Electric power spot transaction auxiliary service decision-making system

Abstract

The invention discloses an electric power spot transaction auxiliary service decision-making system, and relates to the field of electric power spot transaction, and the system comprises a transaction data acquisition module, a transaction data prediction module, a transaction auxiliary service module, a transaction service decision-making module, and an execution feedback correction module. Wherein the transaction data acquisition module is used for acquiring a spot transaction data set in real time; the transaction data prediction module is used for extracting transaction feature data, analyzing a change trend and generating a future market parameter set; the transaction auxiliary service module is used for generating a transaction auxiliary service scheme set; the transaction service decision module is used for selecting a target transaction auxiliary service scheme; the execution feedback correction module is used for executing the target transaction auxiliary service scheme and dynamically correcting the auxiliary service emphasis rule, and the whole-process intelligent electric power spot transaction auxiliary decision-making system is constructed in a modularized mode, so that the intelligence, automation and adaptive adjustment capacity of electric power spot market transaction decision making are improved.

Description

Technical Field

[0001] This invention relates to the field of electricity spot trading, and more particularly to an auxiliary service decision-making system for electricity spot trading. Background Technology

[0002] With the rapid development of the global energy market and the continuous growth of electricity demand, the electricity spot market has become an important part of modern electricity trading. However, electricity prices in the electricity spot market are highly volatile and affected by multiple factors, such as electricity supply and demand, weather changes, energy policies, and market sentiment. This makes electricity spot market trading face high risks and complexities. Therefore, making efficient and accurate trading decisions in a dynamically changing market environment has become an important issue for electricity market operators and participants.

[0003] Meanwhile, ancillary services for electricity spot trading play a vital role in the modern electricity market. Especially as the electricity market gradually moves towards liberalization and marketization, the electricity spot market is becoming a core component of electricity trading. Electricity spot trading ancillary services determine electricity prices in real time based on supply and demand, resulting in very frequent price fluctuations. By optimizing the price discovery process in the electricity market through ancillary services, the efficiency of market allocation of electricity resources can be improved. Furthermore, electricity spot trading ancillary services can help market participants (such as power generation companies and power purchasing companies) make more timely and accurate trading decisions by predicting and analyzing market prices and supply and demand, thereby reducing market uncertainty and information asymmetry.

[0004] Existing power spot trading auxiliary service decision-making systems mainly rely on manual analysis and traditional forecasting models to predict power market prices and demand. However, these systems fail to fully capture the complexity and nonlinearity of market fluctuations, resulting in low forecast accuracy and an inability to effectively respond to sudden market changes. Furthermore, these systems typically focus on a single objective and lack comprehensive evaluation and optimization of multiple objectives and factors, thus failing to provide personalized decision-making solutions. Currently, no effective solutions have been proposed to address these technical issues. Summary of the Invention

[0005] To address the problems in related technologies, this invention proposes an auxiliary service decision-making system for electricity spot trading, thereby overcoming the aforementioned technical problems existing in the current related technologies.

[0006] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows: A power spot trading auxiliary service decision system, comprising: a transaction data acquisition module, a transaction data prediction module, a transaction auxiliary service module, a transaction service decision module, and an execution feedback correction module; The transaction data acquisition module is used to acquire electricity spot transaction data in real time and form a spot transaction dataset through data cleaning and integration. The transaction data prediction module is used to extract transaction feature data from the spot transaction dataset, analyze the changing trend of the transaction feature data, and predict and generate a future market parameter set based on the changing trend. As a preferred embodiment, the transaction data prediction module includes: a data feature extraction module, a data trend analysis module, a future data prediction module, and a result optimization and correction module; The data feature extraction module is used to extract transaction feature data from the spot transaction dataset and to preprocess the transaction feature data. The data trend analysis module is used to perform time series analysis on the preprocessed transaction feature data and classify the transaction feature data into market period fluctuations based on the time series analysis results. The future data prediction module is used to predict the trend of market fluctuation classification results during market periods using a long short-term memory network algorithm, and to generate a set of future market parameters. As a preferred embodiment, the future data prediction module includes: a data processing module, a future prediction module, an evaluation and optimization module, and a result output module; The data processing module is used to normalize the market period fluctuation classification results and fill in missing data. The future prediction module is used to construct a long-short-term prediction model through a long short-term memory network algorithm, and input the normalized market period fluctuation classification results into the long-short-term prediction model to predict market long-short-term data. The evaluation and optimization module is used to evaluate the error value of the market long-term and short-term data forecast results, and to adjust and optimize the market long-term and short-term data forecast results based on the error value evaluation results. The result output module is used to convert the adjusted and optimized market short-term and long-term data prediction results into a future market parameter set output.

[0007] The result optimization and correction module is used to evaluate the data error between the future market parameter set and the historical spot transaction dataset, and to verify and optimize the future market parameter set based on the data error evaluation results and the Bayesian algorithm.

[0008] As a preferred embodiment, the result optimization and correction module includes: an error evaluation module, an error correction module, a correction feedback module, and a verification output module; The error assessment module is used to calculate the error between the future market parameter set and the historical spot transaction dataset, and to assess the accuracy of the parameters of the future market parameter set based on the error calculation results. The error correction module is used to preset an error correction strategy library, match the error correction strategy library with the parameter accuracy, and use the matched error correction strategy to correct the future market parameter set. The correction feedback module is used to compare the corrected future market parameter set with the historical spot transaction dataset, and to evaluate the correction result based on the comparison result. The verification output module is used to add annotations to the future market parameter set based on the correction results and then output the results.

[0009] The transaction assistance service module is used to preset assistance service emphasis rules and combine the assistance service emphasis rules with the spot transaction dataset and the future market parameter set to generate a transaction assistance service scheme set; As a preferred embodiment, the transaction assistance service module includes: a rule setting module, a solution generation module, a solution evaluation module, and a verification output module; The rule setting module is used to set multi-faceted emphasis rules for auxiliary services according to market operation requirements; The solution generation module is used to combine the spot transaction dataset and the future market parameter set, and generate an auxiliary service solution set based on multi-faceted emphasis rules. The solution evaluation module is used to set solution evaluation indicators and perform feasibility evaluation on the ancillary service solution set based on the solution evaluation indicators. As a preferred embodiment, the solution evaluation module includes: a performance evaluation module, an optimization suggestion module, a risk evaluation module, and a feasibility evaluation module; The performance evaluation module is used to set the scheme evaluation indicators and to perform quantitative performance evaluation of the auxiliary service scheme set based on the scheme evaluation indicators. The optimization suggestion module is used to set an optimization suggestion threshold, match the optimization suggestion threshold with the performance quantification evaluation results, and generate an optimization suggestion scheme based on the matching results. The risk assessment module is used to perform quantitative analysis on the ancillary service solution set and assess market volatility risk. The feasibility assessment module is used to combine the optimized proposed solution with market volatility risk to conduct a feasibility assessment, and to verify and output the feasibility assessment results.

[0010] The verification output module is used to verify the feasibility assessment results and select auxiliary service schemes from the auxiliary service scheme set based on the verified feasibility assessment results.

[0011] The transaction service decision module is used to verify and optimize the set of transaction auxiliary service solutions, and select the target transaction auxiliary service solution based on the verification and optimization results. The execution feedback correction module is used to execute the target transaction auxiliary service scheme, record transaction execution data, and dynamically correct the auxiliary service emphasis rules based on the feedback analysis results of the transaction execution data.

[0012] As a preferred embodiment, the execution feedback correction module includes: a real-time monitoring module, a feedback analysis module, a correction strategy module, and a correction execution module; The real-time monitoring module is used to monitor market data and various monitoring parameters during the execution process in real time, and output them as feedback information. The feedback analysis module is used to analyze the deviation value of the feedback information through an anomaly detection algorithm; The correction strategy module is used to preset a deviation value threshold strategy library, match the deviation value threshold strategy library with the deviation value analysis results, and select a deviation value threshold strategy based on the matching results. As a preferred embodiment, the correction strategy module includes: a data receiving module, a strategy selection module, a strategy simulation module, and an adjustment output module; The data receiving module is used to receive the error calculation results from the error evaluation module; The strategy selection module is used to preset a strategy selection library and select an initial correction strategy from the preset strategy selection library based on error assessment results and changes in market parameters. The strategy simulation module is used to simulate and test the selected initial correction strategy using a data recalibration algorithm, and to evaluate the applicability of the initial correction strategy under the current market data. As a preferred embodiment, the strategy simulation module includes: a data relabeling module, an applicability evaluation module, and a simulation result output module; The data recalibration module is used to recalibrate the future market parameter set and integrate the recalibrated future market parameter set with a scale standard. The applicability assessment module is used to modify the initial correction strategy by incorporating the future market parameter set after the scaling standard is integrated, and to calculate the applicability and applicable time period of the strategy correction. As a preferred embodiment, the applicability assessment module includes: an error analysis module, an applicability calculation module, and a time period determination module; The error analysis module is used to apply the initial correction strategy to the future market parameter set after the scale standard is integrated, and to calculate the prediction gap between the predicted value after strategy correction and the expected target value. The applicability calculation module is used to evaluate the performance of the correction strategy based on the prediction gap value, generate an applicability score, and make an applicability judgment based on the applicability score. The time period determination module is used to determine the applicable range of the initial correction strategy in the future time series through a sliding window backtesting algorithm, and generate the corresponding applicable time period information.

