Electric power spot transaction node electricity price prediction method

Abstract

The invention relates to the technical field of electric power market prediction, and discloses an electric power spot transaction node electricity price prediction method, which comprises the following steps of 1, obtaining historical transaction data of an electric power market, operation data of a generator set and interference data influencing electric power market transaction, and classifying the data to form a data set; 2, analyzing the fluctuation condition of the daily electricity price of the electricity market, and generating a corresponding floating index; 3, analyzing the daily supply and demand balance condition of the electricity market and the daily operation fluctuation condition of the generator set, and generating a corresponding monitoring data set; 4, setting a time axis, drawing a curve graph, and evaluating the relevance between different parameters and the electricity price; 5, setting a floating threshold value, judging a time node of abnormal fluctuation of the electricity price, and marking the time node in the curve graph; and step 6, constructing a prediction model according to the curve graph, the market data set, the monitoring data set and the influence data set.

Description

Technical Field

[0001] This invention relates to the field of electricity market forecasting technology, specifically a method for forecasting electricity prices at key points in electricity spot trading. Background Technology

[0002] In the current electricity market system, electricity spot trading refers to the buying and selling of electricity in the short term based on market supply and demand. Among them, nodal pricing, as the core price formation mechanism of the electricity spot market, specifically refers to the marginal price of electricity at a particular node in the power system. This price can directly reflect the scarcity of electricity and the supply cost at that node.

[0003] The electricity spot market exhibits significant price volatility due to its high complexity, with price formation influenced by multiple factors including supply and demand, fuel costs, environmental policies, and grid operation. Nodal prices, as marginal prices at specific nodes in the system, are not only constrained by local supply and demand balances but also coupled with the operational status of other regions within the system and transmission capacity limitations. To improve the accuracy of nodal price forecasts, it is necessary to construct forecasting models with comprehensive multi-dimensional factor analysis capabilities, dynamic market adaptability, and uncertainty information processing mechanisms.

[0004] Currently, traditional electricity spot market price forecasting methods mainly rely on single economic models or time series analysis techniques. These methods often fail to adequately consider multiple coupling factors and neglect key influencing variables such as weather fluctuations, seasonal changes, and unforeseen events. Furthermore, existing forecasting methods lack the ability to identify key data features, easily leading to inaccurate forecasts. The forecasting models also lack dynamic evolution mechanisms, making it difficult to adapt to the time-varying nature of market supply and demand. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for predicting electricity prices at nodes in the electricity spot market. This method has advantages such as strong multi-dimensional analysis and identification capabilities, and high applicability of accurate predictions. It solves the problem that traditional electricity price prediction methods cannot fully consider multiple coupling factors, resulting in inaccurate prediction results.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting electricity prices at key electricity spot trading nodes, comprising the following steps: Step 1: Connect to the big data platform via the network to obtain historical transaction data of the electricity market, operation data of generating units, and interference data affecting electricity market transactions, and classify them into market dataset, operation dataset, and impact dataset; Step 2: Based on the market dataset, analyze the daily fluctuations in electricity prices in the power market and generate corresponding floating indices. ; Step 3: Based on the market dataset and operational dataset, analyze the daily supply and demand balance of the electricity market and the daily operational fluctuations of generating units, and generate corresponding monitoring data sets. ; Step 4: Set the timeline Then, combined with market datasets and monitoring data sets And the impact on the dataset, plotting a curve. And assess the correlation between different parameters and electricity prices; Step 5: Set a fixed floating threshold value Combined with the floating index Determine the time points of abnormal electricity price fluctuations and display them on the curve. The corresponding annotations are made in the following: Step Six: Based on the curve graph Market datasets, monitoring data sets And influence the dataset, build a predictive model .

[0007] Preferably, in step one, the market dataset includes historical electricity prices, transaction volumes, and deficit amounts for each time point in the electricity market.

[0008] Preferably, in step one, the running dataset includes the output power of the generator set at each time point and the transmission power of the transmission line.

[0009] Preferably, in step one, the influencing dataset includes real-time temperature, humidity, wind speed, and the daily coal / gas futures market price.

[0010] Preferably, in step two, the floating index The calculation process is as follows: Based on the market dataset, extract the first electricity market data. Historical transaction data for the day, and the first day In Tianzhong, historical electricity prices at all time points are marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Historical electricity prices at specific time points; According to the standard deviation formula, we obtain the first... Historical electricity price fluctuation index .

