather,t+1 ,…,X weather,T}

[0101] Each X weather,t It is a vector of weather parameters;

[0102] The high-frequency nonlinear components are high-frequency components extracted from the power load time series using EEMD:

[0103] X high ={X high,t ,X high,t+1 ,…,X high,T}

[0104] Then the weather parameters and high-frequency components are normalized to [0,1];

[0105] B2: such as Figure 3 As shown, the CNN-LSTM model based on multi-scale feature extraction consists of three main modules: CNN, LSTM, and Attention mechanism. Based on the weather parameters input in step B1, CNN extracts the local features of the weather parameters and identifies the correlation between weather and load fluctuations. Based on the high-frequency nonlinear components input in step B1, LSTM captures the time series characteristics of the high-frequency components and models their dynamic changes. The attention mechanism is used to improve the model's ability to handle key time steps and important weather parameters.

[0106] The CNN-LSTM model's CNN network structure includes convolutional layers and pooling layers. Convolutional layers extract spatial local patterns of weather parameters, such as the coupling relationship between temperature and humidity. Pooling layers reduce feature dimensionality while preserving key patterns. The CNN extracts local features F of the weather parameters. weather It can reflect the short-term impact of temperature changes on the load, as well as the interaction of parameters such as wind speed and humidity;

[0107] F wetather ={f1,f2,…,fd}

[0108] Among them, f i Let i be the i-th weather feature;

[0109] score(f i ) = W weather *f i +b weather

[0110] Among them, W weather and b weather These are the learning parameters.

[0111] In the CNN-LSTM model, the LSTM network model models high-frequency load components and captures dynamic changes over time. The LSTM network model structure includes a forget gate, an input gate, and an output gate. The features extracted by LSTM are H. high This represents the sequence of hidden states that generate time steps, including periodic fluctuations and random characteristics or outlier changes;

[0112] H high ={h1,h2,…,hT}

[0113] score(h t ) = O t tanh(C t )

[0114] Among them, O t It is the output gate result, C t This is the result of the unit update status.

[0115] The attention mechanism in a CNN-LSTM model includes:

[0116] The importance of using fully connected networks to compute features:

[0117]

[0118] Generate attention-weighted weather features:

[0119]

[0120] Calculate the attention score using contextual information and hidden states:

[0121]

[0122] Generate attention-weighted time series features

[0123]

[0124] Features after incorporating the attention mechanism are then concatenated:

[0125] F combined =Concat(F weather ,attn,H high ,attn)

[0126] Fully connected output:

[0127] y t =W out *F combined +b out .

[0128] The purpose of introducing an attention mechanism is to assign weights to different time steps or feature components, enabling the model to focus more on important input information. Dynamic weights are assigned to local features of weather parameters to focus on important meteorological factors, such as sudden changes in temperature or humidity. Time steps of high-frequency load components are weighted to highlight the fluctuation characteristics of key time periods.

[0129] B3: Use historical power load data to construct training and validation sets, use mean squared error as the loss function, and dynamically adjust the learning rate through the Adam optimization algorithm to improve training efficiency and obtain a well-trained CNN-LSTM model.

[0130] S5: For low-frequency linear components L linear (t) The SARIMA seasonal model is used for independent prediction to capture the long-term trend of the low-frequency linear component and output the predicted value of the low-frequency linear component.

[0131] S6: Construct a composite model consisting of linear and nonlinear components. Add the predicted values ​​of the high-frequency nonlinear components to the predicted values ​​of the low-frequency linear components to obtain the complete load forecast result. Perform error analysis on the predicted results and actual values, and iteratively optimize the model. This overcomes the problem that a single model cannot predict the complexity of power load sequences.

[0132] Error analysis is crucial, not only as a key step in evaluating model performance but also as an important basis for identifying potential directions for model change. Methods of error analysis include:

[0133] D1: Calculate the difference between the actual load value and the predicted load value;

[0134] D2: Commonly used evaluation metrics include mean squared error (MSE), mean absolute error (MAE), absolute percentage error (MAPE), and the absolute coefficient R, which measures the goodness of fit between the predicted and actual load values. 2 ;

[0135] D3: Using visualization tools, plot time series graphs of actual load values, predicted load values, and errors to observe the fluctuation trend of errors; use box plots to show the distribution characteristics of errors and identify the concentrated intervals of outliers.

