Prediction method and system based on price trend of electricity market

By employing data collection, preprocessing, feature selection, and hybrid model prediction, this method addresses the limitations of existing electricity market price trend prediction methods, achieving more efficient and stable electricity market price prediction and enhancing the model's interpretability and adaptability.

CN119048183APending Publication Date: 2024-11-29SICHUAN YILAIKE SMART ENERGY CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411073943.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Existing methods for predicting electricity market price trends fail to adequately consider the characteristics that are highly correlated with price trends, resulting in limited model performance and over-reliance on a single model, making it impossible to comprehensively capture market dynamics.

Method used

The method employs data collection, preprocessing, feature selection, and model building, including data cleaning, standardization, autoencoder feature selection, and hybrid model prediction. It combines time series models and machine learning models for voting or weighted averaging to improve the accuracy and stability of predictions.

Benefits of technology

By improving feature selection and model building, the accuracy and stability of predictions are enhanced, the model structure is simplified, the computational resource requirements are reduced, and the interpretability and adaptability of the model are improved. It can better capture the nonlinear relationships and complex patterns of electricity market prices and provide more accurate decision support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004981665200000041
    Figure BDA0004981665200000041
  • Figure BDA0004981665200000111
    Figure BDA0004981665200000111
  • Figure FDA0004981665190000021
    Figure FDA0004981665190000021
Patent Text Reader

Abstract

The invention discloses an electricity market price trend-based prediction method and system. According to the method, the prediction accuracy can be improved through feature selection. By screening the characteristics highly related to the price trend, the interference of noise and irrelevant information can be reduced, so that the prediction accuracy of the model is improved. Secondly, the complexity of the model is reduced through feature selection, redundant or unimportant features are removed, the model structure is simplified, the risk of overfitting is reduced, and the model is more concise and efficient. In addition, the feature selection also saves computing resources, reduces the number of features, reduces the computing resources required by model training and prediction, and can significantly reduce the computing cost especially when processing a large-scale data set. The interpretability of the model is enhanced through feature selection, and through feature selection, which factors have significant influence on the price trend of the electricity market are better understood, so that the generalization ability of the model is enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of power market forecasting, and specifically relates to a forecasting method and system based on power market price trends. Background Art

[0002] The price trend of the electricity market is an important indicator reflecting the supply and demand relationship of the electricity market, and it has important reference value for power companies, investors and policy makers. The following is an introduction to the recent price trend of the electricity market and its forecast: Recently, the price trend of my country's electricity market has generally shown a fluctuating upward trend. This is mainly due to the rise in the prices of energy such as coal and natural gas, which has led to an increase in the cost of electricity production. In addition, with the economic recovery, the demand for electricity has gradually increased, and the changes in the supply and demand relationship have also driven the rise in electricity prices. Regarding the forecast of the price trend of the electricity market, the following views are worth noting: Short-term fluctuations: Affected by domestic and foreign energy markets, macroeconomics and seasonal factors, the price of the electricity market may still fluctuate in the short term. Medium- and long-term upward trend: With the continuous growth of my country's economy, the demand for electricity will gradually increase, and the price of the electricity market is expected to maintain a medium- and long-term upward trend. The price advantage of new energy electricity is prominent: With the continuous advancement of new energy technology and the reduction of costs, the competitive advantage of new energy electricity prices will become more obvious, which is expected to drive the overall price of the electricity market to fall. Policy regulation role: The government will continue to play a policy regulation role to ensure the balance of supply and demand in the electricity market and stabilize electricity prices. In short, the price trend of the electricity market is affected by many factors and shows a certain degree of volatility. In the coming period, the electricity market will continue to face many challenges, but the overall trend is expected to maintain steady growth. All parties should pay close attention to market dynamics and formulate reasonable strategies to cope with changes in electricity market prices.