[0013] The simulation result output module is used to evaluate the applicability of the initial correction strategy under the current market data based on the applicability of the strategy correction and the applicable time period, and then output the results.

[0014] The adjustment output module is used to adjust and optimize the initial correction strategy based on the simulation test results, form a deviation threshold correction strategy for controlling the prediction deviation within a preset range, and output it as the final correction result.

[0015] The correction execution module is used to perform correction operations according to the selected deviation value threshold strategy and record correction parameters in real time.

[0016] The beneficial effects of this invention are as follows: 1. This invention deeply integrates transaction data acquisition, data prediction, auxiliary service generation, service decision-making, and execution feedback through a modular architecture. It constructs a full-process intelligent power spot market transaction auxiliary decision-making system that spans data collection, market prediction, decision evaluation, and dynamic correction. This enhances the intelligence, automation, and adaptive adjustment capabilities of power spot market transaction decisions. The transaction data acquisition module collects real-time transaction data from the power spot market and constructs a standardized market dataset through data cleaning and integration techniques. This ensures that system decisions are based on high-quality, accurate market data, providing a solid data foundation for subsequent predictions and decisions. Simultaneously, the transaction data prediction module extracts transaction characteristics and predicts fluctuation trends based on historical market data and future trend analysis. It employs a Long Short-Term Memory (LSTM) network algorithm for time-series prediction of market fluctuations and combines this with a Bayesian algorithm to optimize and correct the future market parameter set, improving the accuracy and adaptability of predictions. Furthermore, through error correction and feedback mechanisms, the market prediction results are further optimized, ensuring that the system's output market prediction results remain efficient and accurate in complex market environments.

[0017] 2. This invention sets multi-faceted rules through a transaction support service module, combining market demand and environmental changes to generate a multi-dimensional support service solution set driven by rules. A multi-objective collaborative optimization mechanism is used to quantitatively evaluate and predict the risks of these solutions. In the support service solution evaluation module, the feasibility of the solutions is analyzed based on market fluctuations, and customized support service solutions are provided in conjunction with risk assessment and optimization suggestions, ensuring the stability and feasibility of the solutions in different market scenarios. Simultaneously, an execution feedback correction module constructs a dynamic feedback mechanism for decision execution, monitoring various parameters in the market execution process in real time. Feedback analysis is used to analyze and correct errors in the execution data, dynamically adjusting the execution strategy to ensure continuous adaptability and accuracy of the strategy amidst market fluctuations. Furthermore, by correcting the matching of the strategy library and the execution of deviation-value strategies, the transaction support service strategy can be adaptively adjusted, enhancing the system's self-healing capabilities and long-term stability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a system block diagram of an auxiliary service decision-making system for electricity spot trading according to an embodiment of the present invention.

[0020] In the picture: 1. Transaction data acquisition module; 2. Transaction data prediction module; 3. Transaction auxiliary service module; 4. Transaction service decision module; 5. Execution feedback correction module. Detailed Implementation

[0021] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0022] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0023] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1As shown, the power spot trading auxiliary service decision system according to an embodiment of the present invention includes: a transaction data acquisition module 1, a transaction data prediction module 2, a transaction auxiliary service module 3, a transaction service decision module 4, and an execution feedback correction module 5. The transaction data acquisition module 1 is used to acquire electricity spot transaction data in real time and form a spot transaction dataset through data cleaning and data integration. Specifically, through a data access unit deployed at the interface layer of the power trading system, publicly disclosed market transaction matching data, real-time electricity prices, power generation and consumption plans, and ancillary service quotations are collected in situ. This data access unit consists of an API call module supporting high-concurrency access and a dynamic request scheduling component, capable of acquiring multi-dimensional transaction data, including day-ahead market, real-time market, and ancillary service market, from the power trading platform and dispatch center at a frequency of seconds.

[0024] Furthermore, a distributed data monitoring mechanism is used to continuously monitor external data sources, capturing matching results, electricity price trends, regional load changes, and transmission constraint information released by the power trading center. A multi-threaded asynchronous processing mode is employed to concurrently receive large-scale transaction data without blocking the main process. The received raw data is automatically annotated with precise timestamps within the data access unit using a unified timestamp processing mechanism, and undergoes preliminary layered caching based on market region, trading instrument, and data type to ensure data structure consistency and time series alignment in a multi-data source access environment.

[0025] The incoming data is then guided into the data cleaning module, which incorporates outlier removal algorithms, missing value repair mechanisms, and a set of consistency verification rules. Based on a sliding window historical distribution evaluation model, the system performs volatility analysis on key parameters such as price and trading volume. By setting upper and lower limits, it identifies abrupt changes and anomalies, and combines this with market behavior in adjacent time periods to perform predictive repair on missing or anomalous data using a long short-term memory network algorithm backfilling model. Simultaneously, the system utilizes classification label remapping and unit normalization methods to structurally integrate heterogeneous data sources, transforming them into a unified standard market field structure.

[0026] The cleaned structured data is processed by a format standardization engine. Based on a predefined data dictionary and field templates, this engine formats the cleaned results into a unified data structure, covering basic trading elements (such as nodal electricity prices and marginal transaction volumes), scheduling information (such as supply and demand boundary constraints), and ancillary service attributes (such as reserve frequency regulation parameters). All data types are organized in a time-series format and indexed using multi-dimensional indexes based on geographical region, market type, and category tags. This enables rapid retrieval and efficient loading for subsequent models, ultimately generating a power spot market trading dataset with high temporal resolution and data consistency.

[0027] The transaction data prediction module 2 is used to extract transaction feature data from the spot transaction dataset, analyze the changing trend of the transaction feature data, and predict and generate a future market parameter set based on the changing trend. In this embodiment of the application, the transaction data prediction module 2 includes: a data feature extraction module, a data trend analysis module, a future data prediction module, and a result optimization and correction module; The data feature extraction module is used to extract transaction feature data from the spot transaction dataset and to preprocess the transaction feature data. Specifically, the embedded transaction feature extraction module performs multidimensional analysis on historical and real-time transaction data to extract key feature indicators that reflect market operation status, supply and demand structure fluctuations, price change trends, and participant behavior patterns. Then, composed of feature definition sub-units, sliding window analyzers, and multidimensional aggregation operators, the original fields such as nodal electricity prices, marginal transaction volume, regional load, and ancillary service clearing rates are structured and derived to generate a transaction feature dataset with time-series, trend, and behavioral logic explanatory power.

[0028] The market data is continuously segmented based on a sliding time window mechanism, dynamically generating feature sets at different scales, specifically including: Price change characteristics (such as mean, electricity price volatility, price increase / decrease); Transaction behavior characteristics (such as dominant order direction, frequency distribution, and matching success rate); Market trend characteristics (such as three-day moving average, peak-trough reversal points); Load behavior characteristics (such as demand response magnitude and short-term demand elasticity estimation); and bind the characteristics to their corresponding market time period, node region and market type through a tag indexing mechanism to ensure that the characteristics have positioning capabilities and horizontal comparative analysis capabilities in a multi-dimensional space.

[0029] The transaction feature data immediately enters the preprocessing module after it is generated. This module has an embedded feature regularization mechanism, outlier filtering algorithm and distribution fitting standardization process to improve the stability and generalization ability of the feature data in the subsequent modeling process. Based on the numerical distribution of the feature variables, a stratified sampling method is used to perform statistical analysis on various feature values. A density-based outlier detection model is used to remove or reconstruct abrupt data points that may be caused by extreme market disturbances or collection anomalies.

