Intelligent matching method for long-term transaction in power market

Abstract

The application discloses a kind of long-term transaction intelligent matching method in electric power market, and the specific steps of the method are as follows: collecting multi-source data and standardizing preprocessing;Extract the operation fingerprint of power generation side and power consumption side, and construct the implicit characteristic image of transaction subject;Calculate the matching degree of physical characteristics, and screen to form a physically feasible matching set;With the total remaining maximum of both parties as the goal to solve the initial matching scheme;After decomposing the transaction electric quantity, input digital twin simulation, identify safe risk matching pair;According to priority adjustment and iterative verification, output the final scheme.The application extracts the operation fingerprint of industrial equipment to perform physical pre-matching, reducing contract default;At the same time, the matching scheme is virtually verified throughout the cycle using digital twin, identifying power grid safety risks in advance and avoiding losses caused by after-the-fact remedies.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of electricity market trading and industrial data mining, specifically to a method for intelligent matching of medium- and long-term transactions in the electricity market. Background Technology

[0002] The construction of a unified national electricity market system continues to advance, with medium- and long-term transactions occupying a core position in the electricity market, playing a crucial role in locking in electricity volume, mitigating price risks, and ensuring power supply. As the number of market participants continues to increase and the scale of transactions expands, traditional manual matching methods can no longer meet market demands. Data-driven intelligent matching technology is gradually being applied in electricity trading centers at all levels, effectively improving the overall efficiency of transaction matching.

[0003] Existing intelligent matching methods generally only consider the explicit characteristics of the market and ignore the physical attributes of electricity trading. This can easily lead to matching results that are optimal in the market but cannot be executed in practice, increasing the default risk of trading entities. At the same time, there is a lack of effective pre-verification mechanisms. Problems with the matching scheme can only be discovered and remedied during the execution process, which often results in unnecessary losses. In addition, existing methods make low use of industrial equipment operation data and cannot fully reflect the actual performance capabilities of trading entities. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent matching method for medium- and long-term transactions in the electricity market. This method extracts the operational fingerprints of industrial equipment on the generation and consumption sides to construct implicit feature profiles of trading entities. It first completes physical characteristic pre-matching, then solves for the optimal market matching scheme within the feasible region, and finally completes full-cycle virtual verification and iterative optimization through a digital twin of the power system. This method changes the traditional matching logic that relies solely on market features such as price and electricity volume, filtering out physically infeasible matching pairs from the root. It also shifts from post-event verification to pre-event verification, reducing contract defaults and grid operation fluctuations, adapting to medium- and long-term trading scenarios at different time scales, and making the matching results more closely reflect the actual operating status of the trading entities.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a smart matching method for medium- and long-term transactions in the electricity market, the specific steps of which are as follows: S1: Collect multi-source data from generating units, industrial equipment, power grid operation, and market transactions; perform standardized preprocessing to eliminate data noise and dimensional differences, and ensure the accuracy of subsequent feature extraction and matching calculations. S2: Extract the operational fingerprints of the power generation and consumption sides through industrial data mining, reduce the dimensionality of the original features to obtain the core feature vector, construct the implicit feature profile of the trading entities, and explore the implicit operational characteristics of the trading entities to make up for the deficiencies of explicit market features such as price and electricity. S3: The weighted cosine similarity is used to calculate the physical characteristic matching degree between the trading entities. The matching threshold is adjusted according to market supply and demand, the proportion of new energy and grid congestion. A set of physically feasible matches is selected and physically infeasible matching pairs are filtered out in advance to narrow the solution range of subsequent optimization. S4: Within the physically feasible matching set, construct an optimization model with the goal of maximizing the total surplus of both parties in the transaction. Set constraints on electricity, power grid transmission, and price, and solve to obtain the initial matching scheme. Under the premise of physical feasibility, achieve the optimal market transaction efficiency. S5: Decompose the transaction volume of the initial matching scheme into a preset time granularity, input it into the power system digital twin to simulate the full cycle of the contract, monitor preset security indicators and identify security risk matching pairs, move the security verification process forward, identify power grid operation risks in advance, and avoid post-event remediation. S6: Adjust the safety risk matching pairs according to the preset priority, iterate and verify until all indicators meet the safety threshold, output the final transaction matching scheme, and complete the transaction matching under the premise of ensuring the safe operation of the power grid.