[0003] However, common forecasting methods fail to fully consider features that are highly correlated with price trends, resulting in limited model performance. At the same time, existing methods may rely too much on a single model and fail to fully capture market dynamics. Summary of the invention

[0004] The purpose of the present invention is to provide a prediction method and system based on electricity market price trends in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a prediction method based on the price trend of the electricity market, the method comprising the following steps:

[0006] S1: Collect data, collect power market price data, obtain it from the official website of the power market, industry reports, and historical transaction data channels, and select an appropriate time span, including at least the data of the past 3-5 years;

[0007] S2: Perform data preprocessing, first perform data cleaning: remove invalid, erroneous and duplicate data, then perform data standardization: convert data of different dimensions into a unified scale for subsequent analysis;

[0008] S3: Perform feature selection, first calculate the correlation coefficient between each feature and the electricity market price, and select the features that are highly correlated with the price trend. Then use the autoencoder to perform feature selection, and further select the features that have a significant impact on price prediction;

[0009] S4: Model construction, using a hybrid model combined with a time series model and a machine learning model, and using the prediction results of multiple models for voting or weighted averaging to improve the stability of the prediction;

[0010] S5: Prediction and evaluation: Use the trained model to predict the test set, obtain the prediction results of the power market price trend, calculate the error between the prediction results and the actual values, such as mean square error (MSE), root mean square error (RMSE), etc., to evaluate the model performance;

[0011] S6: Perform result analysis and optimization, analyze the prediction error: analyze the causes of the prediction error, find the problems in the model, and then end the entire prediction process based on the price trend of the electricity market.

[0012] In a preferred embodiment, in step S1, the data types include electricity market price, power generation, load, new energy power generation, weather conditions, and factors affecting electricity market price due to policy changes.

[0013] In a preferred embodiment, in step S2, the data cleaning method includes: the cloud computing server uses a pre-established electricity market price trend data cleaning algorithm to clean the target electricity market price trend data to obtain a first electricity market price trend data cleaning result; the electricity market price trend data cleaning algorithm is trained according to the first electricity market price trend data cleaning result and the second electricity market price trend data cleaning result to obtain an updated electricity market price trend data cleaning algorithm for the next electricity market price trend data cleaning; wherein the second electricity market price trend data cleaning result is the electricity market price trend data cleaning result obtained by manually cleaning the target electricity market price trend data.

[0014] In a preferred embodiment, in step S2, data standardization converts the data into a form with zero mean and unit variance; the specific steps are as follows:

[0015] S1. Calculate the mean and standard deviation: For each feature, calculate its mean (Mean) and standard deviation (Std);

[0016] S2. Apply transformation formula: For each feature value x in the electricity market price trend dataset, use the following formula for transformation:

[0017]

[0018] Among them, x std is the transformed value;

[0019] S3. Transform the dataset: Apply the above transformation to each feature of the entire electricity market price trend dataset; after Z-score standardization, the mean of each feature is 0 and the standard deviation is 1, and the distribution of the dataset will be similar to the standard normal distribution.

[0020] In a preferred embodiment, in step S3, the autoencoder network design includes:

[0021] Network architecture: Design an autoencoder network, which consists of an encoder and a decoder. The encoder compresses the input data into a low-dimensional representation, and the decoder attempts to reconstruct the original data from the low-dimensional representation.

[0022] Encoder design: The encoder usually consists of several fully connected layers, each of which can be followed by an activation function.

[0023] Decoder Design: The decoder structure is similar to the encoder, but the number of layers and neurons is usually opposite to the encoder.

[0024] In a preferred embodiment, in step S3, the specific steps of performing feature selection include:

[0025] Feature extraction: Encoder output: Use the trained encoder to extract low-dimensional representations of the input data, which are new features. Feature importance assessment: You can determine which new features are the most important by observing the weights in the encoder layer or using other feature importance assessment methods.

[0026] Feature selection: Feature screening: Select the most important feature subset based on the feature importance evaluation results. Correlation analysis: Perform correlation analysis on new features and remove highly correlated features to reduce multicollinearity.

[0027] Model Validation: Feature Validation: Train a predictive model using a selected subset of features and evaluate its performance on a validation set. Performance Comparison: Compare the performance of the model after feature selection using the autoencoder with the performance of the model using the original feature set or other feature selection methods.

[0028] In a preferred embodiment, in step S4, prediction is performed using the ARIMA model and the LSTM model respectively, and the specific process method includes:

[0029] Step 1: Data preparation:

[0030] Data collection: Collect historical electricity market price data.

[0031] Data preprocessing: including data cleaning, filling missing values, standardization, etc.

[0032] Step 2: ARIMA model construction:

[0033] Time Series Analysis: Determine the parameters (p, d, q) of the ARIMA model through ACF and PACF plots.

[0034] Model training: Train the ARIMA model using the training set.