[0030] Furthermore, to adapt to the needs of multi-model training and strategy derivation, the system introduces a feature standardization engine. It uses Z-score, Min-Max scaling, and Box-Cox transformation to unify the scale of continuous variables and uses a categorical feature embedding mapping mechanism to vectorize discrete market attributes. All processed features are exported as data interfaces that can be directly used in market forecasting and strategy generation modules according to a unified naming convention and serialization structure. This forms a transaction feature dataset with high robustness, high temporal consistency, and strong expressive power. Finally, the transaction feature dataset serves as the input layer for auxiliary service generation and multi-objective optimization decision-making models, enhancing the system's ability to understand electricity market behavior patterns and adapt to strategies under complex market conditions. This builds a key intermediate bridge between data-driven and intelligent decision-making.

[0031] The data trend analysis module is used to perform time series analysis on the preprocessed transaction feature data and classify the transaction feature data into market period fluctuations based on the time series analysis results. Specifically, based on the standardized and normalized transaction feature dataset, the system performs dynamic evolution modeling of key feature sequences over continuous periods through a deployed time series analysis engine. This engine consists of an embedded sequence trend recognition unit, a periodicity detection module, and a volatility pattern matching mechanism. It identifies trend items, periodic structures, and abrupt change segments with significant market behavior significance in high-frequency transaction data, thereby revealing the volatility characteristics of the electricity spot market at different time scales.

[0032] The system employs a sliding window dynamic modeling approach to perform local linear fitting and trend decomposition of target trading features. It extracts long-term trend components and short-term oscillation components using multi-layer smoothing filters (such as exponential moving averages and low-pass filter banks). Based on this, it combines first-order difference sequences and autoregressive residual structures to extract the rate of change of trend direction and the range of fluctuation amplitude, providing statistical support for subsequent quantification of fluctuation intensity. Subsequently, a composite frequency domain algorithm based on Fourier transform and wavelet analysis is used to decompose the periodicity in trading features at multiple scales, identifying the frequency response components corresponding to intraday cycles (such as peak-valley alternation), intraweekly seasonality (such as weekend electricity consumption pattern changes), and sudden behaviors (such as the influx of peak-shaving resources). Through cycle stability indices and spectral peak intensity analysis, the sustainability and significance of various fluctuation structures are further marked, and a time-period cycle mapping index is established.

[0033] Based on the above analysis results, a volatility classifier is guided to cluster different market periods. This classifier consists of a density-based time-period clustering model (such as a DBSCAN variant) and a multi-feature fusion discriminator. It comprehensively considers multiple indicators such as price volatility, volume variation coefficient, frequency domain spectral peak distribution, and behavioral mutation rate, and automatically divides the 24-hour trading period into structural categories. The final output includes multiple categories of market behavior labels, such as high volatility periods (such as trading critical periods), low activity periods (such as nighttime price stability intervals), and abnormal disturbance periods (such as sudden backup call times). The class labels are embedded into the original trading feature sequence in a time-aligned format, and are accompanied by classification confidence, volatility weight, and trend annotations, forming a time-series structured feature dataset with behavioral meaning.

[0034] The future data prediction module is used to predict the trend of market fluctuation classification results during market periods using a long short-term memory network algorithm, and to generate a set of future market parameters. In this embodiment of the application, the future data prediction module includes: a data processing module, a future prediction module, an evaluation and optimization module, and a result output module; The data processing module is used to normalize the market period fluctuation classification results and fill in missing data. Specifically, the embedded normalization module performs a unified scale transformation on the volatility intensity of different market periods to ensure the stability and comparability of the data in subsequent model training. The normalization module consists of interval scaling, standardization mapping, and hierarchical adjustment mechanisms. It can unify the dimensions of data for different volatility categories and eliminate scale bias caused by differences in market volatility characteristics. Then, the Min-Max normalization method is used to uniformly map the volatility intensity values, normalizing the volatility intensity of high volatility periods (such as periods of large market price fluctuations) to the [0, 1] interval, while scaling the volatility intensity of low volatility periods (such as low demand periods at night) to ensure that their volatility characteristics are processed within the same scale. Specifically, based on the characteristic value range of different categories (such as high volatility, low volatility, and abnormal disturbances), the minimum and maximum values ​​are calculated independently, and the data points are mapped to the standard interval according to the following formula:

[0035] in, This is the original fluctuation intensity data. The value is the normalized value, ensuring that the normalized fluctuation intensity can be directly compared and comprehensively analyzed across all time periods.

[0036] Subsequently, the normalized volatility classification results are Z-score standardized to further adjust the mean and standard deviation of the data, eliminating the interference of volatility intensity from different market periods on model training. Specifically, this is done by calculating the mean and standard deviation of volatility characteristics for each market period, converting each data point into a standard normal distribution, resulting in a mean of 0 and a standard deviation of 1 for the processed data, thus improving the consistency and stability of data processing. After the normalization step, the system proceeds to the missing data imputation stage. Using a built-in missing value imputation algorithm, the system intelligently imputes missing data, ensuring data integrity and continuity. A preliminary imputation scheme based on linear interpolation is used to infer and supplement the volatility intensity within adjacent time periods. Especially for short-term volatility gaps, linear interpolation effectively maintains a smooth transition in time-series data. For gaps over longer periods, the system uses a time-series prediction model for imputation, utilizing historical volatility data and trend components to predict missing data values, ensuring that the imputed data is consistent with the long-term trend of market volatility characteristics.

[0037] Furthermore, to address the lack of data on sudden fluctuations within specific time periods (such as abrupt changes in market behavior caused by extreme weather or equipment failure), a rule-driven filling mechanism is used to analyze the distribution characteristics of similar historical fluctuation events and supplement the data with external data (such as weather forecasts and power grid load). This ensures that the supplemented data reflects reasonable market behavior patterns, and all fluctuation classification results are uniformly mapped to a standardized range, providing high-quality input data for subsequent market behavior analysis and decision-making models.

[0038] The future prediction module is used to construct a long-short-term prediction model through a long short-term memory network algorithm, and input the normalized market period fluctuation classification results into the long-short-term prediction model to predict market long-short-term data. Specifically, the normalized volatility characteristic data is guided into a time series prediction model based on a long short-term memory network. The data preprocessing module performs time series data windowing processing on the normalized market period volatility data. Using a sliding window mechanism, the continuous market volatility characteristic sequence is divided into multiple time series subsequences according to the set time window (such as 24 hours, 48 ​​hours, etc.). Each subsequence represents the market volatility trajectory within a specific time period. The time series prediction model provides historical data input, and during the input data construction process, it will label the features according to the volatility category of the market period (such as high volatility, low volatility), and vectorize key features such as market volatility intensity, trading volume, and price volatility within each period to form a multi-dimensional data input format with time series characteristics.

[0039] Next, we construct and train a time series prediction model. The time series prediction model consists of an input layer, a time series layer, a fully connected layer, and an output layer. The specific process is as follows: Input layer: Receives windowed market fluctuation data (such as normalized volatility intensity, trading volume, etc.) and passes it to the Long Short-Term Memory (LSTM) network algorithm for time series modeling.

[0040] Time series layer: Through multiple time series units, the input time series data is modeled to capture the short-term change trends and long-term dependency features in the data. Each time series unit has multiple gating structures (input gate, forget gate, output gate), which can automatically adjust the network's memory and forgetting mechanism according to the change of time step, thereby achieving an efficient combination of long-term and short-term information in market fluctuation prediction.

[0041] Fully connected layer: The output of the temporal layer is fused through the fully connected layer to obtain a comprehensive feature representation for subsequent prediction output.

[0042] Output layer: Generates predicted values ​​of market volatility intensity for future periods, and outputs electricity price volatility trends, transaction volume predictions, and market volatility classification results for future time points (next hour, next day).

[0043] During model training, the system uses historical market fluctuation data (including actual electricity price fluctuations, transaction volume, demand fluctuations, etc.) as the training set, and optimizes the loss through a regression loss function (such as mean squared error, MSE).

[0044] After training, the system performs real-time predictions on the input time-series features, uses normalized market fluctuation data as input for the current time step, and outputs the predicted fluctuation values ​​for the corresponding time period. In practical applications, the system feeds these long-term and short-term market prediction results back to the decision-making and dispatching module of the power market, providing more accurate decision support for market participants. Furthermore, the system automatically adjusts market participation strategies during strategy generation, such as adjusting power purchase strategies based on predicted high-fluctuation periods or optimizing power generation resource allocation based on low-fluctuation periods, thereby achieving regulation and control of the power market operation.