[0136] S7: Output the power load prediction results through the optimized model.

[0137] As can be seen from the above, short-term power load forecasting is a critical task in power system operation, widely used in generation planning, load dispatching, and electricity market transactions. Power load fluctuations are influenced by numerous factors, including the inherent patterns of historical load (linear and nonlinear characteristics) and external weather conditions (temperature, humidity, etc.). This invention adopts an overall framework of "sequence decomposition + feature extraction + deep learning prediction," combining high-frequency nonlinear components with weather parameters to propose a high-precision and robust short-term power load forecasting method. This method not only captures the complex fluctuation characteristics of load but also fully leverages the driving role of weather parameters, possessing significant theoretical and engineering application value.

[0138] This embodiment also provides a short-term power load forecasting system that incorporates weather parameters. The system includes a network interface, a memory, and a processor. The network interface is used to receive and send signals during information exchange with other external network elements. The memory is used to store computer program instructions that can run on the processor. The processor is used to execute the steps of the consensus method described above when running the computer program instructions.

[0139] This embodiment also provides a computer storage medium storing a computer program that, when executed by a processor, can implement the methods described above. The computer-readable medium can be considered tangible and non-transitory. Non-limiting examples of non-transitory tangible computer-readable media include non-volatile memory circuitry (e.g., flash memory circuitry, erasable programmable read-only memory circuitry, or masked read-only memory circuitry), volatile memory circuitry (e.g., static random access memory circuitry or dynamic random access memory circuitry), magnetic storage media (e.g., analog or digital magnetic tape or hard disk drive), and optical storage media (e.g., CD, DVD, or Blu-ray disc). The computer program includes processor-executable instructions stored on at least one non-transitory tangible computer-readable medium. The computer program may also include or depend on stored data. The computer program may include a basic input / output system (BIOS) for interacting with the hardware of a dedicated computer, device drivers for interacting with specific devices of the dedicated computer, one or more operating systems, user applications, background services, background applications, etc.

[0140] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A short-term power load forecasting method incorporating weather parameters, characterized in that, Includes the following steps: S1: Obtain power load data and corresponding weather parameters within a preset time period; S2: Preprocess the data to ensure data quality and consistency; S3: The integrated empirical mode decomposition (EEMD) method is used to decompose the power load data into high-frequency nonlinear components and low-frequency linear components. S4: Based on high-frequency nonlinear components and combined with weather parameters, train a CNN-LSTM model based on multi-scale feature extraction with an attention mechanism, and output the predicted value of high-frequency nonlinear components for the time period to be predicted. S5: For low-frequency linear components, the time series model SARIMA is used to capture the long-term trend of low-frequency linear components and output the predicted value of low-frequency linear components. S6: Add the predicted values ​​of the high-frequency nonlinear components to the predicted values ​​of the low-frequency linear components to obtain the complete load prediction result. Perform error analysis on the prediction results and the actual values, and iteratively optimize the model. S7: Output the power load prediction results through the optimized model.

2. The short-term power load forecasting method combining weather parameters according to claim 1, characterized in that, In step S1, the power load data obtained from historical data is used to obtain historical temperature, temperature, and rainfall data from the meteorological service platform, and the timestamps are aligned with the load data.

3. The short-term power load forecasting method combining weather parameters according to claim 1, characterized in that, The preprocessing in step S2 includes cleaning, completion, noise reduction, and standardization of the data.

4. The short-term power load forecasting method combining weather parameters according to claim 1, characterized in that, The method for decomposing the power load data in step S3 is as follows: A1: Let the original signal be X(t), and the signal after adding noise be: X i (t)=X(t)+∈ i (t) Where, ∈ i (t) is the white noise superimposed for the i-th time; Each signal was decomposed using EMD to obtain a set of Intrinsic Mode Functions (IMFS). The decomposition results are as follows: Among them, IMF ij (t) is the j-th IMF component of the i-th signal, R i (t) is the residual term; By superimposing different types of white noise and repeating the above steps multiple times, the IMFs of all noise versions are averaged to eliminate the influence of noise on the results; the averaged IMFs are: Where M is the number of noise superpositions; A2: After obtaining the IMF, zero-crossing rate detection is used to analyze the frequency characteristics of each IMF. The zero-crossing rate is calculated using the following formula: Where N is the number of samples, and x[n] is the sample value in the IMF; The EEMD method is used to decompose the power load sequence, and the expression is as follows: L(t)=L linear (t)+L nonlinear (t) Where: L linear (t) is the low-frequency linear component, representing the long-term trend; L nonlinear (t) is a high-frequency nonlinear component, representing the rapid fluctuation characteristics of the load.