[0035] Residual extraction: Calculate the residuals of the ARIMA model, which are the differences between the actual and predicted values.

[0036] Step 3: Feature Engineering:

[0037] Feature construction: In addition to the original time series data, it also includes residuals, trend and seasonal components extracted by the ARIMA model.

[0038] Feature normalization: Make sure all features have the same scale.

[0039] Step 4: LSTM model construction:

[0040] Network design: Design the LSTM network structure, including the input layer, multiple LSTM layers, a fully connected layer, and an output layer.

[0041] Parameter setting: Determine hyperparameters such as learning rate, batch size, and number of iterations.

[0042] Step 5: Hybrid model training:

[0043] Data input: The standardized feature set is input into the LSTM model.

[0044] Model training: The LSTM model is trained using the training set, while the residuals of the ARIMA model are used as additional input features.

[0045] Callback function: Set callback functions such as Early Stopping to prevent overfitting.

[0046] Step 6: Model integration:

[0047] Forecast combination: Use the ARIMA model and LSTM model to make predictions separately, and then combine the prediction results of the two.

[0048] Integration strategy: A simple average method or a weighted average method can be used, and the weights can be determined based on the performance of the model on the validation set.

[0049] Step 7: Model Validation:

[0050] Performance evaluation: Use the validation set to evaluate the performance of the hybrid model and calculate indicators such as RMSE and MAE.

[0051] Model adjustment: Adjust model parameters or structure based on validation results.

[0052] Step 8: Model testing:

[0053] Final Test: Perform a final test on the hybrid model using the test set.

[0054] Performance report: record test results and generate performance reports.

[0055] In a preferred embodiment, in step S5, the flow of the prediction process includes:

[0056] Step 1: Model preparation:

[0057] Load model: Make sure the trained model is loaded into memory correctly.

[0058] Data preparation: Perform necessary preprocessing on the test set data, such as standardization or normalization, to ensure that it is processed in the same way as the training set.

[0059] Step 2: Prediction execution:

[0060] Input data: Input the preprocessed test set data into the model.

[0061] Prediction: Call the model's prediction method to generate prediction results for future electricity market prices.

[0062] Step 3: Save the results:

[0063] Save prediction results: Save the model's prediction results to a file or database for subsequent analysis and use.

[0064] In a preferred embodiment, in step S5, the specific method flow of the evaluation process includes:

[0065] Step 1: Actual value preparation:

[0066] Get actual values: Collect actual electricity market price data corresponding to the test set.

[0067] Data alignment: Ensure that the forecast results are aligned with the actual values ​​in time series.

[0068] Step 2: Error calculation:

[0069] Calculate the error: Use the mean squared error statistic to calculate the difference between the predicted result and the actual value.

[0070] Step 3: Result analysis:

[0071] Analyze errors: Analyze the distribution and pattern of forecast errors to understand when or under what circumstances the model performs poorly.

[0072] Model improvement: Based on the error analysis results, consider further tuning or retraining the model.

[0073] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0074] 1. In the present invention, feature selection can improve the accuracy of prediction. By screening features that are highly correlated with price trends, the interference of noise and irrelevant information can be reduced, thereby improving the prediction accuracy of the model. Secondly, feature selection reduces the complexity of the model, removes redundant or unimportant features, simplifies the model structure, reduces the risk of overfitting, and makes the model more concise and efficient. In addition, feature selection also saves computing resources, reduces the number of features, and reduces the computing resources required for model training and prediction, especially when processing large-scale data sets, which can significantly reduce computing costs. Feature selection enhances the interpretability of the model. Through feature selection, we can better understand which factors have a significant impact on the price trend of the electricity market, thereby improving the interpretability of the model and facilitating market participants to understand and apply the model prediction results. Finally, feature selection improves the generalization ability of the model. Through feature selection, the model can better capture the key information in the data, improve the prediction ability of unknown data, and thus enhance the generalization ability of the model.