[0006] Furthermore, in step S1, the standardization preprocessing includes: using the 3σ criterion combined with temporal continuity detection to identify outliers, which can simultaneously identify numerical deviations and logical anomalies that do not conform to the operating rules; replacing outliers with the average of three consecutive normal data before and after them; processing missing values ​​according to the missing rate using linear interpolation, weighted averaging of similar devices, or elimination methods; performing min-max normalization and unified time granularity alignment on the data to eliminate differences in the units and time granularity of different data and constructing a unified time series dataset.

[0007] Furthermore, in step S2, the power generation side operation fingerprint is divided into four groups: basic operation characteristics, regulation characteristics, economic characteristics, and reliability characteristics. Each group contains eight quantitative characteristics directly related to the operating status of the power generation equipment. Among them, the basic operation characteristics include rated capacity, minimum technical output, maximum technical output, average output, output fluctuation rate, annual utilization hours, number of start-ups and shutdowns, and average start-up and shutdown time. The regulation characteristics include peak shaving depth, peak shaving response speed, ramp rate, AGC regulation accuracy, primary frequency regulation response time, secondary frequency regulation response time, minimum stable fuel load, and start-up and shutdown costs; Economic characteristics include unit coal consumption curve, unit power generation cost, environmental protection cost, operation and maintenance cost, carbon emission intensity per kilowatt-hour, fuel price sensitivity, maintenance cycle, and equipment health. Reliability characteristics include mean time between failures (MTBF), downtime rate, historical performance rate, output forecast accuracy, peak shaving task completion rate, environmental compliance rate, number of safety incidents, and credit rating. The power consumption side operation fingerprint is divided into four groups: basic load characteristics, time sequence characteristics, production process characteristics, and power consumption behavior characteristics. Each group contains eight quantitative characteristics that are directly related to the operating status of the power equipment. Among them, the basic load characteristics include annual power consumption, maximum load, minimum load, average load, load factor, peak-valley difference rate, simultaneity rate, and load fluctuation rate. The time-series characteristics include daily load curve type, weekly load curve type, monthly load curve type, seasonal load factor, peak period percentage, valley period percentage, normal period percentage, and load transfer capacity; Production process characteristics include production continuity level, equipment start-up and shutdown costs, power consumption per product, production cycle, capacity utilization rate, order fluctuation coefficient, interruptible load capacity, and interruptible duration; Electricity consumption behavior characteristics include historical contract fulfillment rate, price sensitivity, demand response participation rate, demand response speed, proportion of self-generated and self-consumed renewable energy, energy storage capacity, electricity consumption growth rate, and credit rating.

[0008] Furthermore, step S2 also includes a dynamic fingerprint update mechanism: collecting the latest operational data at a preset cycle to update the implicit feature profile of the transaction entity; when major unit overhauls, production process adjustments, equipment replacements, environmental protection facility upgrades, and other similar events affecting the operational fingerprint occur, an update is triggered within a preset time after the event ends to ensure the timeliness of the feature profile and avoid the matching results from being out of sync with the actual operational status of the transaction entity.

[0009] Furthermore, in step S3, the formula for calculating the weighted cosine similarity is:

[0010] in, Main generator With the main body of electricity consumption The degree of matching of physical characteristics To sum over all core feature dimensions, The total dimension of the core feature vector. For the first The weight coefficients of the core features, Main generator The Standardized values ​​of core features For the main electricity user The The algorithm standardizes the core feature values; it can distinguish the differences in importance of different feature dimensions, and more accurately reflect the degree of physical matching between trading entities; the first... Weight coefficients of core features The initial weights are determined through the following steps: First, multiple experts in the fields of electricity market and grid operation compare and score the importance of each feature pairwise to construct a judgment matrix using the analytic hierarchy process (AHP) to obtain the initial weights. Then, using historical transaction contract data, the initial weights are optimized using the gradient descent method to maximize the Pearson correlation coefficient between the matching degree and the contract fulfillment rate. Combining expert experience with historical transaction data, the weight values ​​are made to better reflect the actual transaction scenario.