[0011] Preferably, in step three, the monitoring data group The calculation process is as follows: Based on the market dataset, the first In the sky, the transaction volume at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The transaction volume at the 1st time point will be the 1st time point. In the sky, the amount of shortfall at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The amount of shortfall at each time point; Based on the running dataset, extract the generator set number The day's operating data, and the first In the sky, the output power at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... The output power at the nth time point will be the nth In the sky, the transmission power of the transmission lines at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Transmission power of transmission lines at each time point; No. Daily monitoring data group This includes the average ratio of deficit to trading volume, the standard deviation of power output, and the average transmission load rate.

[0012] Preferably, in step four, the curve graph The drawing process is as follows: Visualization software is used to visualize market datasets and monitoring data sets. and the impact dataset by time axis The data curves are arranged such that each data type corresponds to an independent data curve, and each data curve is distinguished by a different color. In the market dataset, the historical electricity price data curves for all time points are shown in Figure 1. Monitoring data group In the middle, the average ratio of the deficit to the transaction volume at all time points corresponds to data curve two; Monitoring data group In the middle, the standard deviation of the output power at all time points corresponds to data curve three; Monitoring data group In the middle, the average transmission load rate at all time points corresponds to data curve four; The data curves corresponding to the temperature at all time points in the dataset are shown in Figure 5. The data curves corresponding to humidity at all time points in the dataset are shown in section six. The data curves corresponding to wind speeds at all time points in the dataset are shown in Figure 7. The data set includes eight curves representing the coal / gas futures market prices at all time points.

[0013] Preferably, in step four, the correlation assessment process is as follows: According to the curve Using data curve one as the reference sequence, one data curve is then extracted from data curve two to data curve eight as the comparison sequence. If the values ​​in the reference sequence and the values ​​in the comparison sequence rise or fall synchronously within the same time period, it indicates that there is a positive correlation between the corresponding data curve and the electricity price. If, within the same time period, the value in the reference sequence increases while the value in the comparison sequence decreases, or vice versa, it indicates an inverse correlation between the corresponding data curve and the electricity price. The value changes in the reference sequence are recorded according to the time period. In the second time period, the value in the reference sequence shows an upward trend. If the value in the comparison sequence also shows the same upward trend in the first or third time period, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. During time period two, the values ​​in the reference sequence show a downward trend. If the values ​​in the comparison sequence also show the same downward trend during time period one or time period three, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. Correlation includes positive correlation, negative correlation, and interval interference.

[0014] Preferably, in step five, the floating index ≥Floating threshold (Dynamically set based on the standard deviation of historical electricity price fluctuations, typically using the floating index of the past 30 days) The 90th percentile indicates abnormal fluctuations in electricity prices. The timeframe of these abnormal price fluctuations (usually hourly or every 15 minutes) should be considered when analyzing the curve. The corresponding annotations are then made.

[0015] Preferably, in step six, the prediction model The build process is as follows: S1. According to the curve graph The time points marked with abnormal electricity price fluctuations are from the monitoring data group. In the impact dataset, parameters that are correlated before the time point of abnormal electricity price fluctuations are extracted as the experimental dataset, and historical electricity prices in the market dataset after the time point of abnormal electricity price fluctuations are extracted as the validation dataset. S2. Based on the experimental dataset, train and generate a prediction model. Output electricity price forecast results; S3. Set a fixed error threshold value. Judge the prediction model The accuracy is as follows: If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is less than the error threshold... This indicates that the prediction model... The high accuracy is the key to the final prediction model. ; If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is greater than or equal to the error threshold... If the result is negative, it indicates that the accuracy of the prediction model is low, and it should be retrained based on S1-S2 to generate a new prediction model. Until the new prediction model The absolute difference between the output electricity price prediction and the historical electricity price in the validation dataset is less than the error threshold. .

[0016] Compared with existing technologies, this invention provides a method for predicting electricity prices at nodes in electricity spot trading, which has the following advantages: 1. This invention connects to a big data platform via a network to acquire historical transaction data of the electricity market, operating data of generating units, and interference data affecting electricity market transactions. These data are then categorized into market datasets, operating datasets, and impact datasets, covering both direct and indirect factors influencing electricity prices. Based on the market datasets, the daily fluctuations in electricity prices are analyzed, generating corresponding floating indices. It sensitively reflects abnormal market phenomena. Based on market and operational datasets, it analyzes the daily supply and demand balance of the electricity market and the daily operational fluctuations of generating units, generating corresponding monitoring data sets. It has strong multidimensional analysis and identification capabilities.