[0075] 2. In the present invention, model construction can improve the stability of prediction. By combining the time series model and the machine learning model, and using the prediction results of multiple models for voting or weighted averaging, the limitations of a single model are reduced, and the stability and reliability of the prediction are improved. Secondly, model construction can capture complex relationships, and the hybrid model can better capture the nonlinear relationships and complex patterns of electricity market prices, and improve the ability to capture market dynamics. In addition, model construction also improves the adaptability of the model. By using different models, the model construction method can better adapt to different types of data and problems, and improves the adaptability and flexibility of the model. Model construction enhances the robustness of the model. Through model integration, the robustness of the model is enhanced, so that it can better cope with uncertainties and external interference in the market. Finally, model construction improves the prediction accuracy of the model. By combining the advantages of different models, the model construction method can improve the prediction accuracy and provide more accurate and reliable decision support for market participants.

[0076] 3. In the present invention, through the two steps of feature selection and model construction, the electricity market price trend prediction method can more effectively capture the key information in the data, improve the accuracy and reliability of the prediction, and provide better decision support for market participants. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0078] Example:

[0079] A prediction method based on power market price trend, the method comprising the following steps:

[0080] S1: Collect data, collect power market price data, obtain it from the official website of the power market, industry reports, and historical transaction data channels, and select an appropriate time span, including at least the data of the past 3-5 years;

[0081] S2: Perform data preprocessing, first perform data cleaning: remove invalid, erroneous and duplicate data, then perform data standardization: convert data of different dimensions into a unified scale for subsequent analysis;

[0082] S3: Perform feature selection, first calculate the correlation coefficient between each feature and the electricity market price, and select the features that are highly correlated with the price trend. Then use the autoencoder to perform feature selection, and further select the features that have a significant impact on price prediction;

[0083] S4: Model construction, using a hybrid model combined with a time series model and a machine learning model, and using the prediction results of multiple models for voting or weighted averaging to improve the stability of the prediction;

[0084] S5: Prediction and evaluation: Use the trained model to predict the test set, obtain the prediction results of the power market price trend, calculate the error between the prediction results and the actual values, such as mean square error (MSE), root mean square error (RMSE), etc., to evaluate the model performance;

[0085] S6: Perform result analysis and optimization, analyze the prediction error: analyze the causes of the prediction error, find the problems in the model, and then end the entire prediction process based on the price trend of the electricity market.

[0086] In step S1, the data types include electricity market price, power generation, load, renewable energy power generation, weather conditions, and factors affecting electricity market price due to policy changes.

[0087] In step S2, the data cleaning method includes: the cloud computing server uses a pre-established electricity market price trend data cleaning algorithm to clean the target electricity market price trend data to obtain a first electricity market price trend data cleaning result; the electricity market price trend data cleaning algorithm is trained according to the first electricity market price trend data cleaning result and the second electricity market price trend data cleaning result to obtain an updated electricity market price trend data cleaning algorithm for the next electricity market price trend data cleaning; wherein the second electricity market price trend data cleaning result is an electricity market price trend data cleaning result obtained by manually cleaning the target electricity market price trend data.

[0088] In step S2, data standardization converts the data into a form with zero mean and unit variance; the specific steps are as follows:

[0089] S1. Calculate the mean and standard deviation: For each feature, calculate its mean (Mean) and standard deviation (Std);

[0090] S2. Apply transformation formula: For each feature value x in the electricity market price trend dataset, use the following formula for transformation:

[0091]

[0092] Among them, x std is the transformed value;

[0093] S3. Transform the dataset: Apply the above transformation to each feature of the entire electricity market price trend dataset; after Z-score standardization, the mean of each feature is 0 and the standard deviation is 1, and the distribution of the dataset will be similar to the standard normal distribution.

[0094] In step S3, the autoencoder network design includes:

[0095] Network architecture: Design an autoencoder network, which consists of an encoder and a decoder. The encoder compresses the input data into a low-dimensional representation, and the decoder attempts to reconstruct the original data from the low-dimensional representation.

[0096] Encoder design: The encoder usually consists of several fully connected layers, each of which can be followed by an activation function.

[0097] Decoder Design: The decoder structure is similar to the encoder, but the number of layers and neurons is usually opposite to the encoder.

[0098] In step S3, the specific steps of feature selection include:

[0099] Feature extraction: Encoder output: Use the trained encoder to extract low-dimensional representations of the input data, which are new features. Feature importance assessment: You can determine which new features are the most important by observing the weights in the encoder layer or using other feature importance assessment methods.

[0100] Feature selection: Feature screening: Select the most important feature subset based on the feature importance evaluation results. Correlation analysis: Perform correlation analysis on new features and remove highly correlated features to reduce multicollinearity.