[0011] Furthermore, in step S3, the matching threshold adjustment rule is as follows: set an initial threshold; reduce the first adjustment amount when the market supply-demand ratio is greater than the first threshold; increase the second adjustment amount when the proportion of new energy output is greater than the second threshold; increase the third adjustment amount when the historical grid congestion probability is greater than the third threshold; the adjustment amounts can be superimposed, and the adjusted threshold fluctuates within a preset upper and lower limit range, which can adapt to different market supply and demand, new energy proportions and grid operating environments, and improve the adaptability of the matching scheme.

[0012] Furthermore, in step S4, the optimization model is a mixed-integer linear programming model, which can simultaneously handle 0-1 matched discrete variables and continuous variables of traded electricity volume, ensuring the rationality of the solution results. The objective function is:

[0013] in, To maximize computation, To sum over all power generation entities, To sum over all electricity users, This represents the total number of power generation entities. This represents the total number of electricity users. The variable is a 0-1 match, with a value of 1 indicating the power generation entity. With the main body of electricity consumption A deal is reached; a value of 0 indicates that a deal has not been reached. For the main electricity user The declared electricity purchase price, Main generator The declared electricity sales price per unit, Main generator With the main body of electricity consumption The method involves calculating the trading volume between different entities. An improved branch-and-bound method is used to solve the problem. Priority is given to matching pairs where the trading volume is greater than a preset volume threshold and the price difference is greater than a preset price difference threshold. This effectively reduces the number of branches, improves the solution efficiency, and shortens the time required for matching calculations.

[0014] Furthermore, in step S5, the power system digital twin consists of a power grid physical model, a meteorological model, an operation control model, and a fault simulation model. The power grid physical model simulates the electrical characteristics of various power grid equipment, the meteorological model simulates the impact of meteorological conditions on the output and load of new energy sources, the operation control model simulates the control strategy of power grid dispatching, and the fault simulation model simulates the occurrence and propagation process of power grid faults, which can comprehensively cover various scenarios of power grid operation. The preset safety indicators include line load rate, node voltage deviation, system frequency deviation, transformer load rate, and power grid congestion probability.

[0015] Furthermore, step S5 also includes a digital twin incremental learning mechanism: after each contract is executed, the actual power grid operation data is collected and compared with the digital twin simulation data of the corresponding time period. The parameters of the power grid physical model and the meteorological model are updated using an incremental learning algorithm. The simulation accuracy can be improved without retraining the complete model, so that the verification results gradually approach the actual operation of the power grid.

[0016] Furthermore, in step S6, the adjustment priority of the security risk matching pair is executed in the following order: First priority, all transaction electricity of the matching pair is transferred to a matching pair in the same area that does not involve risky lines and has a higher physical characteristic matching degree; if the first priority cannot transfer all transaction electricity, the second priority is executed, and the remaining untransferred transaction electricity is transferred to a matching pair in the adjacent area with a higher physical characteristic matching degree; if the second priority still cannot transfer all remaining electricity, the third priority is executed, and the transaction electricity of the matching pair is reduced proportionally until the security threshold requirement is met. The adjustment is completed first through electricity transfer to minimize the impact of the adjustment on both parties to the transaction and ensure the stability of the transaction.

[0017] Compared with existing technologies, this integrated management system for the hot pot industry, from ingredient traceability to end-user services, has the following beneficial effects: I. This invention generates a latent feature profile of the trading entity by extracting the operational fingerprints of industrial equipment on the power generation and consumption sides. It first completes the pre-matching of physical characteristics and then performs the market-optimal matching calculation. This changes the traditional method's matching logic that only relies on market features such as price and electricity volume. It filters out physically infeasible matching pairs from the source and reduces defaults during contract execution. At the same time, by running a dynamic fingerprint update mechanism, it ensures the timeliness of the feature profile and allows the matching results to stably adapt to the operational changes of the trading entity, reducing the frequency of subsequent contract adjustments.

[0018] Second, this invention uses a digital twin of the power system to perform full-cycle virtual verification of the initial matching scheme, identify power grid operation safety risks in advance, and then iteratively adjust the matching scheme according to a preset priority. This closely integrates market transactions with the physical operation of the power grid, avoiding economic losses and power grid operation fluctuations caused by ex-post verification in traditional methods. At the same time, through the incremental learning mechanism of the digital twin, the simulation accuracy is continuously improved, and the power grid adaptability and execution success rate of the matching scheme are gradually improved.

