Real-time electricity price prediction method and system based on classification network

Abstract

The invention discloses a real-time electricity price prediction method and system based on a classification network, and the method comprises the steps: monitoring an electricity market in real time, obtaining multi-source data, and enabling the multi-source data to comprise the time sequence characteristics of the electricity price, the state information of external influence factors, and the fluctuation characteristics of the electricity price; constructing a classification network model based on the multi-source data; based on the classification network model, analyzing the currently acquired multi-source data, generating prediction results of a plurality of electricity price intervals, and evaluating the confidence of the result of each interval; and based on the prediction result and the corresponding confidence, comprehensively considering the market change trend, analyzing the future electricity price trend, generating an electricity price prediction report, and providing support for electricity market decision. And it is ensured that prediction of electricity price changes is more comprehensive and accurate.

Description

Technical Field

[0001] The present invention relates to the field of smart grid technology, and in particular to a real-time electricity price prediction method and system based on a classification network. Background Art

[0002] Real-time electricity price forecasting in the current power market is a key link in power system management and market transactions. With the continuous development of the power market, electricity prices are affected by many factors, including electricity demand, power generation costs, climate conditions, energy policies, etc., so electricity price fluctuations are complex and uncertain. Existing electricity price forecasting methods are mainly based on traditional statistical analysis, time series models (such as ARIMA models), machine learning algorithms (such as regression models, support vector machines), etc. Although these methods can capture the time series characteristics and external influencing factors of electricity prices to a certain extent, they still have some limitations:

[0003] Insufficient multi-source data integration capabilities: When processing multi-source data, traditional methods usually only focus on a single or a small number of data sources and cannot effectively integrate the multi-dimensional characteristics of electricity prices (such as time series characteristics, external influencing factors, fluctuation characteristics, etc.). Therefore, such methods cannot fully reflect the complexity of electricity price fluctuations and the impact of multiple factors. Limited prediction accuracy: Although some advanced machine learning methods can improve prediction accuracy, these methods still perform poorly when faced with drastic changes in electricity price fluctuations. In particular, when external factors (such as policy changes and climate change) suddenly appear, it is difficult for existing models to adjust quickly, resulting in large prediction errors. Insufficient model adaptability: Existing models often lack the ability to adapt to market changes, especially when electricity prices fluctuate frequently and data characteristics are complex. Existing methods cannot adjust model parameters in a timely manner, thus affecting the real-time and accuracy of predictions. Imperfect confidence assessment of electricity price ranges: Current electricity price prediction models usually output a single prediction value, but lack an assessment mechanism for the confidence of the prediction results, making it difficult to provide decision makers with clear risk warnings and response plans.

[0004] Therefore, there is an urgent need for a real-time electricity price prediction method and system based on a classification network. Summary of the invention

[0005] The present invention provides a real-time electricity price prediction method and system based on a classification network to solve the above problems existing in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A real-time electricity price prediction method based on a classification network, characterized by comprising:

[0008] S101: Real-time monitoring of the power market to obtain multi-source data, including the time series characteristics of electricity prices, the status information of external influencing factors, and the characteristics of electricity price fluctuations;

[0009] S102: Construct a classification network model based on multi-source data;

[0010] S103: Based on the classification network model, the currently acquired multi-source data is analyzed to generate prediction results of multiple electricity price intervals, and the confidence of the results of each interval is evaluated;

[0011] S104: Based on the prediction results and the corresponding confidence levels, the market trend is comprehensively considered to analyze the future electricity price trend and generate an electricity price forecast report to provide support for electricity market decision-making.

[0012] Wherein, step S101 includes:

[0013] S1011: when monitoring multi-source real-time data of the power market, obtaining time series characteristic information of the electricity price of the power market;

[0014] S1012: Obtaining status information of external factors affecting the power market;

[0015] S1013: Generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library;

[0016] S1014: Based on the electricity price fluctuation feature detection template, the electricity price fluctuation feature is obtained according to the state information.

[0017] Among them, building a classification network model includes:

[0018] Preprocess multi-source data, extract feature information related to real-time electricity price prediction, and use the feature information as multi-source data features;

[0019] According to the characteristics of multi-source data, the correlation between multi-source data and real-time electricity prices is determined. The correlation is used to represent the impact of time series characteristics, external influencing factors and electricity price fluctuation characteristics on real-time electricity price prediction;

[0020] Based on the correlation between multi-source data and real-time electricity prices, an initial classification network model is constructed. The classification network model predicts real-time electricity prices by inputting multi-source data features. The initial classification network model consists of multiple network layers and neural units. The network layer includes a time series layer, a state information processing layer, and a fluctuation feature extraction layer, which are used to process input data related to the time series features of electricity prices, the state information of external influencing factors, and the fluctuation features of electricity prices, respectively.

[0021] Training the initial classification network model, and adjusting the weights in the classification network model using preset training set data, wherein the training set data includes labeled historical electricity prices and corresponding multi-source data features;

[0022] According to the performance of the trained classification network model, the classification network model is optimized using a preset model optimization strategy. The model optimization strategy includes a back propagation algorithm based on error feedback, which is used to adjust the weight relationship between network layers to improve the accuracy of real-time electricity price prediction and obtain an optimized classification network model.

[0023] Based on the optimized classification network model, the real-time input multi-source data is predicted, and the real-time electricity price prediction result is used as the output of the classification network model;

[0024] In the classification network model, the effect of the classification network model is evaluated according to the preset model detection method to judge the performance of the classification network model in the real-time electricity price forecast, and the evaluation results are used as the basis for further model optimization;

[0025] Based on the evaluation results of the classification network model, the classification network model finally used for real-time electricity price prediction is determined. The classification network model achieves high-precision prediction of real-time electricity prices through the input of multi-source data.

[0026] The currently acquired multi-source data is analyzed, including:

[0027] Based on multi-source data, construct an input data set and input the input data set into the classification network model;

[0028] Based on the analysis results of the classification network model, multiple electricity price intervals are set, and the electricity price interval is the predicted value range of electricity prices in a continuous time period;

[0029] Conduct confidence assessment on the forecast results of the electricity price range, which is based on the fitting degree of the classification network model to the input data and the fluctuation state of external influencing factors;

[0030] If the confidence level of a certain electricity price range is lower than a preset confidence level threshold, the prediction result of the electricity price range is re-analyzed or adjusted to generate a revised electricity price range;

[0031] Output the forecast results of multiple electricity price ranges and the corresponding confidence information for electricity price regulation or market analysis;

[0032] Among them, the setting of multiple electricity price intervals is based on the analysis of the time series characteristics of electricity prices by the classification network model, combined with the state information of external influencing factors, to dynamically generate multiple interval ranges;

[0033] Confidence assessment includes backtesting based on historical data to ensure the accuracy and reliability of the forecast results.

[0034] Among them, the electricity price forecast report is generated, including:

[0035] Generate an electricity price forecast report based on the electricity price forecast results, and provide multiple scenario analyses of future electricity price trends based on the confidence level;

[0036] Compare and analyze the data in the forecast report. When the deviation between the electricity price change trend in the forecast report and the historical or pre-stored data is greater than the first preset threshold, issue a first-level reminder and add analysis suggestions to the forecast report.

[0037] When the deviation of the electricity price forecast result exceeds the second preset threshold, a secondary alarm is triggered, and a market risk warning and related response plan are attached to the forecast report;

[0038] When the deviation of the electricity price forecast result exceeds the third preset threshold, a third-level alarm is triggered, and the report recommends taking emergency regulatory measures to avoid market volatility risks.

[0039] Among them, obtaining the status information of external factors affecting the power market includes:

[0040] Obtain historical electricity price records in the electricity market;

[0041] Based on historical electricity price records, the electricity market is divided into key monitoring market areas and non-key monitoring market areas;

[0042] Adaptive allocation of detection resources for detecting electricity price fluctuation characteristics in key monitored market areas and non-key monitored market areas.