[0101] Model Validation: Feature Validation: Train a predictive model using a selected subset of features and evaluate its performance on a validation set. Performance Comparison: Compare the performance of the model after feature selection using the autoencoder with the performance of the model using the original feature set or other feature selection methods.

[0102] In step S4, the ARIMA model and the LSTM model are used for prediction respectively. The specific process method includes:

[0103] Step 1: Data preparation:

[0104] Data collection: Collect historical electricity market price data.

[0105] Data preprocessing: including data cleaning, filling missing values, standardization, etc.

[0106] Step 2: ARIMA model construction:

[0107] Time Series Analysis: Determine the parameters (p, d, q) of the ARIMA model through ACF and PACF plots.

[0108] Model training: Train the ARIMA model using the training set.

[0109] Residual extraction: Calculate the residuals of the ARIMA model, which are the differences between the actual and predicted values.

[0110] Step 3: Feature Engineering:

[0111] Feature construction: In addition to the original time series data, it also includes residuals, trend and seasonal components extracted by the ARIMA model.

[0112] Feature normalization: Make sure all features have the same scale.

[0113] Step 4: LSTM model construction:

[0114] Network design: Design the LSTM network structure, including the input layer, multiple LSTM layers, a fully connected layer, and an output layer.

[0115] Parameter setting: Determine hyperparameters such as learning rate, batch size, and number of iterations.

[0116] Step 5: Hybrid model training:

[0117] Data input: The standardized feature set is input into the LSTM model.

[0118] Model training: The LSTM model is trained using the training set, while the residuals of the ARIMA model are used as additional input features.

[0119] Callback function: Set callback functions such as Early Stopping to prevent overfitting.

[0120] Step 6: Model integration:

[0121] Forecast combination: Use the ARIMA model and LSTM model to make predictions separately, and then combine the prediction results of the two.

[0122] Integration strategy: A simple average method or a weighted average method can be used, and the weights can be determined based on the performance of the model on the validation set.

[0123] Step 7: Model Validation:

[0124] Performance evaluation: Use the validation set to evaluate the performance of the hybrid model and calculate indicators such as RMSE and MAE.

[0125] Model adjustment: Adjust model parameters or structure based on validation results.

[0126] Step 8: Model testing:

[0127] Final Test: Perform a final test on the hybrid model using the test set.

[0128] Performance report: record test results and generate performance reports.

[0129] In step S5, the flow of the prediction process includes:

[0130] Step 1: Model preparation:

[0131] Load model: Make sure the trained model is loaded into memory correctly.

[0132] Data preparation: Perform necessary preprocessing on the test set data, such as standardization or normalization, to ensure that it is processed in the same way as the training set.

[0133] Step 2: Prediction execution:

[0134] Input data: Input the preprocessed test set data into the model.

[0135] Prediction: Call the model's prediction method to generate prediction results for future electricity market prices.

[0136] Step 3: Save the results:

[0137] Save prediction results: Save the model's prediction results to a file or database for subsequent analysis and use.

[0138] In step S5, the specific method flow of the evaluation process includes:

[0139] Step 1: Actual value preparation:

[0140] Get actual values: Collect actual electricity market price data corresponding to the test set.

[0141] Data alignment: Ensure that the forecast results are aligned with the actual values ​​in time series.

[0142] Step 2: Error calculation:

[0143] Calculate the error: Use the mean squared error statistic to calculate the difference between the predicted result and the actual value.

[0144] Step 3: Result analysis:

[0145] Analyze errors: Analyze the distribution and pattern of forecast errors to understand when or under what circumstances the model performs poorly.

[0146] Model improvement: Based on the error analysis results, consider further tuning or retraining the model.

[0147] A prediction system based on the price trend of the electricity market, the system runs a prediction method based on the price trend of the electricity market.

[0148] In the present invention, feature selection can improve the accuracy of prediction. By screening features that are highly correlated with price trends, the interference of noise and irrelevant information can be reduced, thereby improving the prediction accuracy of the model. Secondly, feature selection reduces the complexity of the model, removes redundant or unimportant features, simplifies the model structure, reduces the risk of overfitting, and makes the model more concise and efficient. In addition, feature selection also saves computing resources, reduces the number of features, and reduces the computing resources required for model training and prediction, especially when processing large-scale data sets, which can significantly reduce computing costs. Feature selection enhances the interpretability of the model. Through feature selection, we can better understand which factors have a significant impact on the price trend of the electricity market, thereby improving the interpretability of the model and facilitating market participants to understand and apply the model prediction results. Finally, feature selection improves the generalization ability of the model. Through feature selection, the model can better capture the key information in the data, improve the prediction ability of unknown data, and thus enhance the generalization ability of the model.