[0019] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0021] Figure 1 This is a flowchart illustrating the core steps of a smart matching method for medium- and long-term transactions in the power market, as described in this invention. Figure 2 This is a flowchart illustrating the filtering of the physically feasible matching set in this invention; Figure 3 This is a flowchart illustrating the safety risk identification and iterative adjustment process of this invention. Detailed Implementation

[0022] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0023] Example 1

[0024] The complete implementation process of the intelligent matching method for medium- and long-term transactions in the power market of this invention involves collecting multi-source data from the generation side, the consumption side, the power grid, and market transactions, performing standardized preprocessing, extracting the operational fingerprints of industrial equipment of both parties in the transaction, and constructing implicit feature profiles. First, physical characteristic pre-matching is completed based on weighted cosine similarity. Then, an initial matching scheme that maximizes the total surplus of both parties in the physical feasible domain is solved. Finally, the matching scheme is iteratively optimized through a digital twin of the power system for full-cycle virtual verification. This achieves the unification of market transaction demand and power grid physical operation constraints.

[0025] like Figure 1As shown, the specific process of this embodiment includes six core steps from data acquisition to final solution output, and the specific implementation process is as follows: First, multi-source data acquisition is conducted. This involves collecting operating data from the industrial IoT platform (generator side) and industrial equipment (consumer side) via standard communication protocols, as well as grid operation data from the real-time database of the power grid dispatch center and market transaction data from the power trading center's trading system. The collected data covers the equipment operating status of trading entities, the grid operating status, and historical trading behavior. All data is transmitted and stored according to a unified interface specification, providing a foundation for subsequent feature extraction and matching calculations.

[0026] After the collection of multi-source data is completed, the standardized preprocessing stage begins.

[0027] The first step is outlier handling. The 3σ criterion is used to identify numerical outliers. This criterion is based on the principle of normal distribution. First, the mean and standard deviation of each time-series data sequence are calculated. Then, data deviating from the mean by more than three standard deviations are marked as numerical outliers. Simultaneously, time-series continuity detection is used to identify logical outliers. By calculating the magnitude of change in data between adjacent time points, when the magnitude of change exceeds the maximum range for normal operation of this type of equipment, it is marked as a logical outlier. All identified outliers are replaced with the average of the three consecutive normal data points before and after them.

[0028] The second step is to handle missing values. Different processing methods are used according to the missing data rate. For time periods with a missing rate of less than 5%, linear interpolation is used to fill the missing values ​​by calculating the missing values ​​through the linear relationship between two adjacent normal data. For time periods with a missing rate between 5% and 20%, the weighted average of data from the same type of equipment in the same time period is used to fill the missing values. The weights are determined according to the similarity of the equipment's operation. Time periods with a missing rate of more than 20% are marked as invalid data and removed.

[0029] The third step is to perform min-max normalization, which maps all numerical data to the interval between 0 and 1 through linear transformation, eliminating the influence of different units on subsequent calculations.

[0030] The fourth step is to perform unified time granularity alignment, aligning all collected data with different time granularities to a 1-hour time granularity. Data with a time granularity greater than 1 hour is split, and data with a time granularity less than 1 hour is aggregated to build a unified time series database.

[0031] After data preprocessing, the operation fingerprints of the power generation and power consumption sides are extracted from the preprocessed time series data using industrial data mining techniques.

[0032] The power generation side operation fingerprint is divided into four groups: basic operation characteristics, regulation characteristics, economic characteristics, and reliability characteristics. Each group contains eight quantitative characteristics that are directly related to the operating status of the power generation equipment.