[0043] Among them, based on the time series feature information and the preset electricity price fluctuation feature template generation library, an electricity price fluctuation feature detection template is generated, including:

[0044] The number of information types used to parse temporal feature information;

[0045] When the number of information types is unique, the electricity price fluctuation detection rule corresponding to the time series characteristic information is determined from the electricity price fluctuation characteristic template generation library;

[0046] Based on the electricity price fluctuation detection rules, a template for detecting electricity price fluctuation characteristics is generated;

[0047] When the number of information types is not unique, the state information feature set of the time series feature information is parsed based on a preset feature parsing template;

[0048] Based on the state information feature set, construct a first feature description vector of the time series feature information;

[0049] Extracting multiple sets of one-to-one corresponding second feature description vectors and electricity price fluctuation detection rule sets from an electricity price fluctuation feature template generation library;

[0050] Calculating the vector similarity between the first feature description vector and any second feature description vector;

[0051] An electricity price fluctuation detection rule set corresponding to the second feature description vector of the maximum vector similarity is used to generate an electricity price fluctuation feature detection template.

[0052] Among them, multiple electricity price ranges are set, including:

[0053] Obtaining a predicted electricity price value for the electricity price range to be set, where the predicted electricity price value is the electricity value predicted based on the electricity price prediction model within a continuous time period;

[0054] Acquire a reference electricity price data set, where the reference electricity price data set is a reference data set containing electricity price forecast values ​​in a preset electricity price data set;

[0055] The division of electricity price intervals is determined based on the time attribute value of the electricity price forecast value recorded in each reference data in the reference electricity price data set. The time attribute value of the electricity price forecast value recorded in the reference data is used to describe the time distribution of the electricity price forecast value.

[0056] Among them, a real-time electricity price prediction system based on a classification network includes:

[0057] A multi-source data acquisition unit is used to monitor the power market in real time and acquire multi-source data, including time series characteristics of electricity prices, status information of external influencing factors, and characteristics of electricity price fluctuations;

[0058] A classification network model unit, used to construct a classification network model based on multi-source data;

[0059] A prediction result acquisition unit is used to analyze the currently acquired multi-source data based on a classification network model, generate prediction results for multiple electricity price intervals, and evaluate the confidence level of the results for each interval;

[0060] Obtain the electricity price forecast report unit, based on the forecast results and the corresponding confidence level, comprehensively consider the market trend, analyze the future electricity price trend, generate an electricity price forecast report, and provide support for electricity market decision-making.

[0061] The obtaining of multi-source data units includes:

[0062] The first multi-source data acquisition module is used to acquire the time series characteristic information of the electricity price in the power market when monitoring the multi-source real-time data in the power market;

[0063] The second multi-source data acquisition module is used to obtain status information of external influencing factors of the power market;

[0064] The third multi-source data acquisition module is used to generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library;

[0065] The fourth multi-source data acquisition module is used to acquire electricity price fluctuation characteristics based on the electricity price fluctuation characteristic detection template and according to the state information.

[0066] Compared with the prior art, the present invention has the following advantages:

[0067] A real-time electricity price forecasting method based on a classification network includes: real-time monitoring of the power market, obtaining multi-source data, the multi-source data including the time series characteristics of electricity prices, the state information of external influencing factors and the characteristics of electricity price fluctuations; building a classification network model based on the multi-source data; analyzing the currently obtained multi-source data based on the classification network model, generating forecast results for multiple electricity price intervals, and evaluating the confidence of each interval result; based on the forecast results and the corresponding confidence, comprehensively considering the market change trend, analyzing the future electricity price trend, generating an electricity price forecast report, and providing support for power market decision-making. Ensure that the forecast of electricity price changes is more comprehensive and accurate.

[0068] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the present invention.

[0069] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0071] Figure 1 This is a flow chart of a real-time electricity price prediction method based on a classification network in an embodiment of the present invention;

[0072] Figure 2 A flowchart for obtaining multi-source data in an embodiment of the present invention;

[0073] Figure 3 The structure diagram of a real-time electricity price prediction system based on a classification network in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0075] The embodiment of the present invention provides a real-time electricity price prediction method based on a classification network, comprising:

[0076] S101: Real-time monitoring of the power market to obtain multi-source data, including the time series characteristics of electricity prices, the status information of external influencing factors, and the characteristics of electricity price fluctuations;

[0077] S102: Construct a classification network model based on multi-source data;

[0078] S103: Based on the classification network model, the currently acquired multi-source data is analyzed to generate prediction results of multiple electricity price intervals, and the confidence of the results of each interval is evaluated;

[0079] S104: Based on the prediction results and the corresponding confidence levels, the market trend is comprehensively considered to analyze the future electricity price trend and generate an electricity price forecast report to provide support for electricity market decision-making.

[0080] The working principle of the above technical solution is as follows: the system will collect data from the power market in real time from multiple sources, including the time series characteristics of electricity prices (such as hourly, daily, weekly, and monthly electricity price change trends), the status information of external influencing factors (such as climate, supply and demand, policies and regulations, etc.), and electricity price fluctuation characteristics (such as sudden price fluctuations, short-term market disturbances, etc.). These data are continuously acquired through smart sensors, market interfaces, and other means. Suppose that a power company monitors the power consumption in multiple regions in real time every day, and combines weather forecast data (such as high temperature weather may cause a surge in electricity consumption) and policy changes (such as new electricity price subsidy policies), and updates these data in hours. These data are archived and timestamped to form a set of time series features containing multiple data types.

[0081] The system uses deep learning technology to build a classification network model based on the collected multi-source data. The classification network model establishes the association between various features and electricity price fluctuations by learning from historical electricity price data, and classifies the ranges that future electricity prices may fall into based on these features. For example, using historical data, the system recognizes that electricity prices usually show a sharp upward trend when hot weather, electricity demand increases, and the government implements new electricity price intervention policies. The model uses these factors as key features to identify price ranges in similar situations in future forecasts.

[0082] The system uses a classification network model and combines the multi-source data currently acquired in real time to predict electricity prices. The system outputs multiple possible electricity price intervals and generates confidence levels for each interval. This confidence level is based on historical data and the accuracy of the model, and can help evaluate the reliability of the forecast results. For example, currently, the data received by the system indicates that the temperature will rise significantly in the next week, and there is a high possibility of an increase in electricity demand. Combined with the supply status, the system predicts that the electricity price will be between 0.8 yuan / kWh and 1.2 yuan / kWh, of which the 0.9 yuan / kWh interval has an 80% confidence level.

[0083] Based on S103, the system further conducts a comprehensive analysis of the changing trends of the electricity market, taking into account short-term policy changes, seasonal factors, macroeconomic environment, etc., revises or strengthens the forecast results, and finally generates a detailed electricity price forecast report for decision makers' reference.

[0084] The multiple electricity price ranges and corresponding confidence levels given in S103 provide basic data for the market trend analysis in S104. In S104, the system not only relies on the existing forecast results, but also comprehensively considers possible changes in the market, such as sudden policy or demand changes. Specifically, it includes:

[0085] The system first selects the prediction interval with a higher confidence level as a preliminary reference. Secondly, the system analyzes the market trend. For example, if the electricity demand in a certain region increases sharply, the government will issue a temporary regulation policy. Based on this trend, the system will re-evaluate the prediction confidence level in S103 and even adjust the prediction result.

[0086] Assume that the system predicts that the future electricity price may fluctuate around 1.0 yuan / kWh in S103 and gives different confidence levels. However, the upcoming power price limit policy in the market will affect the market supply. The system analyzes in S104 that the policy will greatly suppress the increase in electricity prices, so it adjusts the forecast and finally generates a report suggesting that the decision maker will stabilize the electricity price at around 0.95 yuan / kWh.

[0087] The beneficial effects of the above technical solution are: the system can monitor and integrate in real time from multiple data sources, including historical electricity price data, climate change, policy factors, etc., to ensure that the prediction of electricity price changes is more comprehensive and accurate. Through the integration of multi-source data, the system can quickly respond to market changes and provide real-time forecasting support. The classification network model based on machine learning can learn patterns in historical data and automatically classify electricity price ranges under different scenarios. The self-learning and optimization capabilities of the model make the forecast results gradually accurate, and at the same time provide confidence evaluation for each forecast result to help decision makers better understand the reliability of the forecast results. The system not only provides forecast results, but also generates corresponding confidence based on each forecast interval. The generation of confidence allows decision makers to more clearly understand the possibilities of different electricity price ranges. At the same time, the system can comprehensively consider market trends and revise the forecast results to ensure that the final forecast report is more in line with the actual situation. The electricity price forecast report finally generated by the system is not only based on real-time data, but also takes into account long-term market change trends, which can help power market decision makers make more forward-looking and scientific decisions when dealing with market fluctuations.