[0017] 2. This invention achieves its goal by setting a time axis. Then, combined with market datasets and monitoring data sets And the impact on the dataset, plotting a curve. It also evaluates the correlation between different parameters and electricity prices, supporting multi-parameter linkage analysis. A fixed floating threshold value can be set. (Dynamically set based on the standard deviation of historical electricity price fluctuations, typically using the floating index of the past 30 days) (90th percentile), combined with the floating index Determine the time points of abnormal electricity price fluctuations (usually every hour or every 15 minutes) and display them on the graph. Make corresponding annotations within the graph. Market datasets, monitoring data sets And influence the dataset, build a predictive model It provides accurate predictions and has strong applicability. Attached Figure Description

[0018] Figure 1 This is a diagram illustrating the steps of the method of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Traditional methods for predicting electricity prices at key points in the electricity spot market primarily rely on single economic models or time series analysis techniques. These methods often fail to adequately consider multiple coupling factors and neglect key influencing variables such as weather fluctuations, seasonal changes, and unforeseen events. Furthermore, existing prediction methods lack the ability to identify key data features, leading to inaccurate predictions. The prediction models also lack dynamic evolution mechanisms, making it difficult to adapt to the time-varying nature of market supply and demand. Therefore, this invention provides a method for predicting electricity prices at key points in the electricity spot market. Please refer to [link / reference]. Figure 1 This includes the following steps: Step 1: Connect to the big data platform via the network to obtain historical transaction data of the electricity market, operation data of generating units, and interference data affecting electricity market transactions. Then, classify and form market datasets, operation datasets, and impact datasets to cover the direct and indirect factors affecting electricity prices. The market dataset includes historical electricity prices, trading volumes, and deficits for each point in time in the electricity market; The running dataset includes the output power of the generator set at each time point and the transmission power of the transmission line; The dataset is influenced by real-time temperature, humidity, wind speed, and the daily coal / gas futures market price. The big data platform includes, but is not limited to: provincial power trading center database (interface protocol IEEE1547-2018), National Meteorological Administration API (HTTPS / JSON protocol), and power generation group SCADA system (OPCUA protocol). Step 2: Based on the market dataset, analyze the daily fluctuations in electricity prices in the power market and generate corresponding floating indices. ; Floating Index The calculation process is as follows: Based on the market dataset, extract the first electricity market data. Historical transaction data for the day, and the first day In Tianzhong, historical electricity prices at all time points are marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Historical electricity prices at specific time points;

[0021]

[0022] In the formula, Indicates the first The average historical electricity price in Tianzhong. Indicates the first Tianzhong, No. Historical electricity prices at specific points in time , This indicates that, according to the standard deviation formula, the first... Historical electricity price fluctuation index It is sensitive to abnormal market phenomena; Step 3: Based on the market dataset and operational dataset, analyze the daily supply and demand balance of the electricity market and the daily operational fluctuations of generating units, and generate corresponding monitoring data sets. ; Monitoring data group The calculation process is as follows: Based on the market dataset, the first In the sky, the transaction volume at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The transaction volume at the 1st time point will be the 1st time point. In the sky, the amount of shortfall at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The amount of shortfall at each time point; Based on the running dataset, extract the generator set number The day's operating data, and the first In the sky, the output power at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... The output power at the nth time point will be the nth In the sky, the transmission power of the transmission lines at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Transmission power of transmission lines at each time point;

[0023] In the formula, Indicates the first The average output power in the sky;