[0033] Basic operating characteristics are extracted from the equipment's basic parameters and long-term operating statistics, including rated capacity, minimum technical output, maximum technical output, average output, output fluctuation rate, annual utilization hours, number of start-ups and shutdowns, and average start-up and shutdown time. Output fluctuation rate is obtained by calculating the ratio of the standard deviation of output per unit time to the average output. Regulation characteristics are extracted from the equipment's regulation process data, including peak shaving depth, peak shaving response speed, ramp-up rate, AGC regulation accuracy, primary frequency regulation response time, secondary frequency regulation response time, minimum stable fuel load, and start-up and shutdown costs. Peak shaving response speed is obtained by calculating the change in unit output per unit time relative to the rated capacity. The ratio is obtained; economic characteristics are extracted from the economic operation data of the equipment, including unit coal consumption curve, unit power generation cost, environmental protection cost, operation and maintenance cost, carbon emission intensity per kilowatt-hour, fuel price sensitivity, maintenance cycle, and equipment health. The unit coal consumption curve is obtained by fitting coal consumption data under different output levels; reliability characteristics are extracted from the historical operation and transaction data of the equipment, including mean time between failures (MTBF), failure downtime rate, historical performance rate, output prediction accuracy, peak shaving task completion rate, environmental compliance rate, number of safety accidents, and credit rating. The mean time between failures (MTBF) is obtained by statistically analyzing the average running time between two equipment failures.

[0034] The power consumption side operation fingerprint is divided into four groups: basic load characteristics, time sequence characteristics, production process characteristics, and power consumption behavior characteristics. Each group contains eight quantitative characteristics that are directly related to the operating status of the power equipment.

[0035] Basic load characteristics are extracted from statistical data on electricity consumption, including annual electricity consumption, maximum load, minimum load, average load, load factor, peak-valley difference rate, simultaneity rate, and load volatility. Temporal characteristics are extracted from temporal load variation data, including daily load curve type, weekly load curve type, monthly load curve type, seasonal load factor, peak period percentage, valley period percentage, normal period percentage, and load transfer capacity. Production process characteristics are extracted from user production process data, including production continuity level, equipment start-up and shutdown costs, single-product power consumption, production cycle, capacity utilization rate, order volatility factor, interruptible load capacity, and interruptible duration. Electricity consumption behavior characteristics are extracted from users' historical electricity consumption and transaction data, including historical fulfillment rate, price sensitivity, demand response participation rate, demand response speed, proportion of self-generated and self-consumed renewable energy, energy storage capacity, electricity consumption growth rate, and credit rating.

[0036] After extracting the original running fingerprint, principal component analysis was used to reduce the dimensionality of the original features.

[0037] First, the covariance matrix of the original feature matrix is ​​calculated. Then, the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvalues ​​are arranged in descending order, and the top n eigenvectors with a cumulative contribution rate of 90% are selected as principal components. The original feature matrix is ​​projected onto these n eigenvectors to obtain the core feature vectors. Principal component analysis transforms a set of potentially correlated original features into a set of linearly uncorrelated core feature vectors through orthogonal transformation. This reduces the feature dimensionality while retaining the main information of the original features, thus reducing the complexity of subsequent calculations. After obtaining the core feature vectors, a latent feature profile is constructed for each trading entity. Each trading entity's profile corresponds to an n-dimensional core feature vector. Simultaneously, a dynamic update mechanism for operational fingerprints is established. The latest operational data is collected monthly, and the operational fingerprints are re-extracted and the latent feature profiles of trading entities are updated to ensure that the feature profiles reflect the actual operational status of the trading entities in real time. When major unit overhauls, production process adjustments, equipment replacements, environmental facility upgrades, or other similar events affecting the operational fingerprints occur, fingerprint updates are triggered within 72 hours after the event ends to promptly correct any deviations in the feature profiles.

[0038] After completing the construction of the implicit feature profile of the transaction entity, the process moves to the physical characteristic pre-matching stage.

[0039] The weighted cosine similarity algorithm is used to calculate the physical characteristic matching degree between power generation entities and power consumption entities. This algorithm comprehensively considers the differences in importance across different feature dimensions, more accurately reflecting the degree of physical characteristic matching between the trading entities. The calculation formula is as follows:

[0040] in, Main generator With the main body of electricity consumption The degree of matching of physical characteristics To sum over all core feature dimensions, The total dimension of the core feature vector. For the first The weight coefficients of the core features, Main generator The Standardized values ​​of core features For the main electricity user The Standardized values ​​of core features; No. Weight coefficients of core features Determined through the following steps: The first step involves having multiple experts in the fields of electricity market and power grid operation conduct pairwise comparisons and scores of the importance of each feature, constructing a judgment matrix using the analytic hierarchy process (AHP). The elements in the judgment matrix represent the relative importance between two features. Then, the largest eigenvalue and the corresponding eigenvector of the judgment matrix are calculated. After normalizing the eigenvector, the initial weights are obtained. At the same time, the consistency index and consistency ratio are calculated to check the consistency of the judgment matrix. When the consistency ratio is less than 0.1, the judgment matrix is ​​considered to have satisfactory consistency, and the initial weights are considered valid. The second step utilizes historical transaction contract data and employs gradient descent to optimize the initial weights. First, the loss function is defined as the negative of the Pearson correlation coefficient between the matching degree and the contract fulfillment rate. Then, the weight values ​​are iteratively adjusted to minimize the loss function, which in turn maximizes the Pearson correlation coefficient between the matching degree and the contract fulfillment rate. During the iteration process, the learning rate is set to 0.01, and the iteration terminates when the change in the loss function is less than 0.0001 or the maximum number of iterations (1000) is reached, at which point the iteration stops, yielding the final weight coefficients.

[0041] After determining the matching degree calculation method, the matching threshold is adjusted according to market supply and demand, the proportion of new energy sources, and grid congestion.

[0042] First, an initial threshold is set to 0.6. When the market supply-demand ratio exceeds the first threshold, the first adjustment amount is decreased; when the proportion of renewable energy output exceeds the second threshold, the second adjustment amount is increased; and when the historical grid congestion probability exceeds the third threshold, the third adjustment amount is increased. These adjustments can be cumulative, and the adjusted threshold fluctuates within the range of 0.4 to 0.8. Finally, matching pairs with a matching degree greater than the adjusted threshold are selected to form a physically feasible matching set. The selection process is as follows: Figure 2 As shown, physically infeasible matching pairs are filtered out at the source.

[0043] After obtaining the set of physically feasible matches, the process moves on to the market-optimal matching optimization stage.

[0044] Within the physically feasible matching set, a mixed-integer linear programming model is constructed with the objective of maximizing the total surplus of both trading parties. Mixed-integer linear programming models can handle both continuous and integer variables simultaneously, making them suitable for optimization problems involving 0-1 decision variables, such as electricity trading matching. The objective function is:

[0045] in, To maximize computation, To sum over all power generation entities, To sum over all electricity users, This represents the total number of power generation entities. This represents the total number of electricity users. The variable is a 0-1 match, with a value of 1 indicating the power generation entity. With the main body of electricity consumption A deal is reached; a value of 0 indicates that a deal has not been reached. For the main electricity user The declared electricity purchase price, Main generator The declared electricity sales price per unit, Main generator With the main body of electricity consumption Electricity traded between them.

[0046] The model sets four types of constraints: power generation constraints stipulate that the total transaction volume of each power generation entity shall not exceed its declared maximum saleable volume; power consumption constraints stipulate that the total transaction volume of each power consumption entity shall not exceed its declared maximum demand; power grid transmission capacity constraints stipulate that the total transaction volume on each transmission line shall not exceed its maximum transmission capacity; and transaction price constraints stipulate that the declared prices on both the power generation and power consumption sides shall be within the upper and lower limits of the price stipulated by the trading center.

[0047] An improved branch and bound method is used to solve this mixed-integer linear programming model. The improved branch and bound method first sorts all matching pairs in descending order according to the product of the transaction volume and the price difference, and prioritizes branching the matching pairs whose transaction volume is greater than a preset volume threshold and whose price difference is greater than a preset price difference threshold, which can effectively reduce the number of branches.

[0048] During the solution process, the upper and lower bounds of each branch are quickly calculated using the relaxation method. When the upper bound of a branch is less than the currently known optimal lower bound, the branch is pruned and the solution is no longer pursued. At the same time, heuristic rules are introduced to generate initial feasible solutions, thereby improving the solution speed and finally obtaining the initial matching scheme.

[0049] After obtaining the initial matching scheme, the process proceeds to the digital twin virtual verification stage.

