Multi-microgrid operation method and system based on electricity price sensitivity weighted scene reduction

Abstract

The invention discloses a multi-microgrid operation method and system based on electricity price sensitivity weighted scene reduction, and the method comprises the steps: constructing a master-slave game market framework which comprises an upper-layer leader model and a lower-layer follower model; obtaining a typical scene by using an electricity price sensitivity weighted scene reduction method; the upper leader model takes minimization of the expected total comprehensive operation cost of the whole regional power grid as a target function; each micro-grid in the lower-layer follower model receives the internal transaction electricity price given by the upper-layer leader model, and transaction volume data is obtained with the purpose of minimizing the comprehensive operation cost of the micro-grid; in combination with the transaction volume data, the upper leader model calculates the fitness and feedback quality index of the current internal transaction electricity price based on a feedback quality adaptive mechanism, and generates a new generation of internal transaction electricity price according to the feedback quality index; and repeating the iteration process, and outputting an optimal internal transaction electricity price strategy. According to the method, the uncertainty of renewable energy sources and loads can be dealt with more accurately, and optimal configuration of regional system resources is realized.

Description

Technical Field

[0001] This invention belongs to the field of power system optimization operation and market-based trading technology, and specifically relates to a multi-microgrid operation method and system based on electricity price sensitivity weighted scenario reduction. Background Technology

[0002] In recent years, the penetration rate of distributed renewable energy, represented by wind and solar power, on the distribution network side has continued to rise, forming a new pattern of regional energy systems with multiple microgrids coexisting. Achieving coordinated operation and energy sharing among multiple microgrids is of great significance for improving renewable energy consumption, ensuring regional power supply reliability, and reducing system operating costs.

[0003] While master-slave game theory provides an ideal framework for constructing regional energy markets, it still has significant limitations in practical applications, particularly in handling uncertainty, data privacy, and optimizing collaboration. (1) Existing scenario reduction techniques are disconnected from optimization objectives: When considering the uncertainty of renewable energy and load, traditional scenario reduction methods mainly focus on the differences in probability distribution between scenarios, but fail to link the reduction process with subsequent optimization objectives such as electricity price games. For example, the patent "A Distributed Optimization Scheduling Method and System for Distribution Network under Multi-Stakeholder Game" (CN202411085913.8) introduces stochastic programming theory to handle the uncertainty of renewable energy output and load. However, this method does not clearly explain how its scenario reduction process is combined with subsequent game optimization objectives. If scenario reduction is based solely on probability distance, scenarios that have a key impact on game equilibrium may be lost, making subsequent optimization based on a lack of representative scenario sets.

[0004] (2) The difficulty of optimization under incomplete information and the contradiction of data privacy: In the master-slave game framework, the upper-level leader (such as the market operator) usually does not read the detailed internal operating data of the lower-level followers (such as the microgrid) in order to protect the data privacy of the followers. This makes traditional optimization algorithms face the challenge of incomplete information. The patent "Energy Co-optimization Method under Multi-regional Microgrid and Electric Vehicle Load Interaction" (CN120955761A) effectively protects the data privacy of each microgrid through the federated learning mechanism. However, this method does not clearly explain how it solves the problem of low optimization efficiency caused by insufficient information in the upper-level model under the premise of protecting data privacy. How to achieve efficient optimization under limited information is still an unsolved problem.

[0005] (3) Balancing Distributed Optimization and Global Coordination: Multi-microgrid systems involve multiple stakeholders, and distributed optimization algorithms (such as ADMM) are often used to solve game theory models. However, when dealing with uncertain scenarios, how to ensure the balance between the convergence speed of distributed solutions and the global optimization effect still needs further research. The patent "Optimal and Flexible Energy Trading Method and System for Microgrids Based on Two-Level Game Theory" (CN201710053467.6) establishes a multi-master, multi-slave Stankberg game model to improve microgrid revenue and energy utilization. However, in uncertain environments, the coordination mechanism between the distributed solution efficiency and the global optimization performance of this method still has room for improvement.

[0006] (4) The complexity of coordinated operation of multiple microgrid systems: With the increase in the number of microgrids and the complexity of network structure (such as the emergence of AC / DC hybrid microgrid groups), how to achieve efficient coordinated operation and optimal resource allocation of microgrid groups poses a greater challenge to traditional optimization scheduling methods.