[0024] In the formula, Indicates the first Tianzhong, No. Transaction volume at each time point , Indicates the first Tianzhong, No. The amount of shortfall at each time point. Indicates the first The ratio of the shortfall to the transaction volume at each time point. Indicates the first In the middle of the day, the average ratio of the amount of shortfall to the amount of trading volume. Indicates the first Tianzhong, No. Power output at each time point , This represents the standard deviation of the output power on day i. Indicates the first Tianzhong, No. Transmission power of transmission lines at each time point This represents a standard value used to measure the transmission power of power transmission lines. Indicates the first Transmission load rate at each time point Indicates the first The average value of transmission load rate in the sky. Indicates the first The monitoring data set for the day has strong multi-dimensional analysis and identification capabilities; Step 4: Set the timeline Then, combined with market datasets and monitoring data sets And the impact on the dataset, plotting a curve. It also evaluates the correlation between different parameters and electricity prices, and supports multi-parameter linkage analysis; Curve graph The drawing process is as follows: Visualization software is used to visualize market datasets and monitoring data sets. and the impact dataset by time axis The data curves are arranged such that each data type corresponds to an independent data curve, and each data curve is distinguished by a different color. In the market dataset, the historical electricity price data curves for all time points are shown in Figure 1. Monitoring data group In the middle, the average ratio of the deficit to the transaction volume at all time points corresponds to data curve two; This curve reflects the daily electricity deficit ratio, and its calculation formula is: deficit ratio = (Missing amount) / Transaction volume ) × 100%, where: shortfall amount Represented as a time point Electricity supply deficit (unit: MWh), transaction volume Represented as a time point Actual electricity trading volume (unit: MWh). When the deficit ratio is consistently higher than 15%, historical data shows that the electricity price curve will show a significant upward trend (the increase is usually >20%).

[0025] Monitoring data group In the middle, the standard deviation of the output power at all time points corresponds to data curve three; Monitoring data group In the middle, the average transmission load rate at all time points corresponds to data curve four; The data curves corresponding to the temperature at all time points in the dataset are shown in Figure 5. The data curves corresponding to humidity at all time points in the dataset are shown in section six. The data curves corresponding to wind speeds at all time points in the dataset are shown in Figure 7. The data curves corresponding to the coal / gas futures market prices at all time points in the dataset are shown in Figure 8. The correlation assessment process is as follows: According to the curve Using data curve one as the reference sequence, one data curve is then extracted from data curve two to data curve eight as the comparison sequence. If the values ​​in the reference sequence and the values ​​in the comparison sequence rise or fall synchronously within the same time period, it indicates that there is a positive correlation between the corresponding data curve and the electricity price. If, within the same time period, the value in the reference sequence increases while the value in the comparison sequence decreases, or vice versa, it indicates an inverse correlation between the corresponding data curve and the electricity price. The value changes in the reference sequence are recorded according to the time period. In the second time period, the value in the reference sequence shows an upward trend. If the value in the comparison sequence also shows the same upward trend in the first or third time period, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. During time period two, the values ​​in the reference sequence show a downward trend. If the values ​​in the comparison sequence also show the same downward trend during time period one or time period three, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. Correlation includes positive correlation, negative correlation, and interval interference, revealing non-immediate influencing factors and optimizing the time dimension of prediction; Step 5: Set a fixed floating threshold value Combined with the floating index Determine the time points of abnormal electricity price fluctuations and display them on the curve. The corresponding annotations are made in the following: Floating Index ≥Floating threshold (Dynamically set based on the standard deviation of historical electricity price fluctuations, typically using the floating index of the past 30 days) The 90th percentile indicates abnormal fluctuations in electricity prices. The timeframe of these abnormal price fluctuations (usually hourly or every 15 minutes) should be considered when analyzing the curve. The corresponding annotations are made in the following: Step Six: Based on the curve graph Market datasets, monitoring data sets And influence the dataset, build a predictive model ; Predictive Model The build process is as follows: S1. According to the curve graph The time points marked with abnormal electricity price fluctuations are from the monitoring data group. In the impact dataset, parameters that are correlated before the time point of abnormal electricity price fluctuations are extracted as the experimental dataset, and historical electricity prices in the market dataset after the time point of abnormal electricity price fluctuations are extracted as the validation dataset. S2. Based on the experimental dataset, train and generate a prediction model. Output electricity price forecast results; Specifically, this embodiment uses an Attention-LSTM hybrid model for training, and the Attention-LSTM hybrid model includes the following key components: LSTM layer: used to capture the time-series features of electricity price data; Attention mechanism layer: used to dynamically allocate weights and highlight information in the time series that is more valuable for prediction; Output layer: Outputs electricity price prediction results; Specific settings for each layer: LSTM layer parameters: number of hidden units h=128. The number of hidden units determines the capacity and complexity of the LSTM layer. 128 hidden units can effectively extract long-term dependencies in electricity price data, while avoiding overfitting caused by an overly complex model.