[0088] In another embodiment, step S101 includes:

[0089] S1011: when monitoring multi-source real-time data of the power market, obtaining time series characteristic information of the electricity price of the power market;

[0090] S1012: Obtaining status information of external factors affecting the power market;

[0091] S1013: Generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library;

[0092] S1014: Based on the electricity price fluctuation feature detection template, the electricity price fluctuation feature is obtained according to the state information.

[0093] The working principle of the above technical solution is: when monitoring the power market in real time, the system will first collect the time series characteristic information of electricity prices. Time series characteristics refer to the regularity of electricity price changes over time, such as electricity price fluctuations in time periods such as hours, days, weeks, and months. This feature can help the system understand the short-term and long-term trends of electricity prices. For example, the electricity price in a certain power market has fixed peaks and valleys every day. For example, during the day, the demand for electricity is high, and the electricity price is usually higher. At night, the demand decreases and the electricity price is correspondingly lower. These price change patterns that change over time are the time series characteristics of electricity prices.

[0094] Electricity prices are not only affected by time changes, but also by external factors. External factors include climate conditions (such as extreme heat or cold weather), policy changes (such as government intervention in electricity prices), changes in electricity supply and demand, etc. These status information helps the system analyze the reasons for electricity price fluctuations more comprehensively. Suppose the temperature in a region suddenly rises to 40°C, resulting in a surge in air conditioning use, a rapid increase in electricity demand, and a corresponding increase in electricity prices. At this time, the temperature information, as an external influencing factor, is a key data source for electricity price fluctuations.

[0095] The system will generate an electricity price fluctuation feature detection template based on the collected time series feature information and the pre-set electricity price fluctuation feature template library. This template is used to identify specific electricity price fluctuation patterns, such as sharp increases or sudden decreases. The template library is a summary of electricity price fluctuation patterns based on historical data. Suppose the system finds that after 5 consecutive days of high temperature weather in a certain market, electricity prices tend to rise rapidly on the 6th day. Through historical data, the system has recorded this feature in the template library in advance. Now, if the weather forecast shows that there will be similar high temperature weather again, the system will generate a corresponding electricity price fluctuation detection template to cope with possible fluctuations in the future.

[0096] The system uses the generated electricity price fluctuation feature detection template and the current external influencing factor status to analyze the characteristics of electricity price fluctuations. The electricity price fluctuation characteristics reflect abnormal changes in electricity prices, such as large increases or decreases. These characteristics usually correspond to certain specific market conditions. Assume that, at present, meteorological data shows that high temperatures will continue and the problem of insufficient power supply is worsening. The system detects the characteristics of a possible large increase in electricity prices through the template, and matches it with the current external influencing factors to confirm that a sharp electricity price fluctuation is about to occur. This fluctuation feature helps the system and market decision makers take measures in advance, such as adjusting power allocation to balance market supply and demand.

[0097] Electricity price time series characteristic information: This is the cyclical change pattern of electricity prices under normal market conditions, reflecting the natural fluctuations in electricity prices, such as peak hours on weekdays and low hours on holidays.

[0098] Electricity price fluctuation characteristics: refers to abnormal fluctuations caused by external conditions, such as extreme weather and policy interventions that lead to dramatic changes in electricity prices. Such fluctuations are often not cyclical, but rather sudden and unpredictable.

[0099] The beneficial effects of the above technical solution are as follows: by combining time series characteristics and external factors, the system can more accurately identify the changing trend of electricity prices, especially when responding to emergencies or external interference, the system can detect abnormal fluctuations early through the electricity price fluctuation feature detection template, and help the power market prepare for response. The system combines the time law of electricity prices with external influencing factors, not only focusing on the natural fluctuations of electricity prices, but also considering the impact of emergencies on electricity prices. This makes electricity price forecasts more comprehensive and can provide more targeted support for power market decision makers. Through the detection of electricity price fluctuation characteristics, the system can provide early warnings to help market participants quickly respond to potential abnormal fluctuations in electricity prices and avoid losses caused by information lags. For example, by associating high temperature weather with peak electricity consumption, power resources can be dispatched in advance to balance supply and demand and stabilize market prices. The system can dynamically adjust the electricity price fluctuation feature detection template based on actual market data, and continuously enrich the template library to adapt to more complex market environments in the future. This self-learning and optimization capability further improves the accuracy and timeliness of the forecast results.

[0100] In another embodiment, constructing a classification network model includes:

[0101] Preprocess multi-source data, extract feature information related to real-time electricity price prediction, and use the feature information as multi-source data features;

[0102] According to the characteristics of multi-source data, the correlation between multi-source data and real-time electricity prices is determined. The correlation is used to represent the impact of time series characteristics, external influencing factors and electricity price fluctuation characteristics on real-time electricity price prediction;

[0103] Based on the correlation between multi-source data and real-time electricity prices, an initial classification network model is constructed. The classification network model predicts real-time electricity prices by inputting multi-source data features. The initial classification network model consists of multiple network layers and neural units. The network layer includes a time series layer, a state information processing layer, and a fluctuation feature extraction layer, which are used to process input data related to the time series features of electricity prices, the state information of external influencing factors, and the fluctuation features of electricity prices, respectively.

[0104] Training the initial classification network model, and adjusting the weights in the classification network model using preset training set data, wherein the training set data includes labeled historical electricity prices and corresponding multi-source data features;

[0105] According to the performance of the trained classification network model, the classification network model is optimized using a preset model optimization strategy. The model optimization strategy includes a back propagation algorithm based on error feedback, which is used to adjust the weight relationship between network layers to improve the accuracy of real-time electricity price prediction and obtain an optimized classification network model.

[0106] Based on the optimized classification network model, the real-time input multi-source data is predicted, and the real-time electricity price prediction result is used as the output of the classification network model;

[0107] In the classification network model, the effect of the classification network model is evaluated according to the preset model detection method to judge the performance of the classification network model in the real-time electricity price forecast, and the evaluation results are used as the basis for further model optimization;

[0108] Based on the evaluation results of the classification network model, the classification network model finally used for real-time electricity price prediction is determined. The classification network model achieves high-precision prediction of real-time electricity prices through the input of multi-source data.

[0109] The working principle of the above technical solution is as follows: The key steps of data preprocessing are as follows:

[0110] Missing value handling: for example, using mean filling, interpolation, or deleting rows with missing data.

[0111] Data normalization / standardization: The magnitude of data such as electricity prices, loads, and weather is different. In order to make the model converge faster, standardization is required.

[0112] Time window division: Expand the historical electricity price and load data according to the time series, for example, set a 1-hour or 24-hour time window for prediction.

[0113] Feature engineering: Extract features that may affect electricity prices from raw data, such as weather factors: temperature, humidity, rainfall, etc.

[0114] Time series characteristics: time period, working days / holidays, etc.

[0115] Fluctuation characteristics: The volatility of electricity prices over the past few hours or days.

[0116] The final feature matrix serves as the input of the model.

[0117] Based on data analysis, correlation analysis (such as Pearson correlation coefficient) or feature importance analysis (such as XGBoost's feature importance score) is used to find out which external factors (such as temperature, load) are highly correlated with electricity prices. This part of the steps helps the model design a network structure that can capture time series characteristics, state characteristics, and fluctuation characteristics.

[0118] Classification network models usually consist of multiple network layers, including:

[0119] Time series layer: Process time series data, often using LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit) to capture the temporal changes in electricity prices.

[0120] State information processing layer: External states (such as weather and demand) are processed through a fully connected layer (Dense Layer).

[0121] Fluctuation feature extraction layer: The fluctuation characteristics of historical electricity prices are extracted through the convolutional layer (CNN).

[0122] These network layers are combined to form the structure of the model, enabling it to capture complex relationships.

[0123] The training set data consists of labeled historical electricity prices and corresponding multi-source features. During the training process, the cross entropy loss function or mean square error (MSE) is used as the loss function to measure the error between the model output and the actual electricity price.