[0007] In summary, existing technologies, when applying master-slave game theory to the optimal scheduling of multiple microgrids, often struggle to systematically address the precise characterization of uncertain scenarios, efficient optimization while protecting data privacy, and coordinated operation of complex system structures. Therefore, a new operational approach is urgently needed that can more accurately and efficiently address the uncertainties of renewable energy and loads while protecting the data privacy of each microgrid, thereby achieving optimal allocation of regional system resources. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a multi-microgrid operation method and system based on price-sensitivity-weighted scenario reduction. It constructs a master-slave game market framework, including an upper-level leader model and a lower-level follower model. A typical scenario is obtained using a price-sensitivity-weighted scenario reduction method. The upper-level leader model uses minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function. Each microgrid in the lower-level follower model receives an internal transaction price given by the upper-level leader model and aims to minimize its own comprehensive operating cost, thus obtaining transaction volume data. Combining the transaction volume data, the upper-level leader model calculates the fitness of the current internal transaction price and the feedback quality index based on an adaptive mechanism of feedback quality, and generates a new generation of internal transaction prices based on the feedback quality index. This iterative process is repeated to output the optimal internal transaction price strategy. This invention can more accurately address the uncertainties of renewable energy and load, achieving optimized allocation of regional system resources.

[0009] The present invention adopts the following technical solution.

[0010] This invention proposes a multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction, including: Construct a master-slave game market framework, including an upper-level leader model and a lower-level follower model; Several typical scenarios were obtained by using a scenario reduction method that weights scenarios based on electricity price sensitivity; Based on the aforementioned typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost. Combining transaction volume data, the upper-level leader model calculates the fitness and feedback quality indicators of the current internal transaction electricity price based on an adaptive mechanism of feedback quality, and generates a new generation of internal transaction electricity prices based on the feedback quality indicators. Repeat the above iterative process until the fitness meets the iteration condition, and output the optimal internal trading electricity price strategy.

[0011] More preferably, the specific method for obtaining several typical scenarios using the electricity price sensitivity-weighted scenario reduction method includes: Based on the probability distribution of uncertain variables, Monte Carlo simulation is used to generate the original operating scenario; the uncertain variables include wind speed, light intensity, and load. For the uncertain variable at each time step probability distribution Sub-independent random sampling, generating A primitive high-dimensional scene; set up To include all A collection of original scenes generated by Monte Carlo remained unchanged throughout the cut process; The set of scenes to be reduced in the current iteration is initially equal to As the reduction process gradually shrinks; Initialize scene set To include all A collection of original scenes; for Each scene in Calculate the new scene set formed after removing it. With the original scene set The weighted reduction criterion values ​​between; For all calculated weighted reduction criterion values, select the scenario corresponding to the smallest weighted reduction criterion value. Remove; Repeat the above steps until... The number of scenes equals the number of targets. The target quantity is set according to requirements; The final result The scenario described above is a typical scenario.

[0012] More preferably, the method for calculating the weighted reduction criterion value is as follows: for Each scene in ,calculate and remove scene The new scene set that followed Calculate the Wasserstein distance between them to obtain the distance result; Electricity price sensitivity index for computing scenarios; The scenario is obtained by weighting and summing the distance result and the reciprocal of the electricity price sensitivity index. The weighted reduction criterion value.

[0013] More preferably, the method for calculating the electricity price sensitivity index is as follows: In the scene Below, calculate the transaction volume of microgrid i in terms of electricity price over time. The absolute value of the rate of change; The absolute values ​​corresponding to the N microgrids are added together to calculate the scenario. Electricity price sensitivity index.

[0014] More preferably, each microgrid in the lower-level follower model receives an internal transaction price given by the upper-level leader model, aiming to minimize its own overall operating costs, and obtains transaction volume data; the specific process is as follows: The overall operating cost of the company includes the operating cost of internal equipment, the energy trading cost with market operators, and the carbon emission cost generated by the internal conventional generating units. The constraints include power balance constraints, energy storage operation constraints, conventional unit output constraints, V2G aggregation constraints, interruptible load constraints, and network security constraints. Based on the stated objective and constraints, the lower-level follower model is solved to obtain trading volume data, including buy volume. and sales volume , where the subscript i represents the i-th microgrid, s represents the s-th typical scenario, and t represents the t-th time period.