[0045] The evaluation and optimization module is used to evaluate the error value of the market long-term and short-term data forecast results, and to adjust and optimize the market long-term and short-term data forecast results based on the error value evaluation results. Specifically, the embedded prediction error assessment module performs time-by-time accuracy analysis on the prediction results. This module consists of an error calculation engine, an anomaly detector, and a dynamic backtracking optimization mechanism. It can perform fine-grained error diagnosis and prediction offset correction on the output of the time-series prediction model, perform time alignment processing on the prediction results and real market observation data, construct an error assessment comparison matrix between the predicted and actual values, and organize the error assessment results with the time dimension as the main axis to form an error time-series spectrum. This is used to reveal the differences in model performance of the system in different prediction periods and under different market conditions. Through error clustering analysis, typical high-error segments are marked, such as the surge in electricity consumption in the early morning, the period of drastic changes in distributed photovoltaic output in the afternoon, and the market disturbance response window, which are the key areas for subsequent optimization. Then, the error-driven prediction optimization mechanism is launched. This mechanism uses the error evaluation results as feedback signals and corrects and enhances the current model output by dynamically adjusting the model structure and prediction strategy.

[0046] The result output module is used to convert the adjusted and optimized market short-term and long-term data prediction results into a future market parameter set output.

[0047] The result optimization and correction module is used to evaluate the data error between the future market parameter set and the historical spot transaction dataset, and to verify and optimize the future market parameter set based on the data error evaluation results and the Bayesian algorithm.

[0048] In this embodiment of the application, the result optimization and correction module includes: an error evaluation module, an error correction module, a correction feedback module, and a verification output module; The error assessment module is used to calculate the error between the future market parameter set and the historical spot transaction dataset, and to assess the accuracy of the parameters of the future market parameter set based on the error calculation results. Specifically, the future market parameter set is generated by the prediction model and includes key parameters such as electricity price, demand load, and power generation clearing. The historical spot transaction dataset provides real market observation data as a reference standard for the prediction results. The prediction accuracy of the parameters is quantitatively evaluated by calculating the error between the two. Then, through time alignment and field mapping mechanisms, the future market parameter set and the historical spot transaction dataset are preprocessed to ensure that each predicted value corresponds precisely to the actual observed value. Next, multiple error evaluation indicators are used to evaluate the prediction results, mainly including mean squared error, mean absolute error, and symmetric mean absolute percentage error, which evaluate the accuracy of the model output from three dimensions: global accuracy, deviation magnitude, and relative error.

[0049] Next, the error-driven accuracy assessment module judges the accuracy of the prediction results of various parameters. Based on the set error threshold, parameters with large prediction errors are marked, high error areas are identified and analyzed in detail. Then, based on the error distribution map and the fluctuation characteristics of historical data, the system identifies which time periods or regions have larger errors, thus providing a quantitative assessment of parameter accuracy. Finally, based on these error assessment results, a parameter accuracy report is automatically generated, and the assessment results are fed back to the prediction model optimization module. By comparing the error distribution with the actual changes in market behavior, the system will adaptively adjust the model.

[0050] The error correction module is used to preset an error correction strategy library, match the error correction strategy library with the parameter accuracy, and use the matched error correction strategy to correct the future market parameter set. Specifically, through the error correction strategy library construction module, a multi-level correction strategy is preset based on the error characteristics of historical market data and the error distribution of prediction results. The strategy library includes, but is not limited to, the following types of correction strategies: Linear regression correction strategy: used to correct persistent biases, applicable to systematic errors in long-term fluctuations of market parameters; Local fluctuation correction strategy: Local correction is performed for errors caused by short-term sudden changes or large fluctuations in order to reduce the impact of instantaneous fluctuations on the prediction results; Weighted moving average correction strategy: used to smooth the forecast results of market parameters and adapt to the correction needs of short-term intraday fluctuations; Machine learning-assisted correction strategy: By introducing deep learning models, complex nonlinear errors are corrected, which is especially suitable for error adjustment in complex market environments.

[0051] After the error calculation stage is completed, an error assessment report for each prediction parameter is automatically generated. Then, an error correction matching algorithm is used to match each error type with the correction strategies in the strategy library. The matching process is based on the magnitude of the error, the change of the time window, and market characteristics (such as high volatility periods, stable periods, etc.) to determine the optimal correction strategy. Then, according to the set rules, the matched correction strategy is automatically selected and applied to adjust each parameter in the future market parameter set. During the correction process, an adaptive adjustment mechanism is used to dynamically optimize the weight and execution method of the correction strategy by combining the historical error correction effect and the feedback of the current market environment. By continuously learning and adjusting the correction strategy, it can adapt to long-term market changes and can also react sensitively to sudden market events in the short term, ensuring the real-time accuracy and efficiency of the market parameter set.

[0052] The correction feedback module is used to compare the corrected future market parameter set with the historical spot transaction dataset, and to evaluate the correction result based on the comparison result. Specifically, through a time series alignment mechanism, the corrected set of future market parameters is matched with the historical spot transaction dataset on a time-by-time basis. Field mapping rules ensure that core parameters such as electricity price, load, and transaction volume have a strict correspondence across the same time dimension. Subsequently, the system calls upon a multi-dimensional error index set to calculate the errors in the prediction results before and after correction, and compares the magnitude of error reduction to quantify the correction effect. After the error calculation is completed, the system enters the correction result evaluation process.

[0053] The evaluation module makes a comprehensive judgment based on the following dimensions: Error improvement rate assessment: Calculate the percentage decrease in error after correction compared to before correction, and measure the degree of improvement of the correction strategy under different parameters and different time periods; Robustness analysis: By detecting volatility sensitivity, we determine whether the corrected prediction results have higher stability and anti-misalignment ability in high volatility ranges; Consistency test: Using statistical correlation and trend fitting methods, analyze the degree of fit between the corrected parameters and historical data in terms of overall trend; Robustness evaluation: By comparing the correction effects in different market scenarios (such as high-load days and low-demand nights), the cross-scenario adaptability of the correction strategy is evaluated.

[0054] Finally, the comparison and evaluation results are integrated into a correction result evaluation report. The report includes various error improvement indicators, trend alignment maps, and stability analysis curves. This report not only provides quantitative evidence for the effectiveness of the correction strategy, but also serves as input feedback to the strategy optimization module to update the error correction strategy library and adjust the weight parameters of future prediction models.

[0055] The verification output module is used to add annotations to the future market parameter set based on the correction results and then output the results.

[0056] The transaction assistance service module 3 is used to preset the assistance service emphasis rules and combine the assistance service emphasis rules with the spot transaction dataset and the future market parameter set to generate a transaction assistance service scheme set; In this embodiment of the application, the transaction assistance service module 3 includes: a rule setting module, a scheme generation module, a scheme evaluation module, and a verification output module; The rule setting module is used to set multi-faceted emphasis rules for auxiliary services according to market operation requirements; Specifically, the system monitors the current market conditions (such as load levels, power generation output, and renewable energy fluctuations) in real time, and combines historical data with forecasts to identify the types of ancillary services that need to be prioritized in the market. These ancillary services include, but are not limited to, frequency regulation services, standby dispatch services, and load regulation services. The system sets different service emphasis rules based on real-time demand, capacity, and service response speed. Next, through the multi-party emphasis rule setting module, specific rules are set according to the following key market operation requirements: Load tracking and stability requirements: When large-scale load fluctuations occur in the market, the system will prioritize the activation of load regulation services and reserve capacity to ensure market frequency stability and power supply reliability. At this time, the emphasis on load regulation services will be increased to ensure that the dispatch system can flexibly respond to instantaneous fluctuations.

[0057] Renewable energy generation fluctuations: For renewable energy generation (such as wind and solar power) affected by weather fluctuations, the system will increase the emphasis on frequency regulation services during predictable fluctuation periods, and use fast-response dispatching methods to reduce system load fluctuations and improve the system's adaptability to renewable energy fluctuations.

[0058] Market demand and supply balance: During periods of tight market supply and demand (such as peak load periods), the system will prioritize the allocation of auxiliary services to ensure that load demand is met, based on the degree of supply tightness. At this time, the focus of backup scheduling services will increase to ensure sufficient backup capacity to cope with emergencies.

[0059] System operation efficiency optimization: When the power grid load is relatively stable and controllable, the system will reduce its reliance on low-priority ancillary services, such as peak shaving and valley filling load regulation services, thereby improving the overall system operation efficiency and optimizing resource allocation.