[0124] The backpropagation algorithm is used to adjust the weights of the network layer through the gradient information of the error. The specific optimization strategies include:

[0125] Learning rate adjustment: Use an adaptive optimization algorithm (such as Adam or RMSProp) to dynamically adjust the learning rate based on gradient changes.

[0126] Hyperparameter optimization: Use grid search or Bayesian optimization to tune the key parameters of the model (such as learning rate, batch size, number of network layers, etc.). Regularization strategy: such as L2 regularization or Dropout, to avoid overfitting of the model and improve generalization ability.

[0127] After the training is completed, the multi-source data input in real time is predicted through the model, and the electricity price result is output. The latest data is continuously input through the sliding window method to achieve dynamic prediction.

[0128] Model evaluation is a critical step in selecting the final model. The following evaluation methods can be used:

[0129] Mean Squared Error (MSE): Measures the error between the predicted value and the true value.

[0130] Mean absolute percentage error (MAPE): evaluates the model’s ability to capture electricity price fluctuations.

[0131] R 2 (Coefficient of determination): Measures the ability of the model to explain the variance in the data.

[0132] Cross-validation: K-fold cross-validation is used to ensure the generalization ability of the model to different data sets.

[0133] According to the above evaluation results, the model structure and parameters are further adjusted, and finally the optimal model is selected.

[0134] Select the best performing model based on evaluation metrics (such as MSE, MAPE) to ensure its stable prediction accuracy in the test set and real-time data.

[0135] Monitor the performance of your model in real-world applications and regularly update or retrain the model based on new data.

[0136] The beneficial effects of the above technical solution are as follows: through the fusion of multi-source data and the introduction of deep neural networks, the model can capture the complex fluctuation trend of electricity prices and improve the accuracy of real-time electricity price forecasts. The model can handle different types of external data (such as weather, load, etc.) and improve the generalization ability of the model in different time periods through optimization strategies. Regularization strategies and cross-validation and other optimization methods are used to ensure the stable performance of the model in diversified data and avoid overfitting. The model achieves rapid prediction by dynamically inputting real-time data, providing timely information support for electricity market participants (such as power companies and users). The model structure is flexible in design and can be expanded according to emerging external data (such as policy changes, market events) to ensure that the model adapts to market changes. High-precision electricity price forecasts help market participants optimize procurement and production strategies and reduce economic losses caused by electricity price fluctuations.

[0137] In another embodiment, analyzing the currently acquired multi-source data includes:

[0138] Based on multi-source data, construct an input data set and input the input data set into the classification network model;

[0139] Based on the analysis results of the classification network model, multiple electricity price intervals are set, and the electricity price interval is the predicted value range of electricity prices in a continuous time period;

[0140] Conduct confidence assessment on the forecast results of the electricity price range, which is based on the fitting degree of the classification network model to the input data and the fluctuation state of external influencing factors;

[0141] If the confidence level of a certain electricity price range is lower than a preset confidence level threshold, the prediction result of the electricity price range is re-analyzed or adjusted to generate a revised electricity price range;

[0142] Output the forecast results of multiple electricity price ranges and the corresponding confidence information for electricity price regulation or market analysis;

[0143] Among them, the setting of multiple electricity price intervals is based on the analysis of the time series characteristics of electricity prices by the classification network model, combined with the state information of external influencing factors, to dynamically generate multiple interval ranges;

[0144] Confidence assessment includes backtesting based on historical data to ensure the accuracy and reliability of the forecast results.

[0145] The working principle of the above technical solution is to construct an input data set using multi-source data (such as historical electricity price data, weather data, supply and demand conditions, etc.). After processing and standardization, these data are input into the classification network model. The classification network model is a deep learning model, usually including a convolutional neural network (CNN) or a recurrent neural network (RNN), which is used to process time series data. Through the training process of the model, the network can learn the time series characteristics of electricity prices. Suppose we have collected hourly electricity price data for the past five years, combined with external influencing factors such as temperature, electricity demand and economic growth rate in the same period. These data are integrated into a multi-dimensional input data set. The model uses this data for training and extracts the changing pattern of electricity prices.

[0146] Based on the analysis results of the classification network model, the system predicts future electricity prices and generates multiple electricity price ranges. The electricity price range represents the range that the predicted value of the electricity price may fall within a certain continuous time period. For example, it can be set as "low electricity price range", "medium electricity price range" and "high electricity price range". Assuming that the forecast result is that the electricity price range for the next 24 hours is 0.2 yuan / kWh to 0.6 yuan / kWh, the model will divide the electricity price into three ranges based on these results, such as 0.2 yuan to 0.3 yuan for low electricity price, 0.3 yuan to 0.5 yuan for medium electricity price, and 0.5 yuan to 0.6 yuan for high electricity price.

[0147] Confidence is a measure of the credibility of the prediction results, which depends on the degree of fit of the classification network model to the input data and the fluctuation state of external influencing factors. Confidence can usually be calculated through the output probability distribution of the model, which indicates the degree of certainty of the model for a certain prediction result. At the same time, the magnitude of fluctuations in external factors (such as weather and market demand) will also affect the setting of confidence. When the confidence is high, it means that the model is more confident in the current prediction; otherwise, it is more uncertain. Suppose the model predicts that the electricity price will be 0.3 yuan / kWh in a certain period of time, but the weather data shows that the weather will fluctuate greatly in the future and the demand may change greatly, so the confidence may be lower than a certain threshold.

[0148] If the confidence level of a certain electricity price range is lower than the preset threshold, the system will re-analyze or adjust the forecast results for that range. This may involve reintroducing more data or optimizing the parameters of the model to generate a revised electricity price range. If the confidence level of a certain range is low, the system may recalculate the impact of key data such as weather and demand on electricity prices and readjust the electricity price range. For example, the original electricity price forecast range of 0.3 to 0.4 yuan / kWh is changed to 0.25 to 0.35 yuan / kWh.

[0149] Finally, the system will output the prediction results of each electricity price range and its corresponding confidence information. This information can be used for electricity price regulation or provide decision-making basis for market participants.

[0150] Confidence combines the model fit and the algorithm setting of external factors: The confidence setting can usually be comprehensively considered in the following ways:

[0151] C=f(P,V)

[0152] Among them: C is the confidence level; P is the fit of the model to the input data, which can be calculated through the loss function of the model (such as cross entropy loss); V is the volatility of external influencing factors (such as standard deviation or variance of weather, market demand, etc.).

[0153] This formula indicates that the confidence level C is a function of the model's fit to the data P and the external factor volatility V. If the fit is high and the external factor volatility is low, the confidence level is high; otherwise, the confidence level is low.

[0154] The beneficial effects of the above technical solution are as follows: by inputting multi-source data into the classification network model, the system can learn the complex time series characteristics of electricity prices and generate multiple electricity price ranges in combination with external influencing factors. Compared with traditional linear regression or simple time series models, the classification network model can better capture the nonlinear characteristics of electricity price changes, thereby improving the accuracy of prediction. By incorporating the fluctuation state of external influencing factors into the confidence assessment, the system can automatically adjust the prediction results when the external environment changes drastically. This flexibility enables the system to better cope with situations such as extreme weather and sudden market demand, and avoid static predictions that rely too much on historical data. Confidence information provides users with an intuitive sense of the credibility of the prediction results. Users can make more flexible decisions based on confidence information, such as adopting a more conservative market strategy when the confidence is low, or actively participating in the market when the confidence is high. By outputting multiple electricity price ranges and their confidence information, power companies or market regulators can make more refined electricity price regulation based on the prediction results. This helps to balance market supply and demand and avoid the adverse effects of drastic fluctuations in electricity prices on users and the market.

[0155] In another embodiment, generating an electricity price forecast report includes:

[0156] Generate an electricity price forecast report based on the electricity price forecast results, and provide multiple scenario analyses of future electricity price trends based on the confidence level;

[0157] Compare and analyze the data in the forecast report. When the deviation between the electricity price change trend in the forecast report and the historical or pre-stored data is greater than the first preset threshold, issue a first-level reminder and add analysis suggestions to the forecast report.