[0015] More preferably, the V2G aggregation constraint refers to the operational constraints of the aggregated electric vehicle charging and discharging V2G model, including state transition constraints, charging and discharging power constraints, and off-grid demand constraints; In the optimization process of the lower-level follower model, the total adjustable power of the aggregated electric vehicle charging and discharging V2G model is incorporated into the overall power balance of the microgrid for unified optimization.

[0016] More preferably, the specific steps of the adaptive mechanism based on feedback quality are as follows: Define the m-th particle This represents a complete set of internal trading electricity pricing strategies, in which and These represent the buy price and sell price at time interval t, respectively. Calculate the feedback quality index for the k-th iteration. Calculate the cost improvement after the k-th iteration relative to the (k-1)-th iteration; use the total system cost of the k-th iteration as the denominator and the cost improvement as the numerator to obtain a first ratio result; calculate the standard deviation of each microgrid transaction volume data, and use the reciprocal of the standard deviation as a second ratio result; multiply the first ratio result and the second ratio result to obtain the feedback quality index for the k-th iteration. ; According to the feedback quality indicators Adjust the inertial weights that control the global and local search capabilities of particles. and learning factors and According to the adjusted inertia weights and learning factors and For particles speed and location Update; obtain the first The internal transaction price for the next iteration.

[0017] More preferably, the inertial weight The adjustment method is as follows: Set the maximum and minimum inertia weights according to the actual situation, and adjust the sensitivity parameters. ; Calculate the first difference between the maximum and minimum inertia weights, and calculate the parameters. With the aforementioned feedback quality indicators The opposite of the product; Multiply the first difference by the function value with base e and the opposite of e as the exponent, and add the multiplication result to the minimum inertia weight to obtain the inertia weight. ; The learning factor and The adjustment method is as follows: For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient and the feedback quality indicators Add the product of 1 and 1, then add the result to the baseline value of the learning factor. Multiply by the product to obtain the learning factor. The adjusted result; For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient is calculated. and the feedback quality indicators The product result is then subtracted from 1, and the result of the subtraction is compared with the learning factor baseline value. Multiply by the product to obtain the learning factor. The adjusted result.

[0018] This invention also proposes a multi-microgrid operation system based on electricity price sensitivity weighted scenario reduction, including a typical scenario generation module, a master-slave game iteration module, and an optimal internal transaction price strategy output module: A master-slave game market framework construction module, wherein the master-slave game market framework includes an upper-level leader model and a lower-level follower model; The typical scenario generation module uses a scenario reduction method weighted by electricity price sensitivity to generate several typical scenarios; In the master-slave game iterative module, based on the typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost; combining the transaction volume data, the upper-level leader model calculates the fitness and feedback quality index of the current internal transaction price based on the adaptive mechanism of feedback quality, and generates a new generation of internal transaction prices based on the feedback quality index; The optimal internal transaction electricity price strategy output module repeats the above iterative process until the fitness reaches the iteration condition, and then outputs the optimal internal transaction electricity price strategy.

[0019] The present invention also proposes a terminal, including a processor and a storage medium: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0020] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0021] Compared with the prior art, the present invention has the following beneficial effects: (1) Improved overall system economy and operational efficiency. This invention deeply couples the two core innovations of "intelligent scenario reduction" and "adaptive optimization" with a hierarchical master-slave game framework. This framework provides an operating environment for innovative technologies, while the innovative technologies solve the inherent problems within this framework. The two work together to achieve optimized allocation of system resources.

[0022] (2) It enables more accurate and robust optimization decisions in the face of uncertainty. Unlike traditional deterministic optimization or scenario reduction methods that only consider probability similarity, the "electricity price sensitivity-weighted scenario reduction" technology proposed in this invention can screen out typical scenarios that have a key impact on game equilibrium. This allows subsequent optimization decisions to be based on a more representative set of scenarios, and the resulting scheduling scheme can better cope with the random fluctuations of renewable energy and load.

[0023] (3) This invention solves the optimization problem under conditions of incomplete information while protecting data privacy. Existing technologies struggle to achieve efficient collaborative optimization while protecting the commercial secrets of each microgrid. This invention utilizes a two-layer distributed architecture and an "adaptive PSO algorithm based on feedback quality." The upper layer only needs to publish electricity prices and receive transaction volume feedback from the lower layer, without needing to obtain sensitive data such as internal costs and energy storage status of the microgrid. This mechanism, while strictly protecting data privacy, uses macro-level feedback information to infer the system state, achieving efficient and stable optimization under conditions of "incomplete information."