[0026] Attention mechanism layer parameters: Weight calculation formula: , in:

[0027] , , This represents the trainable parameter matrix, used to calculate the attention weights; Let be the LSTM hidden state at time t; s be the context vector, representing the global context information at the current time step; the attention mechanism calculates the weights at each time step. Dynamically allocate information that is more valuable for prediction, thereby improving the model's predictive performance; Loss function type: Using Huber loss function: Calculation formula: , In the electricity price forecasting task, y represents the actual value of electricity; The predicted value represents the value calculated by the model f from the input feature x; δ is the threshold parameter of the Huber loss function. The Huber loss function combines the advantages of mean squared error (MSE) and absolute error (MAE), using squared error for small errors and linear error for large errors, which can effectively balance the model's sensitivity to outliers. Recommended parameter: δ=1 (adjustable hyperparameter), used to control the switching between the squared term and the linear term of the error. The recommended value is 1, but it can be adjusted according to the actual dataset to optimize model performance. The following is the training and prediction process for the Attention-LSTM hybrid model: Data preparation: Collect experimental datasets, including historical electricity price data and related characteristics (such as weather data, electricity demand, etc.). Data preprocessing, including normalization and handling of missing values, is performed to ensure data quality.

[0028] Model training: The preprocessed dataset is divided into a training set and a validation set; the Attention-LSTM model is trained using the training set to optimize the LSTM layer parameters and the attention mechanism layer parameters; During training, the prediction error is calculated using the Huber loss function and the model parameters are updated via backpropagation; the model performance is evaluated periodically on the validation set, and hyperparameters (such as δ) are tuned to optimize the model. Model prediction: Using a trained model to predict electricity prices based on new input data; Output forecast results: The forecast results are given in numerical form, representing the predicted future electricity prices.

[0029] During training, if the validation set loss does not decrease for 10 consecutive rounds, an early stopping mechanism is triggered.

[0030] S3. Set a fixed error threshold value. Judge the prediction model The accuracy is as follows: If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is less than the error threshold... This indicates that the prediction model... The high accuracy is the key to the final prediction model. ; If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is greater than or equal to the error threshold... If the result is negative, it indicates that the accuracy of the prediction model is low, and it should be retrained based on S1-S2 to generate a new prediction model. Until the new prediction model The absolute difference between the output electricity price prediction and the historical electricity price in the validation dataset is less than the error threshold. .

[0031] Example 1 Curve definition explanation:

[0032] 1. Electricity price time series curve (curve one) Data source: Historical electricity price data Data Description: This data records the changes in electricity prices over a past period, presented as a time series. The curve reflects the dynamic trend of electricity price changes and is the primary target variable for the electricity price forecasting model.

[0033] Purpose: To serve as the target value for model training and to evaluate the model's predictive performance.

[0034] 2. Supply-demand gap ratio curve (curve two) Data source: Electricity market transaction data Data Description: The calculation formula represents the daily average ratio of the power deficit to the trading volume, reflecting the degree of imbalance between supply and demand in the electricity market. The power deficit refers to the portion of electricity supply that is insufficient, while the trading volume refers to the total electricity traded in the market.

[0035] Function: As one of the important factors affecting electricity prices, the supply-demand gap ratio can reflect the supply and demand relationship in the electricity market and has important reference value for electricity price forecasting.

[0036] 3. Power generation output fluctuation curve (curve three) Data source: Power generation system operation data Data Description: Obtained by calculating the standard deviation of power generation, it reflects the fluctuation of power output. The larger the standard deviation of power generation, the worse the stability of power output.

[0037] Function: Fluctuations in power generation output affect the supply and demand balance of the power system and electricity prices, and are one of the important features of the model input.

[0038] 4. Transmission load rate curve (curve four) Data source: Power transmission system operation data Data Description: The calculation formula is the ratio of the transmission power of a transmission line to its rated capacity, reflecting the load condition of the transmission line. A higher transmission load rate indicates a higher utilization rate of the transmission line, but may also mean greater operational pressure on the power system.

[0039] Function: Transmission load factor is an important indicator for assessing the operating status of the power system and has an indirect impact on electricity price forecasting.