[0158] When the deviation of the electricity price forecast result exceeds the second preset threshold, a secondary alarm is triggered, and a market risk warning and related response plan are attached to the forecast report;

[0159] When the deviation of the electricity price forecast result exceeds the third preset threshold, a third-level alarm is triggered, and emergency regulatory measures are recommended in the report to avoid market volatility risks;

[0160] Among them, the first-level warning indicates that the predicted electricity price fluctuations are within the controllable range and will not have a significant impact on the short-term operation of the electricity market. The report provides suggestions to optimize market response;

[0161] Level 2 alarm means that the predicted price fluctuations may cause certain risks to the operation of the power market and temporary regulatory measures are required;

[0162] A Level 3 alert means that the predicted electricity price fluctuations may have a significant impact on the security of the electricity market, and emergency response measures are required to avoid instability in the electricity market.

[0163] The working principle of the above technical solution is: scenario analysis generates multiple scenario forecasts based on different assumptions (such as climate change, market fluctuations, policy changes). For example:

[0164] Forecast of electricity price trends under normal circumstances;

[0165] The potential for dramatic fluctuations in electricity prices during extreme weather conditions;

[0166] The impact of a sharp increase in electricity demand (such as holidays) on electricity prices.

[0167] The model uses the confidence level of the predicted results as the basis for judging future trends. High confidence corresponds to a more stable trend, while low confidence requires special attention to possible risks.

[0168] Compare the predicted electricity price with historical data or stored reference data.

[0169] If the electricity price deviation exceeds the first preset threshold, a first-level reminder is triggered;

[0170] When the second preset threshold is exceeded, a second level alarm is triggered;

[0171] When the third preset threshold is exceeded, a level 3 alarm is triggered.

[0172] The first preset threshold is usually set based on the short-term fluctuation range of the electricity market (such as 5%-10%). Fluctuations within this range will not have a significant impact on the market.

[0173] The second preset threshold is set at 10%-20%, indicating that temporary regulatory measures are needed to cope with larger market fluctuations.

[0174] The third preset threshold: set at more than 20%, which means that electricity price fluctuations may pose a serious threat to market stability and emergency regulatory measures need to be taken.

[0175] By regularly backtesting the model’s forecast accuracy, the threshold range can be adjusted dynamically. For example, during high-demand seasons or periods of market volatility, the threshold can be automatically relaxed to reduce the frequency of unnecessary alerts.

[0176] Level 1 reminder: Provide suggestions to operators, such as adjusting peak and valley electricity price policies, optimizing dispatch plans, etc. Example: If the predicted electricity price is 8% higher than the historical average, provide market optimization suggestions, such as encouraging users to use electricity during off-peak hours.

[0177] Level 2 alarm: Market operators are required to take temporary regulatory measures, such as increasing backup power supplies and restricting some transactions. For example, if the electricity price forecast deviates from the reference value by 15%, it may pose a risk to the market supply and demand balance, and it is recommended to appropriately regulate the electricity load.

[0178] Level 3 alarm: When the electricity price fluctuates by more than 20%, it may lead to market instability or power shortage, and emergency response measures need to be taken, such as starting emergency backup power supply and intervening in market transactions. Example: If the electricity price is predicted to increase by 30%, the emergency plan needs to be activated immediately to avoid market collapse or large-scale power outages for users.

[0179] The beneficial effects of the above technical solution are: through the combination of multiple models and scenario analysis, the prediction system can more accurately grasp the future trend of electricity prices and quickly capture fluctuation signals in the market. The prediction results and analysis reports provide scientific decision-making basis for market participants (such as power generation companies and power sales companies), helping them to arrange power production and scheduling plans more reasonably. Through reminder and alarm mechanisms, potential market risks can be discovered in time, and preventive measures can be taken in advance to reduce the possibility of market out of control. The threshold adaptive adjustment mechanism avoids frequent false alarms and effectively balances the sensitivity of early warning and market stability. Different levels of reminders and alarms correspond to corresponding market response strategies to ensure that participants can respond in time and reduce losses when the market fluctuates. Emergency regulatory measures can effectively prevent market imbalances and maintain the security and continuity of power supply when major anomalies occur in the market.

[0180] In another embodiment, obtaining status information of external influencing factors of the power market includes:

[0181] Obtain historical electricity price records in the electricity market;

[0182] Based on historical electricity price records, the electricity market is divided into key monitoring market areas and non-key monitoring market areas;

[0183] Adaptive allocation of detection resources for detecting electricity price fluctuation characteristics in key monitoring market areas and non-key monitoring market areas;

[0184] Based on historical electricity price records, the electricity market is divided into key monitoring market areas, including:

[0185] Extract the abnormal electricity price fluctuation records generated in the market area in history from the historical electricity price records;

[0186] Based on the abnormal electricity price fluctuation records, determine the abnormal frequency of electricity price fluctuations in the market area;

[0187] When the abnormal frequency of electricity price fluctuation is greater than or equal to the preset abnormal frequency threshold of electricity price fluctuation, the corresponding market area will be taken as a key monitoring market area;

[0188] and / or,

[0189] Obtain forecast basis for external factors influencing the supply and demand relationship in the market area, weather changes, fuel prices, policies, and holidays;

[0190] Input the prediction basis of external influencing factors into the preset electricity price fluctuation upper limit prediction model to determine the electricity price fluctuation upper limit;

[0191] Extracting the first continuous electricity price change within a recent preset time range in the electricity market area from the historical electricity price records;

[0192] Based on the first continuous electricity price change, determine whether the market area has reached the upper limit of electricity price fluctuation;

[0193] When the answer is yes, the corresponding market area will be taken as the key monitored market area.

[0194] The working principle of the above technical solution is: each area of ​​the power market has detailed historical electricity price records, recording the fluctuations of electricity prices in different time periods. These records are usually provided by power market operators or government agencies, covering relevant information such as power supply demand, weather, fuel costs, policy changes, etc. Assume that the daily fluctuation records of electricity prices in area A for the past five years have been saved. With this data, we can analyze the fluctuation pattern of electricity prices in area A.

[0195] Based on the historical electricity price fluctuations, the electricity market is divided into two categories: key monitoring areas and non-key monitoring areas. The specific division standard is based on whether the abnormal frequency of electricity price fluctuations is higher than the preset threshold. The abnormal frequency of electricity price fluctuations refers to the frequency of abnormal fluctuations in electricity prices relative to normal levels in the market area. The threshold is set by experts in the electricity market based on actual conditions. If area A has had a large electricity price fluctuation every month in the past year, while the electricity price fluctuations in area B are relatively stable, area A can be divided into a key monitoring area, while area B is a non-key monitoring area.

[0196] Extract abnormal fluctuation records of a certain market area from historical electricity price records to determine the frequency of abnormal electricity price fluctuations. Abnormal fluctuations usually refer to a sudden and sharp rise or fall in electricity prices. Assume that in the historical electricity price records of area A, there is a sharp increase in electricity prices every three months (for example, an increase of more than 10%). This increase will be recorded as abnormal electricity price fluctuations.

[0197] External factors such as supply and demand, weather changes, fuel prices, and policy adjustments may affect electricity price fluctuations. By predicting these factors and inputting the predicted data into the preset electricity price fluctuation upper limit prediction model, the upper limit of future electricity price fluctuations in the market area can be estimated. In the summer, electricity prices in Region B usually rise due to high temperatures and increased use of air conditioners. Therefore, weather changes are a key external factor in electricity price fluctuations in Region B. Through the weather prediction model, the upper limit of electricity price fluctuations in Region B can be estimated in advance.

[0198] The recent electricity price changes are extracted from the historical electricity price records. If the electricity price in a certain market area changes continuously and reaches or exceeds the predicted fluctuation limit, the area will be regarded as a key monitoring market area. In area A, if the electricity price continues to rise in the past week and reaches the predicted fluctuation limit, the area will be classified as a key monitoring area to prevent greater electricity price fluctuations in the future.

[0199] For key monitoring market areas, the system will automatically adjust the allocation of detection resources, such as increasing the frequency of data monitoring, raising the risk warning level, etc., in order to better capture abnormal fluctuations in electricity prices. For non-key monitoring areas, the allocation of detection resources is relatively reduced, thereby saving resources. Since Area A is classified as a key monitoring market area, the power regulatory agency may increase the sampling frequency of the monitoring system to promptly detect new abnormal fluctuations in electricity prices.