[0024] (4) A closed-loop optimization of "scenario-electricity price-feedback" is formed, generating a synergistic enhancement effect. The various steps of this invention are not simply connected in series, but constitute a deeply coupled intelligent closed-loop system. The intelligent scenario reduction at the front end provides high-quality input for the adaptive optimization at the back end; the feedback information generated by the back end optimization verifies and realizes the value of the scenario reduction at the front end. This closed-loop collaborative mechanism makes the overall optimization effect significantly better than the traditional method of simply superimposing various technical modules. Attached Figure Description

[0025] Figure 1 This is a flowchart of the multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction of the present invention; Figure 2 This is a flowchart of the method according to Embodiment 1 of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0027] This invention provides the following technical solution: like Figure 1 As shown, this invention proposes a multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction, including the following steps: Construct a master-slave game market framework, including an upper-level leader model and a lower-level follower model; Several typical scenarios were obtained by using a scenario reduction method that weights scenarios based on electricity price sensitivity; The specific methods for obtaining several typical scenarios using the electricity price sensitivity weighted scenario reduction method include: Based on the probability distribution of uncertain variables, Monte Carlo simulation is used to generate the original operating scenario; the uncertain variables include wind speed, light intensity, and load. For the uncertain variable at each time step probability distribution Sub-independent random sampling, generating A primitive high-dimensional scene; set up To include all A collection of original scenes generated by Monte Carlo remained unchanged throughout the cut process; The set of scenes to be reduced in the current iteration is initially equal to As the reduction process gradually shrinks; Initialize scene set To include all A collection of original scenes; for Each scene in Calculate the new scene set formed after removing it. With the original scene set The weighted reduction criterion values ​​between; For all calculated weighted reduction criterion values, select the scenario corresponding to the smallest weighted reduction criterion value. Remove; Repeat the above steps until... The number of scenes equals the number of targets. The target quantity is set according to requirements; The final result The scenario described above is a typical scenario.

[0028] The method for calculating the weighted reduction criterion value is as follows: for Each scene in ,calculate and remove scene The new scene set that followed Calculate the Wasserstein distance between them to obtain the distance result; Electricity price sensitivity index for computing scenarios; The scenario is obtained by weighting and summing the distance result and the reciprocal of the electricity price sensitivity index. The weighted reduction criterion value.

[0029] The calculation method for the electricity price sensitivity index is as follows: In the scene Below, calculate the transaction volume of microgrid i in terms of electricity price over time. The absolute value of the rate of change; The absolute values ​​corresponding to the N microgrids are added together to calculate the scenario. Electricity price sensitivity index.

[0030] Based on the aforementioned typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost. Each microgrid in the lower-level follower model receives an internal transaction price from the upper-level leader model, aiming to minimize its own overall operating costs and obtain transaction volume data; the specific process is as follows: The overall operating cost of the company includes the operating cost of internal equipment, the energy trading cost with market operators, and the carbon emission cost generated by the internal conventional generating units. The constraints include power balance constraints, energy storage operation constraints, conventional unit output constraints, V2G aggregation constraints, interruptible load constraints, and network security constraints. Based on the stated objective and constraints, the lower-level follower model is solved to obtain trading volume data, including buy volume. and sales volume , where the subscript i represents the i-th microgrid, s represents the s-th typical scenario, and t represents the t-th time period.

[0031] The V2G aggregation constraints refer to the operational constraints of the aggregated electric vehicle charging and discharging V2G model, including state transition constraints, charging and discharging power constraints, and off-grid demand constraints. In the optimization process of the lower-level follower model, the total adjustable power of the aggregated electric vehicle charging and discharging V2G model is incorporated into the overall power balance of the microgrid for unified optimization.