5. The short-term power load forecasting method combining weather parameters according to claim 4, characterized in that, The modeling and prediction steps for the high-frequency nonlinear components in step S4 are as follows: B1: The high-frequency nonlinear component L nonlinear (t) and the corresponding weather parameters for the time period are used as model inputs; the weather parameters are organized into a matrix according to the time steps: X weather ={X weather,t ,X weather,t +1,...,X weather,T } Each X weather,t It is a vector of weather parameters; The high-frequency nonlinear components are high-frequency components extracted from the power load time series using EEMD: X high ={X high,t ,X high,t+1 ,...,X high,T } Then the weather parameters and high-frequency components are normalized to [0, 1]; B2: The CNN-LSTM model based on multi-scale feature extraction consists of three main modules: CNN, LSTM, and Attention mechanism. Based on the weather parameters input in step B1, CNN extracts the local features of the weather parameters and identifies the correlation between weather and load fluctuations. Based on the high-frequency nonlinear components input in step B1, LSTM captures the time series characteristics of the high-frequency components and models their dynamic changes. Attention mechanisms are used to improve the model's ability to focus on key time steps and important weather parameters; B3: Use historical power load data to construct training and validation sets, use mean squared error as the loss function, and dynamically adjust the learning rate through the Adam optimization algorithm to improve training efficiency and obtain a well-trained CNN-LSTM model.

6. The short-term power load forecasting method combining weather parameters according to claim 5, characterized in that, In step B2, the CNN-LSTM model includes convolutional layers and pooling layers. The convolutional layers extract spatial local patterns of weather parameters, while the pooling layers reduce feature dimensionality while retaining key patterns. The CNN extracts local features F of the weather parameters. weather : F weather ={f1,f2,...,fd} Among them, f i Let i be the i-th weather feature; score(f i )=W weather *f i +b weather Among them, W weather and b weather These are the learning parameters.

7. The short-term power load forecasting method combining weather parameters according to claim 5, characterized in that, In step B2, the CNN-LSTM model uses an LSTM network to model high-frequency load components and capture dynamic changes over time. The LSTM network model structure includes a forget gate, an input gate, and an output gate. The LSTM extracts features H. high This represents the sequence of hidden states that generate time steps, including periodic fluctuations and random characteristics or outlier changes; H high ={h1,h2,…,hT} score(h t )=O t fishy(C) t ) Among them, O t It is the output gate result, C t This is the result of the unit update status.

8. The short-term power load forecasting method combining weather parameters according to claim 5, characterized in that, The operation of the attention mechanism in the CNN-LSTM model in step B2 includes: The importance of using fully connected networks to compute features: Calculate the attention score using contextual information and hidden states: Generate attention-weighted time series features Features after incorporating the attention mechanism are then concatenated: F combined =Concat(F weather,attn ,H high,attn ) Fully connected output: y t =W out *F combined +b out 。 9. The short-term power load forecasting method and system combining weather parameters according to claim 1, characterized in that, In step S5, the low-frequency linear component L is targeted. linear (t) Independent predictions were made using the SARIMA model.

10. The short-term power load forecasting method and system combining weather parameters according to claim 1, characterized in that, The error analysis method in step S6 includes: D1: Calculate the difference between the actual load value and the predicted load value; D2: Evaluation metrics include mean squared error, mean absolute error, absolute percentage error, and the absolute coefficient R, which measures the goodness of fit between the predicted and actual load values. 2 ; D3: Using visualization tools, draw time series graphs of actual load values, predicted load values, and errors, observe the fluctuation trend of errors, and use box plots to show the distribution characteristics of errors and identify the concentrated intervals of outliers.

Images