[0200] The beneficial effects of the above technical solution are as follows: through the analysis of historical data and external factors, the market areas where electricity prices fluctuate abnormally frequently can be effectively identified and divided into key monitoring areas. In this way, power market operators can concentrate monitoring resources in these areas, discover and respond to potential risks in advance, and ensure the stability of the market. By distinguishing between key monitoring areas and non-key monitoring areas, detection resources can be reasonably allocated. More monitoring and detection resources are concentrated in key monitoring areas, while monitoring input is moderately reduced in non-key areas with less fluctuations, thereby improving the efficiency and cost-effectiveness of system operation. The impact of external factors such as weather, supply and demand, and fuel prices on electricity prices is fully considered in the model, which enhances the accuracy of electricity price fluctuation prediction. This can help market participants better plan electricity consumption strategies and reduce the uncertainty caused by electricity price fluctuations. Through real-time monitoring of continuous electricity price changes, once the electricity price in a certain market area approaches or reaches the predicted upper limit of fluctuations, the system can automatically issue an early warning to remind market participants to take measures to reduce the risks brought by market fluctuations.

[0201] In another embodiment, based on the time series feature information and a preset electricity price fluctuation feature template generation library, an electricity price fluctuation feature detection template is generated, including:

[0202] The number of information types used to parse temporal feature information;

[0203] When the number of information types is unique, the electricity price fluctuation detection rule corresponding to the time series characteristic information is determined from the electricity price fluctuation characteristic template generation library;

[0204] Based on the electricity price fluctuation detection rules, a template for detecting electricity price fluctuation characteristics is generated;

[0205] When the number of information types is not unique, the state information feature set of the time series feature information is parsed based on a preset feature parsing template;

[0206] Based on the state information feature set, construct a first feature description vector of the time series feature information;

[0207] Extracting multiple sets of one-to-one corresponding second feature description vectors and electricity price fluctuation detection rule sets from an electricity price fluctuation feature template generation library;

[0208] Calculating the vector similarity between the first feature description vector and any second feature description vector;

[0209] An electricity price fluctuation detection rule set corresponding to the second feature description vector of the maximum vector similarity is used to generate an electricity price fluctuation feature detection template.

[0210] The working principle of the above technical solution is as follows: the system first classifies the time series feature information and determines the number of its types. If the types of this information are unique, the system directly extracts the applicable electricity price fluctuation detection rules from the pre-built electricity price fluctuation feature template generation library.

[0211] Generation of electricity price fluctuation detection rules: Once the number of types of time series feature information is determined to be unique, the system will use the detection rules obtained from the template library to generate an electricity price fluctuation feature detection template. This template is used to monitor the fluctuation of electricity prices within a specific period of time.

[0212] Application of feature analysis template: When the number of types of time series feature information is not unique, the system will turn to the preset feature analysis template to further analyze the state information feature set of the time series feature information. The "state information feature set" here is understood as a set of data that describes the change pattern or fluctuation characteristics of electricity prices over a period of time.

[0213] Construction of feature description vector: By parsing the state information feature set, the system generates the first feature description vector of the time series feature information. This is a numerical description used to represent the overall pattern of these feature information.

[0214] Matching of feature description vectors: The system extracts multiple sets of one-to-one corresponding second feature description vectors and the set of power price fluctuation detection rules associated with these vectors from the power price fluctuation feature template generation library. Next, the system calculates the similarity between the first feature description vector and any second feature description vector.

[0215] Generate detection template based on maximum similarity: After finding the second feature description vector with the largest similarity, the system will extract rules from the corresponding electricity price fluctuation detection rule set to generate the final electricity price fluctuation feature detection template. This template will be used for monitoring and analysis of electricity price fluctuations.

[0216] Suppose we want to monitor the fluctuation of electricity prices in a city's electricity market over a certain period of time:

[0217] In the first step, we will first collect time series data on electricity prices, such as the changes in electricity prices every day.

[0218] In the second step, if we find that the characteristics of these data are single, such as the daily electricity price fluctuation patterns are similar, the system will directly extract the existing rules from the library for analysis and generate a simple electricity price fluctuation detection template for future monitoring.

[0219] In the third step, if there are multiple different patterns of electricity price fluctuations (for example, different peak and valley electricity prices, and different fluctuation characteristics in different time periods), the system will use complex feature parsing templates to analyze the characteristics of these multiple patterns and generate a detailed state information feature set.

[0220] In the fourth step, based on these feature sets, the system generates a first feature description vector.

[0221] In the fifth step, the system extracts multiple known feature description vectors from the template library and compares their similarity with the first feature description vector we generated one by one.

[0222] In the sixth step, the system will finally select the detection rule corresponding to the feature description vector with the highest similarity and generate a feature detection template suitable for the current electricity price fluctuations.

[0223] The beneficial effects of the above technical solution are as follows: the method can automatically generate an electricity price fluctuation detection template based on time series feature information without human intervention. This significantly improves efficiency, especially when faced with a large amount of complex electricity price fluctuation data, and can quickly generate a suitable monitoring template. By comparing the similarity of feature description vectors, the system can accurately match the most suitable electricity price fluctuation detection rule. The maximum similarity matching method ensures the accuracy of template generation and effectively reduces the possibility of false alarms or omissions. Regardless of whether the fluctuation of electricity prices is a single mode or a variety of modes, the method can automatically adjust. Especially when faced with complex and diverse information, the system can parse features according to preset templates and respond flexibly. By automatically generating electricity price fluctuation detection templates, the need for manual data processing is reduced, and the operation and maintenance costs of enterprises in power market monitoring are reduced.

[0224] In another embodiment, multiple electricity price ranges are set, including:

[0225] Obtaining a predicted electricity price value for the electricity price range to be set, where the predicted electricity price value is the electricity value predicted based on the electricity price prediction model within a continuous time period;

[0226] Acquire a reference electricity price data set, where the reference electricity price data set is a reference data set containing electricity price forecast values ​​in a preset electricity price data set;

[0227] Determine the division of the electricity price interval based on the time attribute value of the electricity price prediction value recorded in each reference data in the reference electricity price data set, the time attribute value of the electricity price prediction value recorded in the reference data being used to describe the time distribution of the electricity price prediction value;

[0228] Based on the time attribute value of the electricity price forecast value recorded in each reference data in the reference electricity price data set, the division of the electricity price range is determined, including:

[0229] According to the time attribute value of the electricity price prediction value recorded in the reference electricity price data set, the attribute value for dividing the electricity price interval is determined; wherein the electricity price prediction values ​​recorded in the reference electricity price data set are sorted according to the size of the attribute value, and for the sorted time attribute value sequence, the mean of any two adjacent time attribute values ​​in the sequence is calculated, and each mean value is used as the attribute value for dividing the electricity price interval;

[0230] Determine the division of the electricity price range based on the division attribute value of the electricity price range;

[0231] Based on the division attribute value of the electricity price range, the division of the electricity price range is determined, including:

[0232] Based on the division attribute values ​​of the electricity price interval, at least one interval division sequence of the electricity price is determined, and each interval division sequence of the electricity price includes at least one value of the division attribute values ​​of the electricity price interval; wherein all possible combinations of the division attribute values ​​of the electricity price interval are combined to obtain at least one interval division sequence of the electricity price, or some combinations of the division attribute values ​​of the electricity price interval are determined based on business knowledge to obtain at least one interval division sequence of the electricity price;

[0233] For each interval partition sequence, calculate the score corresponding to the interval partition sequence;

[0234] Determine the division of electricity price intervals based on the interval division sequence with the highest score;

[0235] The calculated score corresponding to the interval partition sequence includes a score used to indicate the order of the interval partition sequence.

[0236] The calculation of the attribute value of the electricity price interval is as follows: given the time attribute value sequence T in the reference electricity price data set = {t1, t2, ..., t n}, sort the time attribute value sequence to get T sorted ={t′1,t′2,……,t′ n}, where t′1≤t′2≤…≤t′ n .

[0237] Calculate the mean of any two adjacent time attribute values ​​to obtain the partition attribute value set:

[0238]

[0239] The working principle of the above technical solution is: the electricity price forecast value is obtained through the electricity price forecast model, which is based on the forecast of future electricity price trends within a certain period of time. Usually, the forecast model can be based on historical electricity price data, economic factors (such as supply and demand), weather conditions, etc. For example, we use a time series model to predict the hourly electricity price in the next week.