[0040] 5. Temperature Influence Curve (Curve 5) Data source: Weather station temperature data Data Description: This data records temperature measurements from weather stations and is presented as a time series. Temperature changes affect electricity demand (such as the use of air conditioning and heating equipment), which in turn affects electricity prices.

[0041] Function: As one of the meteorological factors, temperature has a significant impact on electricity demand and electricity prices, and is an important feature of the model input.

[0042] 6. Humidity Effect Curve (Curve Six) Data source: Humidity data from weather stations Data Description: This data records humidity measurements taken at a weather station and is presented as a time series. Changes in humidity can affect the operating efficiency of electrical equipment and electricity demand.

[0043] Function: Humidity, as one of the meteorological factors, has a certain impact on the operation of the power system and electricity prices, and is an auxiliary feature for model input.

[0044] 7. Wind speed influence curve (curve seven) Data source: Wind speed data from weather stations Data Description: This data records wind speed measurements taken at a weather station and is presented as a time series. Changes in wind speed affect the output of wind power generation, which in turn affects the supply and demand balance of the power system and electricity prices.

[0045] Function: Wind speed is an important factor affecting renewable energy power generation and has important reference value for electricity price forecasting.

[0046] 8. Fuel cost transmission curve (curve eight) Data source: Energy market data Data Description: This data records changes in the coal / gas price index, reflecting dynamic changes in fuel costs. Fuel costs are a major component of power generation costs, and their changes are reflected in electricity prices.

[0047] Function: The fuel cost transmission curve is an important indicator for assessing changes in power generation costs and has a direct impact on electricity price forecasting.

[0048] In this experiment, actual transaction data from the East China Power Grid in January 2023 was selected as an example. The historical electricity price fluctuation range was 420-580 yuan / MWh. Statistics show that the historical electricity price changed five times on March 1st, at 500 yuan / MWh, 520 yuan / MWh, 510 yuan / MWh, 490 yuan / MWh, and 530 yuan / MWh. The fluctuation index of the historical electricity price on March 1st... The calculation process is as follows:

[0049] In the formula, This represents the average historical electricity price on March 1st. Using the standard deviation formula, the fluctuation index of the historical electricity price on March 1st is obtained. Approximately Floating threshold Set at 10 yuan / MWh. Based on the historical electricity price fluctuation index on March 1st... > Floating threshold This indicates that electricity prices fluctuated abnormally on March 1st. The graph should be adjusted based on the timing of this abnormal price fluctuation. The corresponding annotations are then made.

[0050] Example 2 In this experiment, a curve graph was selected. Two types of parameters were used as experimental subjects, based on the curve. Using data curve one as the reference sequence and data curve eight as the comparison sequence, statistics show that within three hours, the historical electricity price was 300 yuan / MWh in the first hour, 300 yuan / MWh in the second hour, and 400 yuan / MWh in the third hour. The fuel price was 900 yuan / ton in the first hour, 1000 yuan / ton in the second hour, and 1000 yuan / ton in the third hour. The correlation assessment process between fuel price and electricity price is as follows: The value changes in the reference sequence are recorded according to the time period. From the first hour to the second hour, the value in the reference sequence time period one remains unchanged. From the second hour to the third hour, the value in the reference sequence time period two shows an upward trend. The changing trends of values ​​in the comparison sequence are recorded according to time periods. From the first hour to the second hour, the values ​​in the first time period of the comparison sequence show an upward trend, while from the second hour to the third hour, the values ​​in the second time period of the comparison sequence remain unchanged. It was determined that when the value in the reference sequence time period two showed an upward trend, the value in the comparison sequence time period one also showed the same upward trend. This indicates that there is an interval interference relationship between data curve eight and data curve one, that is, there is an interval interference relationship between fuel price and electricity price.

[0051] Example 3 Test data: This experiment selected full-year data from the East China regional electricity market in 2023, covering 8,760 hours of electricity price data and related characteristics, to evaluate the predictive performance of the model.

[0052] To verify the prediction model of this invention To assess the effectiveness, the following two traditional models are selected for comparison: ARIMA (Autoregressive Integral Moving Average): A classic linear time series forecasting model, widely used in economic and financial market forecasting; XGBoost (Gradient Boosting Tree Based on Feature Engineering): A tree-based ensemble learning model with powerful feature learning capabilities and predictive performance; The following evaluation metrics are used to quantitatively compare model performance: Mean Absolute Error (MAE): Measures the average error between the predicted and actual values, expressed in yuan / MWh.