[0060] Finally, through a dynamic adjustment and feedback mechanism, the system continuously adjusts the emphasis rules of ancillary services based on real-time changes and scheduling results during market operation. The system also monitors the response speed and effectiveness of various ancillary services in real time and fine-tunes the rules online based on actual operational results. This adaptive adjustment ensures that various ancillary services can be flexibly applied in different market scenarios and consistently meet the stability and economic requirements of the electricity market.

[0061] The solution generation module is used to combine the spot transaction dataset and the future market parameter set, and generate an auxiliary service solution set based on multi-faceted emphasis rules. Specifically, the data fusion module is invoked to align and integrate historical spot trading datasets and future market parameter sets in both time and parameter dimensions. Historical trading data provides a quantitative description of typical market behaviors, including load response, price fluctuations, and generation clearing records. Future market parameters contain a series of predicted values ​​for upcoming operating conditions (such as predicted electricity prices, volatility, and renewable energy output). The fused composite dataset forms a market operating status data map. Based on this map, key scenarios with ancillary service demands in the current and future periods are identified, such as: The risk of frequency stability increases during periods of high volatility. There is a gap in backup scheduling in areas with insufficient clearing capacity; Drastic changes in renewable energy output have led to an increase in the demand for regulating resources.

[0062] After the ancillary service requirements are identified, the multi-party emphasis rule-driven engine is activated. The module prioritizes and matches the service requirements according to the previously set emphasis rules (such as stability priority, economic priority, response speed priority, etc.).

[0063] The weighting of the emphasis rule is based on the following factors: Time-related characteristics: Short-response services will have increased weight during peak load periods; Resource characteristics: Fast response resource priority matching and frequency adjustment services; Regional characteristics: Reserve regulation services will be prioritized for areas with high local grid pressure; Economic cost: Strategies are weighted based on marginal service costs to optimize overall market economics.

[0064] Through the rule matching process, the system establishes a mapping table between demand type and service solution, providing input for the next step of service combination.

[0065] After completing rule matching, the system calls the auxiliary service solution generator to construct a complete auxiliary service combination based on the matched service strategy. The generation process adopts a heuristic combination algorithm and constraint filtering mechanism to ensure that the solution meets the following requirements: service response time constraints, resource capacity boundaries, market economic requirements, and maximization of service type synergy. Each generated service solution includes service type, target area, response window, resource call list, and estimated cost, forming a complete auxiliary service description structure. The final output auxiliary service solution set will have highly structured and schedulable characteristics, and can be directly connected to the scheduling system for dynamic invocation or published to market participants through an interface to realize service bidding and response.

[0066] The solution evaluation module is used to set solution evaluation indicators and perform feasibility evaluation on the ancillary service solution set based on the solution evaluation indicators. In this embodiment of the application, the scheme evaluation module includes: a performance evaluation module, an optimization suggestion module, a risk evaluation module, and a feasibility evaluation module; The performance evaluation module is used to set the scheme evaluation indicators and to perform quantitative performance evaluation of the auxiliary service scheme set based on the scheme evaluation indicators. Specifically, based on the key requirements of market operation and the functional characteristics of auxiliary services, a comprehensive evaluation index system is defined to quantify the performance of auxiliary service solutions. The evaluation index mainly covers the following dimensions: Response speed: This measures the time from when a service request is issued to when a response is actually received, reflecting the timeliness of the solution. In services with high timeliness requirements, such as frequency regulation and load balancing, response speed will be given higher weight.

[0067] Resource utilization rate: This assesses the actual use of resources (such as reserve capacity and regulated power generation) in the evaluation plan, reflecting the efficiency of resource utilization. The level of resource utilization directly affects the cost-effectiveness and supply stability of the market.

[0068] Cost-effectiveness: The economics of a solution are measured by calculating the total cost required to perform the service against the benefits (such as reduced electricity price volatility and improved system stability). A cost-effective solution can achieve the expected performance goals without increasing costs excessively.

[0069] System stability improvement: This metric assesses the degree to which the implemented solution improves grid stability and quantifies the solution's response to frequency fluctuations and load changes. This indicator is crucial for frequency regulation and load tracking services.

[0070] Implementation Risk: Assess the potential risks of the plan during actual implementation, including factors such as insufficient resources and response delays, reflecting the stability and reliability of the plan's execution. Plans with higher risks may face greater uncertainty in actual operation.

[0071] Service coverage: This measures the solution's ability to cover different regions and time periods within the market, ensuring effective service delivery under varying loads and high volatility conditions. Solutions with high service coverage ensure reliable grid operation in dynamic environments.

[0072] After the evaluation metrics are set, each ancillary service solution is quantitatively scored using a performance evaluation engine. The evaluation engine operates based on the following process: Individual scoring and weighting: Each evaluation metric is assigned a specific numerical value through a quantitative method (e.g., response time in milliseconds, resource utilization rate as a percentage), and then different weights are assigned to each metric based on its importance. For example, the response speed and system stability improvement capabilities of frequency adjustment services will be given higher weights, while cost-effectiveness will be adjusted according to the urgency of market demand.

[0073] Overall Performance Scoring: The system synthesizes the scores of various indicators using a weighted average method or a weighted comprehensive index method to generate an overall performance score for each solution. This score comprehensively reflects the solution's overall performance in terms of response speed, resource utilization, cost-effectiveness, and other aspects.

[0074] Risk Adjustment Score: For solutions that show higher risk in the assessment, the system will adjust the score using a risk adjustment coefficient. This coefficient is adjusted based on the implementation risks and potential problems of the solution, reducing the final score of the higher-risk solution.

[0075] Scenario Adaptability Assessment: The system assesses the adaptability of the solution to different market operating conditions (such as peak load, low load, and large fluctuations in renewable energy). The stronger the adaptability of the solution to different market scenarios, the higher the score.

[0076] After completing the performance quantification evaluation, the system uses a dynamic feedback mechanism to send the evaluation results back to the solution generation and adjustment module. The specific process is as follows: Solution optimization suggestions: Based on the evaluation score, the system will generate optimization suggestions for each solution. If a solution performs poorly in terms of cost-effectiveness or resource utilization, the system will suggest adjusting resource allocation or priority order to improve the overall effectiveness of the solution.

[0077] Solution Iteration and Optimization: After obtaining market feedback, the system will adjust the evaluation model and weights online to optimize subsequently generated solutions. This process involves calibration using historical evaluation results and real-time data to ensure that the solutions always adapt to market changes.

[0078] Real-time tracking and feedback adjustment: The system will also track the actual performance of the plan in real time. If the actual performance is lower than expected, the system will automatically trigger adjustments to the plan and provide feedback for correction.

[0079] The optimization suggestion module is used to set an optimization suggestion threshold, match the optimization suggestion threshold with the performance quantification evaluation results, and generate an optimization suggestion scheme based on the matching results. Specifically, the threshold setting module defines recommended optimization thresholds for various performance quantification evaluation indicators. These thresholds are determined based on market operation needs, system security requirements, and historical evaluation experience, including: Response speed threshold: Sets the upper limit for the response time of auxiliary services. For example, if the response delay exceeds a certain millisecond or second level, optimization suggestions will be triggered.

[0080] Resource utilization threshold: Sets a minimum resource utilization standard. If the resource utilization is too low, it will prompt you to optimize resource configuration to avoid redundancy.

[0081] Cost-effectiveness threshold: Sets the minimum ratio of unit benefit to execution cost. If the solution is below this threshold, a cost structure optimization suggestion is triggered.

[0082] Stability Improvement Threshold: Defines the minimum requirement for system frequency stability or fluctuation reduction rate. If the threshold is not met, measures to enhance stability are suggested.

[0083] Implement risk threshold: Set the maximum tolerable risk value. When the risk score exceeds the threshold, the system generates strategy suggestions to reduce the risk.

[0084] Scene adaptability threshold: Defines the minimum adaptability score of the solution in multiple scenes. If the solution performs poorly in a specific scene, a multi-scene optimization strategy is triggered.

[0085] By establishing a multi-dimensional optimization trigger standard system based on the above thresholds, a clear criterion is provided for subsequent solution improvements.

[0086] After the performance quantification evaluation is completed, the system calls the result matching engine to compare the evaluation index results of each solution with the corresponding thresholds item by item: Single index matching: Checking whether the score of each index is lower or higher than the threshold, forming a matching matrix of qualified / unqualified. Comprehensive deviation analysis: Counting the number and weight of unqualified indicators, calculating the overall deviation of the solution, reflecting the overall optimization space of the solution. Priority determination: Determining the priority of optimization suggestions based on the degree of deviation and index weight. For example, if both response speed and stability indicators are lower than the threshold, the solution is marked as high-priority optimization.