[0149] In the present invention, model construction can improve the stability of prediction. By combining the time series model and the machine learning model, and using the prediction results of multiple models for voting or weighted averaging, the limitations of a single model are reduced, and the stability and reliability of the prediction are improved. Secondly, model construction can capture complex relationships, and the hybrid model can better capture the nonlinear relationships and complex patterns of electricity market prices, and improve the ability to capture market dynamics. In addition, model construction also improves the adaptability of the model. By using different models, the model construction method can better adapt to different types of data and problems, and improves the adaptability and flexibility of the model. Model construction enhances the robustness of the model. Through model integration, the robustness of the model is enhanced, so that it can better cope with uncertain factors and external interference in the market. Finally, model construction improves the prediction accuracy of the model. By combining the advantages of different models, the model construction method can improve the prediction accuracy and provide more accurate and reliable decision support for market participants.

[0150] In the present invention, through the two steps of feature selection and model construction, the power market price trend prediction method can more effectively capture key information in the data, improve the accuracy and reliability of the prediction, and provide better decision support for market participants.

[0151] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0152] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A prediction method based on electricity market price trends, Features: The method comprises the following steps: S1: Collect data, collect power market price data, obtain it from the official website of the power market, industry reports, and historical transaction data channels, and select an appropriate time span, including at least the data of the past 3-5 years; S2: Perform data preprocessing, first perform data cleaning: remove invalid, erroneous and duplicate data, then perform data standardization: convert data of different dimensions into a unified scale for subsequent analysis; S3: Perform feature selection, first calculate the correlation coefficient between each feature and the electricity market price, and select the features that are highly correlated with the price trend; then use the autoencoder to perform feature selection, and further select the features that have a significant impact on price prediction; S4: Model construction, using a hybrid model combined with a time series model and a machine learning model, and using the prediction results of multiple models for voting or weighted averaging to improve the stability of the prediction; S5: Prediction and evaluation: Use the trained model to predict the test set, obtain the prediction results of the power market price trend, calculate the error between the prediction results and the actual value, and evaluate the model performance; S6: Perform result analysis and optimization, analyze the prediction error: analyze the causes of the prediction error, find the problems in the model, and then end the entire prediction process based on the price trend of the electricity market.

2. A method for predicting power market price trends according to claim 1, Features: In step S1, the data types include electricity market price, power generation, load, new energy power generation, weather conditions, and factors affecting electricity market price due to policy changes.

3. A method for predicting power market price trends according to claim 1, Features: In the step S2, the data cleaning method includes: the cloud computing server uses a pre-established electricity market price trend data cleaning algorithm to clean the target electricity market price trend data to obtain a first electricity market price trend data cleaning result; the electricity market price trend data cleaning algorithm is trained according to the first electricity market price trend data cleaning result and the second electricity market price trend data cleaning result to obtain an updated electricity market price trend data cleaning algorithm for the next electricity market price trend data cleaning; wherein the second electricity market price trend data cleaning result is the electricity market price trend data cleaning result obtained by manually cleaning the target electricity market price trend data.

4. A method for predicting power market price trends according to claim 1, Features: In step S2, data standardization converts the data into a form with zero mean and unit variance; the specific steps are as follows: S1. Calculate the mean and standard deviation: For each feature, calculate its mean (Mean) and standard deviation (Std); S2. Apply transformation formula: For each feature value x in the electricity market price trend dataset, use the following formula for transformation: Among them, x std is the transformed value; S3. Transform the dataset: Apply the above transformation to each feature of the entire electricity market price trend dataset; after Z-score standardization, the mean of each feature is 0 and the standard deviation is 1, and the distribution of the dataset will be similar to the standard normal distribution.