[0053] Root mean square error (RMSE): measures the standard deviation of prediction error, measured in yuan / MWh.

[0054] Peak error rate: measures the proportion of error in the model during peak electricity price periods.

[0055] Training time: Records the time required to train the model, in hours.

[0056] Experimental comparison results: (The optimal order for the ARIMA model is (2,1,1)), XGBoost parameter selection: max_depth=6 is the maximum depth of a single decision tree; learning_rate=0.1 is the weight reduction factor for each tree; subsample=0.8 randomly samples 80% of the training data; colsample_bytree=0.8 forces each tree to use only 80% of the features.

[0057]

[0058] As can be seen from the table: Predictive Model It significantly outperforms ARIMA and XGBoost in MAE and RMSE, indicating that it has higher prediction accuracy; Predictive Model Its peak error rate is only 12.3%, far lower than ARIMA's 34.6% and XGBoost's 28.1%, indicating that its prediction performance is more stable during peak electricity price periods; Despite the predictive model Its training time (2.1 hours) is slightly longer than XGBoost (1.5 hours) and ARIMA (0.3 hours), but its advantages in prediction accuracy and stability make it more valuable in practical applications.

[0059] Key Scenario Analysis: Normal Market Conditions: Predictive Model The error of the predictive model is ±8 yuan / MWh, while the error of the traditional model is ±15 yuan / MWh, indicating that the predictive model... It has higher forecasting accuracy under normal market conditions.

[0060] Extreme Typhoon Events:

[0061] As can be seen from the table: Predictive Model The maximum instantaneous error under extreme events is 52 yuan / MWh, which is significantly lower than ARIMA's 89 yuan / MWh and XGBoost's 76 yuan / MWh.

[0062] Predictive Model The recovery time to steady state is only 3 hours, far less than ARIMA's 12 hours and XGBoost's 8 hours, indicating that it responds faster under extreme events and can recover to a steady state more quickly.

[0063] Experimental results show that the prediction model of this invention... It outperforms traditional methods in all aspects, including prediction accuracy, peak error rate, and response speed to extreme events. Compared to the ARIMA model, the prediction model... The prediction accuracy improved by 48% (MAE 9.8 vs 18.7), and the response speed to extreme events was 4 times faster. Furthermore, the prediction model... By automatically capturing key features through the attention mechanism, the model eliminates the need for manually designed lag terms, further improving its applicability and flexibility.

[0064] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A method for predicting electricity prices at key nodes in a spot electricity trading market, characterized in that, Includes the following steps: Step 1: Connect to the big data platform via the network to obtain historical transaction data of the electricity market, operation data of generating units, and interference data affecting electricity market transactions, and classify them into market dataset, operation dataset, and impact dataset; Step 2: Based on the market dataset, analyze the daily fluctuations in electricity prices in the power market and generate corresponding floating indices. ; Step 3: Based on the market dataset and operational dataset, analyze the daily supply and demand balance of the electricity market and the daily operational fluctuations of generating units, and generate corresponding monitoring data sets. ; Step 4: Set the timeline Then, combined with market datasets and monitoring data sets And the impact on the dataset, plotting a curve. And assess the correlation between different parameters and electricity prices; Step 5: Set a fixed floating threshold value Combined with the floating index Determine the time points of abnormal electricity price fluctuations and display them on the curve. The corresponding annotations are made in the following: Step Six: Based on the curve graph Market datasets, monitoring data sets And influence the dataset, build a predictive model .

2. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: In step one, the market dataset includes historical electricity prices, transaction volumes, and deficits for each time point in the electricity market.

3. The method for predicting electricity prices at electricity spot trading nodes according to claim 1, characterized in that: In step one, the running dataset includes the output power of the generator set at each time point and the transmission power of the transmission line.

4. The method for predicting electricity prices at electricity spot trading nodes according to claim 1, characterized in that: In step one, the influencing dataset includes real-time temperature, humidity, wind speed, and the daily coal / gas futures market price.

5. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: In step two, the floating index The calculation process is as follows: Based on the market dataset, extract the first electricity market data. Historical transaction data for the day, and the day's data. In Tianzhong, historical electricity prices at all time points are marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Historical electricity prices at specific time points; According to the standard deviation formula, we obtain the first... Historical electricity price fluctuation index .

6. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: In step three, the monitoring data group The calculation process is as follows: Based on the market dataset, the first In the sky, the transaction volume at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The transaction volume at the 1st time point will be the 1st time point. In the sky, the amount of shortfall at all time points is marked as , to Indicates the first In the middle of the sky, the first to arrive at the... The amount of shortfall at each time point; Based on the running dataset, extract the generator set number The day's operating data, and the first In the sky, the output power at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... The output power at the nth time point will be the nth In the sky, the transmission power of the transmission lines at all time points is marked as follows: , to Indicates the first In the middle of the sky, the first to arrive at the... Transmission power of transmission lines at each time point; No. Daily monitoring data group This includes the average ratio of deficit to trading volume, the standard deviation of power output, and the average transmission load rate.

7. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: In step four, the curve graph The drawing process is as follows: Visualization software is used to visualize market datasets and monitoring data sets. and the impact dataset by time axis The data curves are arranged such that each data type corresponds to an independent data curve, and each data curve is distinguished by a different color. In the market dataset, the historical electricity price data curves for all time points are shown in Figure 1. Monitoring data group In the middle, the average ratio of the deficit to the transaction volume at all time points corresponds to data curve two; Monitoring data group In the middle, the standard deviation of the output power at all time points corresponds to data curve three; Monitoring data group In the middle, the average transmission load rate at all time points corresponds to data curve four; The data curves corresponding to the temperature at all time points in the dataset are shown in Figure 5. The data curves corresponding to humidity at all time points in the dataset are shown in section six. The data curves corresponding to wind speeds at all time points in the dataset are shown in Figure 7. The data set includes eight curves representing the coal / gas futures market prices at all time points.

8. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: In step four, the correlation assessment process is as follows: According to the curve Using data curve one as the reference sequence, one data curve is then extracted from data curve two to data curve eight as the comparison sequence. If the values ​​in the reference sequence and the values ​​in the comparison sequence rise or fall synchronously within the same time period, it indicates that there is a positive correlation between the corresponding data curve and the electricity price. If, within the same time period, the value in the reference sequence increases while the value in the comparison sequence decreases, or vice versa, it indicates an inverse correlation between the corresponding data curve and the electricity price. The value changes in the reference sequence are recorded according to the time period. In the second time period, the value in the reference sequence shows an upward trend. If the value in the comparison sequence also shows the same upward trend in the first or third time period, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. During time period two, the values ​​in the reference sequence show a downward trend. If the values ​​in the comparison sequence also show the same downward trend during time period one or time period three, it indicates that there is an interval interference relationship between the corresponding data curve and the electricity price. Correlation includes positive correlation, negative correlation, and interval interference.

9. The method for predicting electricity prices at nodes in a spot electricity trading market according to claim 1, characterized in that: Step five, the floating index ≥Floating threshold (Dynamically set based on the standard deviation of historical electricity price fluctuations, typically using the floating index of the past 30 days) The 90th percentile indicates abnormal fluctuations in electricity prices. The timeframe of these abnormal price fluctuations (usually hourly or every 15 minutes) should be considered when analyzing the curve. The corresponding annotations are then made.

10. The method for predicting electricity prices at electricity spot trading nodes according to claim 1, characterized in that: In step six, the prediction model The build process is as follows: S1. According to the curve graph The time points marked with abnormal electricity price fluctuations are from the monitoring data group. In the impact dataset, parameters that are correlated before the time point of abnormal electricity price fluctuations are extracted as the experimental dataset, and historical electricity prices in the market dataset after the time point of abnormal electricity price fluctuations are extracted as the validation dataset. S2. Based on the experimental dataset, train and generate a prediction model. Output electricity price forecast results; S3. Set a fixed error threshold value. Judge the prediction model The accuracy is as follows: If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is less than the error threshold... This indicates that the prediction model... The high accuracy is the key to the final prediction model. ; If the absolute difference between the predicted electricity price and the historical electricity price in the validation dataset is greater than or equal to the error threshold... If the result is negative, it indicates that the accuracy of the prediction model is low, and it should be retrained based on S1-S2 to generate a new prediction model. Until the new prediction model The absolute difference between the output electricity price prediction and the historical electricity price in the validation dataset is less than the error threshold. .

Images