[0087] After threshold matching is completed, the system enters the optimization suggestion generator stage. Based on the matching results, specific optimization suggestions are generated, including: Targeted optimization suggestion generation: For indicators below the threshold, the system provides targeted suggestions. For example: insufficient response speed suggests introducing rapid adjustment resources or shortening the scheduling path; insufficient cost-effectiveness suggests optimizing resource combinations or introducing low-cost backup capacity; inadequate stability suggests increasing the priority of frequency adjustment or expanding the stability resource pool. Combined optimization strategy: When multiple indicators fail to meet the standard simultaneously, the system generates a comprehensive solution through combined optimization algorithms, ensuring a balance among multiple indicators and avoiding new negative effects from single-item optimization. Hierarchical output of optimization suggestions: The generated optimization suggestions are categorized by priority, forming a three-level structure of urgent optimization items, important optimization items, and general optimization items. Feedback and iteration mechanism: After the optimization suggestion is implemented, the system continuously tracks its effect and feeds back the actual execution results to the threshold setting module to dynamically adjust the threshold and improve the accuracy of future suggestions.

[0088] The risk assessment module is used to perform quantitative analysis on the ancillary service solution set and assess market volatility risk. Specifically, the system conducts detailed quantitative analysis on each ancillary service solution in the set, evaluating its effectiveness and adaptability based on multiple performance indicators. Each solution is analyzed using a performance quantification model, including the following key dimensions: Response Latency Analysis: The system quantifies the response speed of each solution to assess its ability to activate promptly during periods of drastic market demand fluctuations. This process generates accurate response latency assessment results by collecting and processing data such as the solution's trigger time and execution time. Resource Utilization: The system quantifies resource utilization efficiency based on the ratio between the actual resources called upon by the solution and the available resources. High resource utilization indicates that the solution can more efficiently schedule market resources, reducing system operating costs. Service Cost-Benefit Assessment: The system calculates the scheduling cost of each solution and its resulting market benefits (such as electricity price fluctuation suppression and system stability improvement), quantifying the solution's economic viability through the cost-benefit ratio. Stability Gain Analysis: By modeling the suppression effects of frequency fluctuations and load changes, the system quantifies each solution's contribution to grid stability and assesses its performance under abnormal fluctuations. Risk Control Capability Assessment: The system evaluates the execution reliability of each solution in high-risk market scenarios. The system calculates the resilience of each scheme to abnormal fluctuations based on historical execution data. Market adaptability analysis: The system assesses the adaptability of each scheme under different market scenarios to ensure its continued effective operation in a volatile electricity market.

[0089] After completing the quantitative analysis of the solution set, a market risk assessment will be conducted using the market volatility risk assessment module, combined with real-time market data. The main steps are as follows: Fluctuation Scenario Simulation: Based on historical data and predictive models, the system constructs various typical market fluctuation scenarios, such as high load fluctuations, drastic changes in renewable energy output, and sharp price increases and decreases. Each fluctuation scenario corresponds to one or more potential risks.

[0090] Risk Exposure Analysis: By simulating the implementation of the plan under market volatility scenarios, the system can assess the risk exposure of the plan under specific market volatility conditions. The higher the exposure, the more difficult the plan is to implement under market volatility, and the greater the potential for system instability.

[0091] Disturbance immunity score: The system quantifies the disturbance immunity of each scheme by matching the intensity of fluctuations with the response capability of the scheme. This score reflects the scheme's ability to maintain grid stability under market fluctuations.

[0092] Redundancy and Backup Assessment: This assesses whether the proposed solution possesses sufficient resource redundancy and backup mechanisms in high-volatility scenarios. A highly redundant solution can mitigate the risks associated with market volatility through the allocation of additional resources.

[0093] Market adaptability of the solution: The system evaluates the adaptability of each solution by taking into account different market conditions and disturbance characteristics, so as to ensure that the solution can respond normally and operate effectively in various market fluctuations.

[0094] By comprehensively evaluating performance indicators and market volatility risks, a comprehensive risk score and optimization recommendations can be generated for each scheme. These evaluation results provide a scientific basis for power market operation, dispatch, and ancillary service decisions. The specific process is as follows: Overall Score and Prioritization: The system calculates an overall score based on the performance evaluation results and risk exposure of the proposed solutions. Solutions with high scores typically exhibit stronger responsiveness, lower implementation costs, and greater market adaptability, and are therefore recommended for priority use.

[0095] Optimization suggestion generation: Based on the market volatility risk assessment results, the system generates targeted optimization suggestions for each solution. If a solution has a high risk of exposure under specific market volatility, the system will suggest adjusting resource allocation, increasing redundancy, or activating backup solutions.

[0096] Optimize the feedback mechanism: During the implementation of the plan, the system will track the operation effect of the plan in real time, adjust the evaluation model and thresholds according to the execution results, optimize subsequent decisions, and continuously improve risk assessment and resource scheduling strategies.

[0097] The feasibility assessment module is used to combine the optimized proposed solution with market volatility risk to conduct a feasibility assessment, and to verify and output the feasibility assessment results.

[0098] Specifically, the optimized solutions will be integrated with the market volatility risk assessment results to construct a complete feasibility assessment mechanism. First, the optimization measures will be mapped to key risk indicators, and their adaptability in a highly volatile environment will be tested through dynamic market scenario simulation. Then, a feasibility assessment engine will be used to quantitatively score the optimization solutions from multiple dimensions, such as performance improvement, risk adaptability, cost change rate, and system compatibility. A weighted model will be used to calculate the comprehensive feasibility score. Finally, the assessment results will be cross-validated through historical comparison verification, parameter backtesting verification, and real-time tracking verification to ensure that the assessment model has accuracy and predictive ability, thereby providing a scientific, stable, and controllable decision-making basis for the implementation of the auxiliary service optimization solutions.

[0099] The verification output module is used to verify the feasibility assessment results and select auxiliary service schemes from the auxiliary service scheme set based on the verified feasibility assessment results.

[0100] The transaction service decision module 4 is used to verify and optimize the set of transaction auxiliary service solutions, and select the target transaction auxiliary service solution based on the verification and optimization results. The execution feedback correction module 5 is used to execute the target transaction auxiliary service scheme, record transaction execution data, and dynamically correct the auxiliary service emphasis rules based on the feedback analysis results of the transaction execution data.

[0101] In this embodiment of the application, the execution feedback correction module 5 includes: a real-time monitoring module, a feedback analysis module, a correction strategy module, and a correction execution module; The real-time monitoring module is used to monitor market data and various monitoring parameters during the execution process in real time, and output them as feedback information. The feedback analysis module is used to analyze the deviation value of the feedback information through an anomaly detection algorithm; Specifically, the feedback information is processed using anomaly detection algorithms and compared with predetermined normal behavior patterns to identify abnormal deviations in the data. The specific process is as follows: Feature extraction and modeling: By analyzing historical feedback data, normal behavior patterns are constructed. This pattern is based on multi-dimensional features, such as time series changes, resource utilization, and execution results, to form a statistical model of the dataset.

[0102] Benchmark model establishment: Based on historical data and current operating status, the system constructs a dynamic benchmark model that represents the normal behavior range under ideal conditions, and the model can capture the regular fluctuation patterns and trends of feedback data.

[0103] Anomaly detection algorithm application: Common anomaly detection algorithms, such as Isolation Forest, Support Vector Machine, and k-means clustering, are used to compare the feedback data with the benchmark model to identify outliers that deviate from the normal range. The output of the anomaly detection result is an anomaly score, which reflects the degree of deviation between the feedback data and the normal pattern.

[0104] After anomaly detection is completed, the system uses the deviation value calculation module to quantitatively analyze the detected anomaly feedback information. The core process of deviation value calculation is as follows: Deviation value definition: The system defines deviation value as a measure of the difference between the feedback data and the baseline model. Commonly used calculation methods include Euclidean distance, Manhattan distance, and cosine similarity. The larger the deviation value, the more significant the deviation between the feedback information and the normal pattern, which may represent a potential problem or anomaly.

[0105] Deviation threshold setting: Based on historical data and business needs, the system sets a reasonable deviation threshold. If the deviation value exceeds the threshold, the system will mark the feedback as abnormal and activate the corresponding response mechanism.

[0106] Anomaly Classification: Based on the magnitude of the deviation, the system further categorizes anomalies into different types, such as minor anomalies, moderate anomalies, and severe anomalies. Each anomaly category corresponds to a different response strategy to ensure that problems are handled promptly.