[0240] Get reference electricity price data set: The reference electricity price data set is selected from the preset electricity price data set, which contains the electricity value predicted by the model. For example, if the preset electricity price data set contains electricity price data for the past five years and the predicted year, we can select the data of a specific month as the reference electricity price data set.

[0241] Determine the division of electricity price intervals based on the reference electricity price data set: In this step, we need to determine the division of electricity price intervals based on the time attribute values ​​of the electricity price forecast values ​​recorded in each reference data (such as the electricity price forecast value for a certain hour). These time attribute values ​​are used to describe the distribution of electricity price forecast values ​​over time. For example, assuming that the reference data includes hourly electricity price forecast values, we can divide the electricity price intervals based on the distribution of these time attributes (such as 8 am and 3 pm).

[0242] Determine the partition attribute value of the electricity price interval: First, sort the electricity price forecast values ​​in the reference electricity price data set by time attribute value (such as by hourly order). Then, for the sorted time attribute value sequence, calculate the mean of adjacent time points. For example, if there are electricity values ​​at 8 am and 10 am in the reference data, then calculate the mean of 8 am and 10 am (i.e. 9 am), and use this mean as the partition attribute value.

[0243] Determine the electricity price interval based on the partition attribute value of the electricity price interval: The partition attribute value is used to determine the boundary of the electricity price interval. For example, if we get multiple partition attribute values ​​such as 9 o'clock, 11 o'clock, 13 o'clock, etc., then we can use these values ​​as the boundaries of different electricity price intervals to get intervals such as 9 o'clock to 11 o'clock, 11 o'clock to 13 o'clock, etc.

[0244] Determine the interval division sequence of electricity prices: In this step, we need to combine the attribute values ​​for the electricity price interval division. All possible combinations can be combined, or some combinations can be selected based on business experience. For example, we can choose to divide into intervals such as "morning peak (6 am to 9 am)", "off-peak (9 am to 6 pm)" and "evening peak (6 pm to 10 pm)".

[0245] Calculate the score corresponding to the interval partition sequence: Each interval partition sequence will have a score to indicate its order. The score can be calculated based on multiple factors, such as the fluctuation of electricity prices in each interval, the concentration of data in the interval, etc. If a certain partition can make the electricity price fluctuation in each interval smaller, then the score of this partition will be higher.

[0246] Based on the interval division sequence with the highest score, the final electricity price interval division is determined: the scores of all interval division sequences are compared, and the division with the highest score is selected as the final electricity price interval division result. This ensures that the divided interval is more scientific and reasonable in practical applications.

[0247] The beneficial effect of the above technical solution is that by accurately dividing the electricity price interval, the prediction accuracy of the electricity price prediction model can be improved. The characteristics of electricity price fluctuations in different time periods are different, and reasonable interval division can allow the model to capture these changes more accurately. Reasonable electricity price interval division is helpful for power dispatching and load forecasting. For example, power companies can optimize power supply strategies, reduce costs, and improve energy efficiency based on the difference in electricity prices during peak and trough periods. For enterprises participating in electricity market transactions, more flexible trading strategies can be formulated through interval division, such as reducing purchases during high-price periods and increasing purchases during low-price periods, thereby optimizing the cost of purchasing electricity. Reasonable electricity price interval division can better support the implementation of demand response measures. When the electricity price is higher in a certain interval, price signals can be used to guide users to reduce electricity consumption, and vice versa, to guide them to increase electricity consumption and achieve a balance in power load.

[0248] In another embodiment, a real-time electricity price prediction system based on a classification network includes:

[0249] A multi-source data acquisition unit is used to monitor the power market in real time and acquire multi-source data, including time series characteristics of electricity prices, status information of external influencing factors, and characteristics of electricity price fluctuations;

[0250] A classification network model unit, used to construct a classification network model based on multi-source data;

[0251] A prediction result acquisition unit is used to analyze the currently acquired multi-source data based on a classification network model, generate prediction results for multiple electricity price intervals, and evaluate the confidence level of the results for each interval;

[0252] Obtain the electricity price forecast report unit, based on the forecast results and the corresponding confidence level, comprehensively consider the market trend, analyze the future electricity price trend, generate an electricity price forecast report, and provide support for electricity market decision-making.

[0253] The working principle of the above technical solution is as follows: the multi-source data acquisition unit acquires various types of data from the power market, which are the basis for electricity price prediction. The classification network model unit, once the multi-source data is collected and processed, the classification network model unit models the data. The classification network is a machine learning model that can identify the trend of electricity prices under different circumstances by learning the patterns of historical data. The model divides electricity prices into different intervals (such as low electricity prices, medium electricity prices, and high electricity prices) based on time series data, external states, and fluctuation characteristics. The prediction result acquisition unit, through the classification network model, the system analyzes the current multi-source data and predicts which interval the electricity price will fall into. The system will give prediction results for multiple electricity price intervals and evaluate the confidence of each prediction result. For example, the system may predict that there is a 60% probability that the electricity price is in the medium price interval, a 30% probability that it is in the high price interval, and a 10% probability that it is in the low price interval.

[0254] Get the electricity price forecast report unit. Based on the forecast results and their corresponding confidence levels, the system will generate an electricity price forecast report. This report will not only provide forecasts for electricity prices in different intervals, but also analyze them based on market trends. The report will refer to changes in external factors, such as policies or energy supply and demand, to provide decision support for participants in the electricity market.

[0255] The beneficial effects of the above technical solution are: the system can obtain multi-source data of the power market in real time, not only limited to the time series characteristics of electricity prices, but also including external influencing factors. This enables the forecast to be adjusted in more dimensions, reflecting more realistic market changes. Through the introduction of the classification network model, the system can use the patterns of historical data and real-time status information to accurately classify and predict electricity prices. The self-learning and adjustment capabilities of the model enable it to optimize the forecast results based on new data at any time, thereby improving the accuracy of the forecast. The system not only gives a forecast of the electricity price range, but also provides confidence for each forecast result. This helps decision makers judge the reliability of the forecast, so as to make more reliable market operations, such as adjusting the electricity procurement strategy or formulating price response plans. The generated electricity price forecast report comprehensively considers multiple market dimensions, such as time series characteristics, fluctuation characteristics, external influencing factors, etc. Through comprehensive analysis, the report can provide detailed electricity price trends for participants in the power market, provide support for decisions such as electricity price pricing and purchasing power resources, and reduce market risks.

[0256] In another embodiment, obtaining the multi-source data unit includes:

[0257] The first multi-source data acquisition module is used to acquire the time series characteristic information of the electricity price in the power market when monitoring the multi-source real-time data in the power market;

[0258] The second multi-source data acquisition module is used to obtain status information of external influencing factors of the power market;

[0259] The third multi-source data acquisition module is used to generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library;

[0260] The fourth multi-source data acquisition module is used to acquire electricity price fluctuation characteristics based on the electricity price fluctuation characteristic detection template and according to the state information.

[0261] The working principle of the above technical solution is as follows: First, the multi-source data acquisition module is responsible for collecting the time series characteristic information of electricity prices. Time series characteristics refer to the pattern of electricity prices changing over time. For example, the fluctuation of electricity prices every hour or every day. The system monitors these changes in real time and records them.

[0262] The second multi-source data acquisition module focuses on the status information of external influencing factors. External influencing factors include but are not limited to weather conditions, energy supply, policy adjustments, etc. This information usually comes from different sensors or databases, and can update the external conditions of the power market in real time.

[0263] The third multi-source data acquisition module generates a template for detecting electricity price fluctuation characteristics based on the previously collected time series characteristic information and the preset electricity price fluctuation characteristic template generation library. In other words, it builds a model through historical data to identify the fluctuation pattern of electricity prices.

[0264] The fourth multi-source data acquisition module captures the characteristics of current electricity price fluctuations by applying the electricity price fluctuation feature detection template and combining it with external state information. This helps predict possible changes in electricity prices in the short term.