[0050] The transaction volume in the initial matching scheme is decomposed into hourly time granularities to obtain the power generation and consumption plans for the entire contract execution cycle, which are then input into a pre-constructed digital twin of the power system. The digital twin of the power system consists of a power grid physical model, a meteorological model, an operation control model, and a fault simulation model. The power grid physical model includes mathematical models of all power grid equipment, such as generators, transformers, transmission lines, and loads. The model parameters are derived from the factory parameters and historical operating data of the power grid equipment, accurately simulating the electrical characteristics of the equipment. The meteorological model, based on historical meteorological data and weather forecast data, simulates the impact of different meteorological conditions on renewable energy output and load. The operation control model simulates the automatic generation control and automatic voltage control strategies of the power grid dispatch, accurately simulating the power grid operation control process. The fault simulation model simulates common power grid faults and their propagation processes, capable of simulating various fault scenarios such as N-1 faults.

[0051] The digital twin simulates the entire lifecycle of the power grid operation through contract execution. Within each hourly time granularity, it calculates the power flow distribution based on generation and consumption plans, and monitors five preset safety indicators in real time: line load rate, node voltage deviation, system frequency deviation, transformer load rate, and grid congestion probability. When any indicator exceeds its corresponding safety threshold, the transaction pair that caused the exceedance is recorded and identified as a safety risk pair. Simultaneously, an incremental learning mechanism is established for the digital twin. After each contract execution, actual power grid operation data is collected and compared with the corresponding time period's digital twin simulation data. The deviation between the simulated and actual data is calculated, and an incremental learning algorithm is used to update the parameters of the power grid physical model and meteorological model, continuously improving the simulation accuracy of the digital twin. The incremental learning algorithm can update model parameters using new operational data without retraining the entire model, ensuring that the model can adapt to changes in the power grid's operating state.

[0052] After identifying security risk matching pairs, such as Figure 3 As shown, it is adjusted according to the preset priority.

[0053] The first priority is to transfer all traded electricity from the matched pair to another matched pair within the same region that does not involve risky lines and has a high degree of physical characteristic matching, prioritizing regional balance in the trading. If the first priority cannot transfer all traded electricity, the second priority is executed, transferring the remaining untransferred traded electricity to a matched pair with a high degree of physical characteristic matching in an adjacent region. If the second priority still cannot transfer all remaining electricity, the third priority is executed, proportionally reducing the traded electricity of the matched pair until the safety threshold requirements are met. After each adjustment, a new matching scheme is generated and input into the power system digital twin for verification. This verification and adjustment process is repeated until all monitoring indicators meet the safety threshold requirements, at which point the final trading matching scheme is output. The maximum number of iterations during the iteration process is set to 50. If the maximum number of iterations still cannot meet all safety threshold requirements, the total remaining target value for both parties in the transaction is appropriately reduced, prioritizing the safety of power grid operation.

[0054] This embodiment fully describes the implementation process of the intelligent matching method for medium- and long-term electricity market transactions proposed in this invention. By combining physical characteristic pre-matching with market-optimal matching, it changes the traditional method's matching logic that relies solely on market characteristics. Furthermore, it achieves pre-emptive security verification of the matching scheme through digital twin virtual verification. This method can effectively filter physically infeasible matching pairs, reduce contract defaults, and identify grid operation safety risks in advance, avoiding losses from post-event remediation.

[0055] This method is applicable to medium- and long-term electricity market transactions at different time scales, such as annual, quarterly, and monthly transactions, and is also applicable to regional electricity markets of different sizes.

[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent matching of medium- and long-term transactions in the electricity market, characterized in that, The specific steps of this method are as follows: S1: Collect multi-source data from generating units, industrial equipment on the power consumption side, power grid operation, and market transactions, and perform standardized preprocessing; S2: Extract the operational fingerprints of the power generation and consumption sides through industrial data mining, reduce the dimensionality of the original features to obtain the core feature vector, and construct the implicit feature profile of the transaction entity; S3: Use weighted cosine similarity to calculate the physical characteristic matching degree between trading entities, adjust the matching threshold according to market supply and demand, the proportion of new energy and grid congestion, and screen to form a set of physically feasible matches; S4: Within the set of physically feasible matches, construct an optimization model with the goal of maximizing the total surplus of both parties in the transaction. Set constraints on electricity consumption, power grid transmission, and price, and solve for the initial matching scheme. S5: Decompose the transaction volume of the initial matching scheme into a preset time granularity, input it into the power system digital twin to simulate the full cycle of the contract, monitor preset security indicators and identify security risk matching pairs; S6: Adjust the security risk matching pairs according to the preset priority, iterate and verify until all indicators meet the security threshold, and output the final transaction matching scheme.

2. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S1, the standardization preprocessing includes: identifying outliers using the 3σ criterion combined with temporal continuity detection, replacing outliers with the average of the three consecutive normal data before and after them; processing outliers using linear interpolation, weighted averaging of similar devices, or elimination methods according to the missing value rate; and performing min-max normalization and uniform time granularity alignment on the data.

3. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S2, the power generation side operation fingerprint is divided into four groups: basic operation characteristics, regulation characteristics, economic characteristics, and reliability characteristics. Each group contains eight quantitative characteristics directly related to the operating status of the power generation equipment. The power consumption side operation fingerprint is divided into four groups: basic load characteristics, time-series characteristics, production process characteristics, and power consumption behavior characteristics. Each group contains eight quantitative characteristics directly related to the operating status of the power consumption equipment.

4. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, Step S2 also includes a dynamic fingerprint update mechanism: collecting the latest operational data at a preset cycle to update the implicit feature profile of the transaction entity; when major unit overhaul, production process adjustment, equipment replacement, environmental protection facility upgrade, or other similar events affecting the operational fingerprint occur, an update is triggered within a preset time after the event ends.

5. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S3, the formula for calculating the weighted cosine similarity is: in, Main generator With the main body of electricity consumption The degree of matching of physical characteristics To sum over all core feature dimensions, The total dimension of the core feature vector. For the first The weight coefficients of the core features, Main generator The Standardized values ​​of core features For the main electricity user The Standardized numerical values ​​of core features; Weight coefficients of core features The initial weights are determined through the following steps: First, multiple experts in the fields of electricity market and grid operation compare and score the importance of each feature pairwise to construct a judgment matrix using the analytic hierarchy process (AHP) to obtain the initial weights; then, using historical transaction contract data, the initial weights are optimized using the gradient descent method to maximize the Pearson correlation coefficient between the matching degree and the contract fulfillment rate.

6. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S3, the matching threshold adjustment rule is as follows: set an initial threshold; reduce the first adjustment amount when the market supply-demand ratio is greater than the first threshold; increase the second adjustment amount when the proportion of new energy output is greater than the second threshold; increase the third adjustment amount when the historical grid congestion probability is greater than the third threshold; the adjustment amounts can be superimposed, and the adjusted threshold fluctuates within the preset upper and lower limits.

7. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S4, the optimization model is a mixed-integer linear programming model, and the objective function is: in, To maximize computation, To sum over all power generation entities, To sum over all electricity users, This represents the total number of power generation entities. This represents the total number of electricity users. The variable is a 0-1 match, with a value of 1 indicating the power generation entity. With the main body of electricity consumption A deal is reached; a value of 0 indicates that a deal has not been reached. For the main electricity user The declared electricity purchase price, Main generator The declared electricity sales price per unit, Main generator With the main body of electricity consumption The transaction volume between them is determined by an improved branch and bound method, prioritizing matching pairs where the transaction volume is greater than a preset volume threshold and the price difference is greater than a preset price difference threshold.

8. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S5, the power system digital twin consists of a power grid physical model, a meteorological model, an operation control model, and a fault simulation model; the preset safety indicators include line load rate, node voltage deviation, system frequency deviation, transformer load rate, and power grid congestion probability.

9. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, Step S5 also includes a digital twin incremental learning mechanism: after each contract is executed, the actual power grid operation data is collected and compared with the digital twin simulation data of the corresponding time period, and the parameters of the power grid physical model and the meteorological model are updated using an incremental learning algorithm.

10. The intelligent matching method for medium- and long-term transactions in the electricity market according to claim 1, characterized in that, In step S6, the adjustment priority of the security risk matching pair is executed in the following order: First priority, all transaction volume of the matching pair is transferred to the matching pair in the same area that does not involve risk lines and has a higher physical characteristic matching degree; if the first priority cannot transfer all transaction volume, the second priority is executed, and the remaining untransferred transaction volume is transferred to the matching pair in the adjacent area with a higher physical characteristic matching degree; if the second priority still cannot transfer all remaining volume, the third priority is executed, and the transaction volume of the matching pair is reduced proportionally until the security threshold requirement is met.

Images