[0107] Once a deviation is detected and its value is calculated, the system will initiate a feedback adjustment mechanism to adjust the decision-making strategy based on the severity of the anomaly. The specific process is as follows: Minor anomalies: For minor anomalies, the system will trigger an alert mechanism to remind operations personnel to monitor system operation, keep an eye on the situation, and prevent the problem from escalating further.

[0108] Moderate anomaly: When the abnormal deviation is significant, the system will automatically adjust some parameters and optimize the configuration or scheduling strategy to restore the stability of system operation.

[0109] Critical Anomaly: For critical anomalies, the system will automatically take measures to isolate the fault or roll back to prevent further impact, and conduct a comprehensive investigation to ensure that the problem is completely resolved.

[0110] The correction strategy module is used to preset a deviation value threshold strategy library, match the deviation value threshold strategy library with the deviation value analysis results, and select a deviation value threshold strategy based on the matching results. In this embodiment of the application, the correction strategy module includes: a data receiving module, a strategy selection module, a strategy simulation module, and an adjustment output module; The data receiving module is used to receive the error calculation results from the error evaluation module; The strategy selection module is used to preset a strategy selection library and select an initial correction strategy from the preset strategy selection library based on error assessment results and changes in market parameters. The strategy simulation module is used to simulate and test the selected initial correction strategy using a data recalibration algorithm, and to evaluate the applicability of the initial correction strategy under the current market data. In this embodiment of the application, the strategy simulation module includes: a data relabeling module, an applicability evaluation module, and a simulation result output module; The data recalibration module is used to recalibrate the future market parameter set and integrate the recalibrated future market parameter set with a scale standard. Specifically, the future market parameter set is recalibrated using a recalibration module to ensure that the forecast results accurately reflect dynamic market changes. The specific steps are as follows: Historical data backtesting: The system analyzes historical market data to identify the fluctuation patterns and influencing factors of different market parameters (such as electricity prices, demand, and renewable energy output) over specific time periods, and establishes a benchmark model. This historical data provides a reference for the recalibration of future market parameters.

[0111] Future Prediction Model Construction: Combining historical data and market trends, the system employs time series analysis, regression analysis, and machine learning methods to construct a predictive model for future market parameters. This model can predict the possible range of changes in various market parameters over future time periods based on current market conditions.

[0112] Recalibration rule setting: Based on the dynamic characteristics of the forecasting model and market parameters, the system sets recalibration rules. These rules include: adjusting forecasting parameter values ​​according to actual market fluctuations; making compensatory corrections for specific anomalies; and correcting forecast results using real-time market feedback.

[0113] Recalibration process: Within each time period, the system recalibrates the parameters based on the deviation between the predicted and actual values. This process can be performed in real-time or periodically to ensure the accuracy and timeliness of future market parameters.

[0114] After completing the recalibration process, the system uses the scaling standard integration module to standardize the recalibrated future market parameters, ensuring comparability between the parameters and facilitating subsequent analysis and decision-making. The specific steps are as follows: Scale conversion model construction: The system constructs a scale conversion model based on the dimensions and magnitudes of different market parameters. For example, electricity prices are usually measured in yuan per kilowatt-hour, while loads may be measured in megawatts. The system needs to convert market parameters in different units into a unified standard format, such as normalizing all parameters to the range of 0-1, or converting them according to a relative standardization method.

[0115] Standardization Method Selection: The system employs common standardization methods, such as min-max standardization and Z-score standardization, to transform the recalibrated set of future market parameters into a unified scale. This process maps parameters from different dimensions to a standard scale for further comprehensive analysis and optimization.

[0116] Parameter weighting: During the standardization process, the system assigns weights to each parameter based on its impact on the decision outcome. For example, electricity price changes have a significant impact on decisions, while load fluctuations may have a smaller impact. By setting weights, the system ensures that the final standardization result accurately reflects the relative importance of different market parameters.

[0117] Scale-integrated output: Through scale standard integration, the system unifies all market parameters into comparable standardized values ​​and outputs a set of future market parameters in a unified format. This integrated data can be used for further decision support and strategy optimization.

[0118] After completing recalibration and scaling standard integration, the system outputs the processed future market parameter set through a standardized output mechanism for subsequent decision-making and optimization strategies. The specific output format is as follows: Unified data format: The system converts all market parameters into a unified data format to ensure that the output results can be seamlessly integrated with other decision-making modules.

[0119] Dynamic feedback mechanism: The system continuously recalibrates and scales the future market parameter set based on real-time market feedback to adapt to dynamic market changes and ensure that the parameter set always remains consistent with the current market situation.

[0120] The applicability assessment module is used to modify the initial correction strategy by incorporating the future market parameter set after the scaling standard is integrated, and to calculate the applicability and applicable time period of the strategy correction. In this embodiment of the application, the applicability assessment module includes: an error analysis module, an applicability calculation module, and a time period determination module; The error analysis module is used to apply the initial correction strategy to the future market parameter set after the scale standard is integrated, and to calculate the prediction gap between the predicted value after strategy correction and the expected target value. The applicability calculation module is used to evaluate the performance of the correction strategy based on the prediction gap value, generate an applicability score, and make an applicability judgment based on the applicability score. Specifically, based on the analysis of the prediction gap, the performance of the correction strategy is quantitatively evaluated, which can accurately reflect the effectiveness of each strategy in reducing prediction errors. First, the difference between the predicted and actual values ​​before and after strategy correction is calculated, gap characteristics are extracted, and an error analysis model is constructed by combining indicators such as trend deviation and volatility deviation. Then, by setting evaluation dimensions (such as gap convergence rate, error reduction ratio, stability factor, and cost change rate), the strategy performance is transformed into a comprehensive scoring model, thereby generating a standardized suitability score. This score integrates different dimensional indicators based on a weighting function, reflecting the relative advantages and risk control capabilities of the correction strategy in a specific market scenario. Finally, based on the score and set thresholds, suitability is judged and categorized into highly suitable, conditionally suitable, and unsuitable categories. The decision support module prioritizes strategies with better performance and stronger adaptability and outputs feedback suggestions, forming a data-driven strategy selection closed loop.

[0121] The time period determination module is used to determine the applicable range of the initial correction strategy in the future time series through the sliding window backtesting algorithm, and generate the corresponding applicable time period information.

[0122] Specifically, a sliding window backtesting algorithm is introduced to dynamically test the applicability of the initial correction strategy in future time series, identify its performance trends across different time periods, and generate corresponding applicable time period information accordingly. This process first constructs a sliding window of fixed length, dividing the entire prediction time series into multiple continuous sub-intervals. The system runs the correction strategy within each window and records its ability to suppress prediction errors and its stability within that time period, forming a window-level performance score sequence. Then, based on the performance score and a set applicability threshold, the window results are filtered, identifying window regions with good strategy performance and effective bias control, and marking them as applicable intervals. If multiple consecutive windows meet the applicability criteria, the system merges them into a longer applicable time period and outputs its start and end timestamps as the final applicability labeling information. Finally, all effective segments are combined and output as a standardized dataset of applicable time periods for the strategy, providing a structured reference for subsequent dynamic strategy switching and local deployment.

[0123] The simulation result output module is used to evaluate the applicability of the initial correction strategy under the current market data based on the applicability of the strategy correction and the applicable time period, and then output the results.

[0124] Specifically, the system first retrieves the generated applicability score, which is based on the strategy's performance in historical and simulated data, reflecting its comprehensive capabilities in error control, cost-effectiveness, and stability. Then, it calls upon the applicable time period data generated by sliding window backtesting, matching the current market data's time point with each applicable interval. If the current timestamp falls within a strategy's applicable time period and the strategy's applicability score is higher than a set threshold, the system determines that the strategy is applicable under the current market conditions. If the time matches but the score is low, it is deemed conditionally applicable. If neither the time nor the score matches, it is marked as inapplicable. Finally, a structured applicability result is output, including the strategy's adaptation status at the current time point, score details, matching interval number, and suggested operation prompts, serving as the basis for dynamic strategy selection and switching.

[0125] The adjustment output module is used to adjust and optimize the initial correction strategy based on the simulation test results, form a deviation threshold correction strategy for controlling the prediction deviation within a preset range, and output it as the final correction result.

[0126] Specifically, the initial correction strategy is first tested multiple times through simulation to evaluate its performance under different market scenarios, particularly its ability to control forecast bias. By comparing the differences between predicted and actual values, the deviation characteristics of the initial correction strategy in different time periods are identified, and the error fluctuation range of each stage is recorded. Based on this, a preliminary deviation analysis report is generated, reflecting the deviation control effectiveness of the initial correction strategy.