[0265] The beneficial effect of the above technical solution is that the system can obtain multi-source data of the power market in real time, which is not limited to the time series characteristics of electricity prices, but also includes external influencing factors. This enables the forecast to be adjusted in more dimensions and reflect more realistic market changes.

[0266] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A real-time electricity price prediction method based on a classification network, characterized in that: include: S101: Real-time monitoring of the power market to obtain multi-source data, including the time series characteristics of electricity prices, the status information of external influencing factors, and the characteristics of electricity price fluctuations; S102: Construct a classification network model based on multi-source data; S103: Based on the classification network model, the currently acquired multi-source data is analyzed to generate prediction results of multiple electricity price intervals, and the confidence of the results of each interval is evaluated; S104: Based on the prediction results and the corresponding confidence levels, the market trend is comprehensively considered to analyze the future electricity price trend and generate an electricity price forecast report to provide support for electricity market decision-making.

2. A real-time electricity price prediction method based on a classification network according to claim 1, characterized in that: Step S101 includes: S1011: when monitoring multi-source real-time data of the power market, obtaining time series characteristic information of the electricity price of the power market; S1012: Obtaining status information of external factors affecting the power market; S1013: Generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library; S1014: Based on the electricity price fluctuation feature detection template, the electricity price fluctuation feature is obtained according to the state information.

3. The real-time electricity price prediction method based on classification network according to claim 1 is characterized in that: Build a classification network model, including: Preprocess multi-source data, extract feature information related to real-time electricity price prediction, and use the feature information as multi-source data features; According to the characteristics of multi-source data, the correlation between multi-source data and real-time electricity prices is determined. The correlation is used to represent the impact of time series characteristics, external influencing factors and electricity price fluctuation characteristics on real-time electricity price prediction; Based on the correlation between multi-source data and real-time electricity prices, an initial classification network model is constructed. The classification network model predicts real-time electricity prices by inputting multi-source data features. The initial classification network model consists of multiple network layers and neural units. The network layer includes a time series layer, a state information processing layer, and a fluctuation feature extraction layer, which are used to process input data related to the time series features of electricity prices, the state information of external influencing factors, and the fluctuation features of electricity prices, respectively. Training the initial classification network model, and adjusting the weights in the classification network model using preset training set data, wherein the training set data includes labeled historical electricity prices and corresponding multi-source data features; According to the performance of the trained classification network model, the classification network model is optimized using a preset model optimization strategy. The model optimization strategy includes a back propagation algorithm based on error feedback, which is used to adjust the weight relationship between network layers to improve the accuracy of real-time electricity price prediction and obtain an optimized classification network model. Based on the optimized classification network model, the real-time input multi-source data is predicted, and the real-time electricity price prediction result is used as the output of the classification network model; In the classification network model, the effect of the classification network model is evaluated according to the preset model detection method to judge the performance of the classification network model in the real-time electricity price forecast, and the evaluation results are used as the basis for further model optimization; Based on the evaluation results of the classification network model, the classification network model finally used for real-time electricity price prediction is determined. The classification network model achieves high-precision prediction of real-time electricity prices through the input of multi-source data.

4. The real-time electricity price prediction method based on classification network according to claim 1 is characterized in that: Analyze the currently acquired multi-source data, including: Based on multi-source data, construct an input data set and input the input data set into the classification network model; Based on the analysis results of the classification network model, multiple electricity price intervals are set, and the electricity price interval is the predicted value range of electricity prices in a continuous time period; Conduct confidence assessment on the forecast results of the electricity price range, which is based on the fitting degree of the classification network model to the input data and the fluctuation state of external influencing factors; If the confidence level of a certain electricity price range is lower than a preset confidence level threshold, the prediction result of the electricity price range is re-analyzed or adjusted to generate a revised electricity price range; Output the forecast results of multiple electricity price ranges and the corresponding confidence information for electricity price regulation or market analysis; Among them, the setting of multiple electricity price intervals is based on the analysis of the time series characteristics of electricity prices by the classification network model, combined with the state information of external influencing factors, to dynamically generate multiple interval ranges; Confidence assessment includes backtesting based on historical data to ensure the accuracy and reliability of the forecast results.

5. The real-time electricity price prediction method based on classification network according to claim 1 is characterized in that: Generate electricity price forecast reports, including: Generate an electricity price forecast report based on the electricity price forecast results, and provide multiple scenario analyses of future electricity price trends based on the confidence level; Compare and analyze the data in the forecast report. When the deviation between the electricity price change trend in the forecast report and the historical or pre-stored data is greater than the first preset threshold, issue a first-level reminder and add analysis suggestions to the forecast report. When the deviation of the electricity price forecast result exceeds the second preset threshold, a secondary alarm is triggered, and a market risk warning and related response plan are attached to the forecast report; When the deviation of the electricity price forecast result exceeds the third preset threshold, a third-level alarm is triggered, and the report recommends taking emergency regulatory measures to avoid market volatility risks.

6. A real-time electricity price prediction method based on a classification network according to claim 2, characterized in that: Obtain status information of external factors affecting the power market, including: Obtain historical electricity price records in the electricity market; Based on historical electricity price records, the electricity market is divided into key monitoring market areas and non-key monitoring market areas; Adaptive allocation of detection resources for detecting electricity price fluctuation characteristics in key monitored market areas and non-key monitored market areas.

7. The real-time electricity price prediction method based on classification network according to claim 2 is characterized in that: Based on the time series feature information and the preset electricity price fluctuation feature template generation library, an electricity price fluctuation feature detection template is generated, including: The number of information types used to parse temporal feature information; When the number of information types is unique, the electricity price fluctuation detection rule corresponding to the time series characteristic information is determined from the electricity price fluctuation characteristic template generation library; Based on the electricity price fluctuation detection rules, a template for detecting electricity price fluctuation characteristics is generated; When the number of information types is not unique, the state information feature set of the time series feature information is parsed based on a preset feature parsing template; Based on the state information feature set, construct a first feature description vector of the time series feature information; Extracting multiple sets of one-to-one corresponding second feature description vectors and electricity price fluctuation detection rule sets from an electricity price fluctuation feature template generation library; Calculating the vector similarity between the first feature description vector and any second feature description vector; An electricity price fluctuation detection rule set corresponding to the second feature description vector of the maximum vector similarity is used to generate an electricity price fluctuation feature detection template.

8. The real-time electricity price prediction method based on classification network according to claim 4 is characterized in that: Set multiple electricity price ranges, including: Obtaining a predicted electricity price value for the electricity price range to be set, where the predicted electricity price value is the electricity value predicted based on the electricity price prediction model within a continuous time period; Acquire a reference electricity price data set, where the reference electricity price data set is a reference data set containing electricity price forecast values ​​in a preset electricity price data set; The division of electricity price intervals is determined based on the time attribute value of the electricity price forecast value recorded in each reference data in the reference electricity price data set. The time attribute value of the electricity price forecast value recorded in the reference data is used to describe the time distribution of the electricity price forecast value.

9. A real-time electricity price prediction system based on a classification network, characterized in that: include: A multi-source data acquisition unit is used to monitor the power market in real time and acquire multi-source data, including time series characteristics of electricity prices, status information of external influencing factors, and characteristics of electricity price fluctuations; A classification network model unit, used to construct a classification network model based on multi-source data; A prediction result acquisition unit is used to analyze the currently acquired multi-source data based on a classification network model, generate prediction results for multiple electricity price intervals, and evaluate the confidence level of the results for each interval; Obtain the electricity price forecast report unit, based on the forecast results and the corresponding confidence level, comprehensively consider the market trend, analyze the future electricity price trend, generate an electricity price forecast report, and provide support for electricity market decision-making.

10. A real-time electricity price prediction system based on classification network according to claim 9, characterized in that: The unit for obtaining multi-source data includes: The first multi-source data acquisition module is used to acquire the time series characteristic information of the electricity price in the power market when monitoring the multi-source real-time data in the power market; The second multi-source data acquisition module is used to obtain status information of external influencing factors of the power market; The third multi-source data acquisition module is used to generate an electricity price fluctuation feature detection template based on the time series feature information and a preset electricity price fluctuation feature template generation library; The fourth multi-source data acquisition module is used to acquire electricity price fluctuation characteristics based on the electricity price fluctuation characteristic detection template and according to the state information.

Images