5. A method for predicting power market price trends according to claim 1, Features: In step S3, the autoencoder network design includes: Network architecture: Design an autoencoder network, which consists of an encoder and a decoder; the encoder compresses the input data into a low-dimensional representation, and the decoder tries to reconstruct the original data from the low-dimensional representation; Encoder design: The encoder usually consists of several fully connected layers, each followed by an activation function; Decoder Design: The decoder structure is similar to the encoder, but the number of layers and neurons is usually opposite to that of the encoder.

6. A method for predicting power market price trends according to claim 1, Features: In step S3, the specific steps of performing feature selection include: Feature extraction: Encoder output: Use the trained encoder to extract low-dimensional representations of the input data, which are new features; Feature importance evaluation: Determine which new features are most important by observing the weights in the encoder layer or using other feature importance evaluation methods; Feature selection: Feature screening: Select the most important feature subset based on the feature importance evaluation results; Correlation analysis: Perform correlation analysis on new features and remove highly correlated features to reduce multicollinearity; Model Validation: Feature Validation: Train a predictive model using a selected feature subset and evaluate its performance on a validation set; Performance Comparison: Compare the performance of the model after feature selection using the autoencoder with the performance of the model using the original feature set or other feature selection methods.

7. A method for predicting power market price trends according to claim 1, Features: In step S4, the ARIMA model and the LSTM model are used to perform predictions respectively, and the specific process method includes: Step 1: Data preparation: Data collection: Collect historical electricity market price data; Data preprocessing: including data cleaning, filling missing values, and standardization; Step 2: ARIMA model construction: Time series analysis: Determine the parameters (p, d, q) of the ARIMA model through ACF and PACF plots; Model training: Use the training set to train the ARIMA model; Residual extraction: Calculate the residual of the ARIMA model, that is, the difference between the actual value and the predicted value; Step 3: Feature Engineering: Feature construction: In addition to the original time series data, it also includes residuals, trend and seasonal components extracted by the ARIMA model; Feature normalization: ensuring that all features have the same scale; Step 4: LSTM model construction: Network design: Design the LSTM network structure, including the input layer, multiple LSTM layers, fully connected layers, and output layers; Parameter setting: determine the learning rate, batch size, and number of iterations hyperparameters; Step 5: Hybrid model training: Data input: Input the standardized feature set into the LSTM model; Model training: Use the training set to train the LSTM model, and use the residual of the ARIMA model as an additional input feature; Callback function: Set the Early Stopping callback function to prevent overfitting; Step 6: Model integration: Forecast combination: Use the ARIMA model and LSTM model to make predictions separately, and combine the prediction results of the two; Ensemble strategy: Use simple averaging or weighted averaging, with weights determined based on the model's performance on the validation set; Step 7: Model Validation: Performance evaluation: Use the validation set to evaluate the performance of the hybrid model and calculate the RMSE and MAE indicators; Model adjustment: adjust model parameters or structure according to the verification results; Step 8: Model testing: Final test: Use the test set to perform a final test on the hybrid model; Performance report: record test results and generate performance reports.

8. A method for predicting power market price trends as claimed in claim 1, Features: In step S5, the flow of the prediction process includes: Step 1: Model preparation: Load model: ensure that the trained model is correctly loaded into memory; Data preparation: Perform necessary preprocessing on the test set data, including standardization or normalization, to ensure that it is processed in the same way as the training set; Step 2: Prediction execution: Input data: Input the preprocessed test set data into the model; Prediction: Call the model's prediction method to generate prediction results for future electricity market prices; Step 3: Save the results: Save prediction results: Save the model's prediction results to a file or database for subsequent analysis and use.

9. A method for predicting power market price trends according to claim 1, Features: In step S5, the specific method flow of the evaluation process includes: Step 1: Actual value preparation: Get actual values: Collect actual electricity market price data corresponding to the test set; Data alignment: ensure that the forecast results are aligned with the actual values ​​in time series; Step 2: Error calculation: Calculate the error: Use the mean square error statistic to calculate the difference between the predicted result and the actual value; Step 3: Result analysis: Analyze errors: Analyze the distribution and pattern of forecast errors to understand at which points or in which situations the model performs poorly. Model improvement: Based on the error analysis results, consider further tuning or retraining the model.

10. A prediction system based on electricity market price trends, Features: The system runs the method according to any one of claims 1 to 9.

Citation Information

Cited By

  • Power market network attack identification method, system and device based on gating circulation unit, and storage medium

    CN119449482A