[0127] Next, based on the simulation results, the correction strategy is adjusted and optimized. A deviation threshold is calculated based on the magnitude of the deviation in each simulation period, and parameters in the correction strategy are optimized according to a preset deviation control range (such as adjusting the weights of the prediction model, correcting resource scheduling strategies, or adjusting response latency). The goal of the optimization process is to limit the prediction deviation within a preset error range, while improving the adaptability and stability of the correction strategy under different market environments.

[0128] During the optimization process, reverse testing is used to verify the performance of the corrected strategy over future time periods, and further adjustments are made to the optimization parameters. Reverse testing not only ensures that the deviation remains within a preset range but also ensures that the corrected strategy has sufficient robustness in the real market to cope with different market fluctuations. Finally, the adjusted and verified deviation threshold correction strategy will be output as the final correction result. A detailed report on the correction strategy will be generated, including the optimization objectives, adjusted parameters, error control range, and optimized deviation threshold for each stage.

[0129] The correction execution module is used to perform correction operations according to the selected deviation value threshold strategy and record correction parameters in real time.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A decision-making system for auxiliary services in electricity spot trading, characterized in that, The power spot trading auxiliary service decision system includes: a transaction data acquisition module, a transaction data prediction module, a transaction auxiliary service module, a transaction service decision module, and an execution feedback correction module; The transaction data acquisition module is used to acquire electricity spot transaction data in real time and form a spot transaction dataset through data cleaning and integration. The transaction data prediction module is used to extract transaction feature data from the spot transaction dataset, analyze the changing trend of the transaction feature data, and predict and generate a future market parameter set based on the changing trend. The transaction assistance service module is used to preset assistance service emphasis rules and combine the assistance service emphasis rules with the spot transaction dataset and the future market parameter set to generate a transaction assistance service scheme set; The transaction service decision module is used to verify and optimize the set of transaction auxiliary service solutions, and select the target transaction auxiliary service solution based on the verification and optimization results. The execution feedback correction module is used to execute the target transaction auxiliary service scheme, record transaction execution data, and dynamically correct the auxiliary service emphasis rules based on the feedback analysis results of the transaction execution data.

2. The power spot trading auxiliary service decision system according to claim 1, characterized in that, The transaction data prediction module includes: a data feature extraction module, a data trend analysis module, a future data prediction module, and a result optimization and correction module. The data feature extraction module is used to extract transaction feature data from the spot transaction dataset and to preprocess the transaction feature data. The data trend analysis module is used to perform time series analysis on the preprocessed transaction feature data and classify the transaction feature data into market period fluctuations based on the time series analysis results. The future data prediction module is used to predict the trend of market fluctuation classification results during market periods using a long short-term memory network algorithm, and to generate a set of future market parameters. The result optimization and correction module is used to evaluate the data error between the future market parameter set and the historical spot transaction dataset, and to verify and optimize the future market parameter set based on the data error evaluation results and the Bayesian algorithm.

3. The power spot trading auxiliary service decision system according to claim 1, characterized in that, The transaction support service module includes: a rule setting module, a solution generation module, a solution evaluation module, and a verification output module; The rule setting module is used to set multi-faceted emphasis rules for auxiliary services according to market operation requirements; The solution generation module is used to combine the spot transaction dataset with the future market parameter set and generate an auxiliary service solution set based on multi-faceted emphasis rules. The solution evaluation module is used to set solution evaluation indicators and perform feasibility evaluation on the ancillary service solution set based on the solution evaluation indicators. The verification output module is used to verify the feasibility assessment results and select auxiliary service schemes from the auxiliary service scheme set based on the verified feasibility assessment results.

4. The power spot trading auxiliary service decision system according to claim 1, characterized in that, The execution feedback correction module includes: a real-time monitoring module, a feedback analysis module, a correction strategy module, and a correction execution module; The real-time monitoring module is used to monitor market data and various monitoring parameters during the execution process in real time, and output them as feedback information. The feedback analysis module is used to analyze the deviation value of the feedback information through an anomaly detection algorithm; The correction strategy module is used to preset a deviation value threshold strategy library, match the deviation value threshold strategy library with the deviation value analysis results, and select a deviation value threshold strategy based on the matching results. The correction execution module is used to perform correction operations according to the selected deviation value threshold strategy and record correction parameters in real time.

5. The power spot trading auxiliary service decision system according to claim 2, characterized in that, The future data prediction module includes: a data processing module, a future prediction module, an evaluation and optimization module, and a result output module; The data processing module is used to normalize the market period fluctuation classification results and fill in missing data. The future prediction module is used to construct a long-short-term prediction model through a long short-term memory network algorithm, and input the normalized market period fluctuation classification results into the long-short-term prediction model to predict market long-short-term data. The evaluation and optimization module is used to evaluate the error value of the market long-term and short-term data forecast results, and to adjust and optimize the market long-term and short-term data forecast results based on the error value evaluation results. The result output module is used to convert the adjusted and optimized market short-term and long-term data prediction results into a future market parameter set output.

6. The power spot trading auxiliary service decision system according to claim 2, characterized in that, The result optimization and correction module includes: an error evaluation module, an error correction module, a correction feedback module, and a verification output module; The error assessment module is used to calculate the error between the future market parameter set and the historical spot transaction dataset, and to assess the accuracy of the parameters of the future market parameter set based on the error calculation results. The error correction module is used to preset an error correction strategy library, match the error correction strategy library with the parameter accuracy, and use the matched error correction strategy to correct the future market parameter set. The correction feedback module is used to compare the corrected future market parameter set with the historical spot transaction dataset, and to evaluate the correction result based on the comparison result. The verification output module is used to add annotations to the future market parameter set based on the correction results and then output the results.

7. The power spot trading auxiliary service decision system according to claim 3, characterized in that, The solution evaluation module includes: a performance evaluation module, an optimization suggestion module, a risk evaluation module, and a feasibility evaluation module; The performance evaluation module is used to set the scheme evaluation indicators and to perform quantitative performance evaluation of the auxiliary service scheme set based on the scheme evaluation indicators. The optimization suggestion module is used to set an optimization suggestion threshold, match the optimization suggestion threshold with the performance quantification evaluation results, and generate an optimization suggestion scheme based on the matching results. The risk assessment module is used to perform quantitative analysis on the ancillary service solution set and assess market volatility risk. The feasibility assessment module is used to combine the optimized proposed solution with market volatility risk to conduct a feasibility assessment, and to verify and output the feasibility assessment results.

8. The power spot trading auxiliary service decision system according to claim 4, characterized in that, The correction strategy module includes: a data receiving module, a strategy selection module, a strategy simulation module, and an adjustment output module; The data receiving module is used to receive the error calculation results from the error evaluation module; The strategy selection module is used to preset a strategy selection library and select an initial correction strategy from the preset strategy selection library based on error assessment results and changes in market parameters. The strategy simulation module is used to simulate and test the selected initial correction strategy using a data recalibration algorithm, and to evaluate the applicability of the initial correction strategy under the current market data. The adjustment output module is used to adjust and optimize the initial correction strategy based on the simulation test results, form a deviation threshold correction strategy for controlling the prediction deviation within a preset range, and output it as the final correction result.

9. The power spot trading auxiliary service decision system according to claim 8, characterized in that, The strategy simulation module includes: a data recalibration module, an applicability evaluation module, and a simulation result output module; The data recalibration module is used to recalibrate the future market parameter set and integrate the recalibrated future market parameter set with a scale standard. The applicability assessment module is used to modify the initial correction strategy by incorporating the future market parameter set after the scaling standard is integrated, and to calculate the applicability and applicable time period of the strategy correction. The simulation result output module is used to evaluate the applicability of the initial correction strategy under the current market data based on the applicability of the strategy correction and the applicable time period, and then output the results.

10. The power spot trading auxiliary service decision system according to claim 9, characterized in that, The applicability assessment module includes: an error analysis module, an applicability calculation module, and a time period determination module; The error analysis module is used to apply the initial correction strategy to the future market parameter set after the scale standard is integrated, and to calculate the prediction gap between the predicted value after strategy correction and the expected target value. The applicability calculation module is used to evaluate the performance of the correction strategy based on the prediction gap value, generate an applicability score, and make an applicability judgment based on the applicability score. The time period determination module is used to determine the applicable range of the initial correction strategy in the future time series through the sliding window backtesting algorithm, and generate the corresponding applicable time period information.

Images