[0032] Combining transaction volume data, the upper-level leader model calculates the fitness and feedback quality indicators of the current internal transaction electricity price based on an adaptive mechanism of feedback quality, and generates a new generation of internal transaction electricity prices based on the feedback quality indicators. The specific steps of the adaptive mechanism based on feedback quality are as follows: Define the m-th particle This represents a complete set of internal trading electricity pricing strategies, in which and These represent the buy price and sell price at time interval t, respectively. Calculate the feedback quality index for the k-th iteration. Calculate the cost improvement after the k-th iteration relative to the (k-1)-th iteration; use the total system cost of the k-th iteration as the denominator and the cost improvement as the numerator to obtain a first ratio result; calculate the standard deviation of each microgrid transaction volume data, and use the reciprocal of the standard deviation as a second ratio result; multiply the first ratio result and the second ratio result to obtain the feedback quality index for the k-th iteration. ; According to the feedback quality indicators Adjust the inertial weights that control the global and local search capabilities of particles. and learning factors and According to the adjusted inertia weights and learning factors and For particles speed and location Update; obtain the first The internal transaction price for the next iteration.

[0033] The inertial weight The adjustment method is as follows: Set the maximum and minimum inertia weights according to the actual situation, and adjust the sensitivity parameters. ; Calculate the first difference between the maximum and minimum inertia weights, and calculate the parameters. With the aforementioned feedback quality indicators The opposite of the product; Multiply the first difference by the function value with base e and the opposite of e as the exponent, and add the multiplication result to the minimum inertia weight to obtain the inertia weight. ; The learning factor and The adjustment method is as follows: For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient and the feedback quality indicators Add the product of 1 and 1, then add the result to the baseline value of the learning factor. Multiply by the product to obtain the learning factor. The adjusted result; For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient is calculated. and the feedback quality indicators The product result is then subtracted from 1, and the result of the subtraction is compared with the learning factor baseline value. Multiply by the product to obtain the learning factor. The adjusted result.

[0034] Repeat the above iterative process until the fitness meets the iteration condition, and output the optimal internal trading electricity price strategy.

[0035] This invention also proposes a multi-microgrid operation system based on electricity price sensitivity weighted scenario reduction, including a typical scenario generation module, a master-slave game iteration module, and an optimal internal transaction price strategy output module: A master-slave game market framework construction module, wherein the master-slave game market framework includes an upper-level leader model and a lower-level follower model; The typical scenario generation module uses a scenario reduction method weighted by electricity price sensitivity to generate several typical scenarios; In the master-slave game iterative module, based on the typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost; combining the transaction volume data, the upper-level leader model calculates the fitness and feedback quality index of the current internal transaction price based on the adaptive mechanism of feedback quality, and generates a new generation of internal transaction prices based on the feedback quality index; The optimal internal transaction electricity price strategy output module repeats the above iterative process until the fitness reaches the iteration condition, and then outputs the optimal internal transaction electricity price strategy.

[0036] The present invention also proposes a terminal, including a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0037] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the steps of the above-described method.

[0038] Example 1 As an embodiment of this application, a specific implementation method for a multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction is disclosed. The execution flow of the method embodiment is as follows: Figure 2 .

[0039] To address the problems of existing flattened game theory models, such as lack of macro-control, amplified uncertainty risks, and difficulty in finding the optimal solution under incomplete information, as described in the background art, this invention provides a hierarchical master-slave game theory operation method. This method first constructs a market framework, handles uncertainty through Monte Carlo scenario generation (S1) and scenario reduction incorporating electricity price sensitivity (S2), and uses an adaptive algorithm based on feedback quality (S3) for optimization. When the microgrid only provides feedback on interactive electricity (S4), iterative game theory (S5) is used to obtain the optimal electricity price and operating strategy.

[0040] The following is a detailed explanation of each step: S1: Electricity price sensitivity weighted scenario reduction This step generates and reduces uncertainties. By introducing electricity price sensitivity as a reduction criterion, the original scenarios generated by the Monte Carlo simulation are filtered, retaining those that have a significant impact on microgrid trading behavior. The typical scenario set output by this step reflects the system's price response characteristics under different resource and load conditions, thereby improving the targeting and efficiency of subsequent optimization calculations.

[0041] 1.1: In a further implementation, Monte Carlo simulation is used to generate original operating scenarios. A large number of original operating scenarios are generated by randomly sampling the probability distribution function of the uncertain variables.

[0042] For wind speed modeling, the Weibull distribution is typically used, and its probability density function is:

[0043] In the formula, v is the wind speed, k is the wind speed shape parameter, and c is the scale parameter, reflecting the magnitude of the average wind speed. The parameters k and c are determined through the following steps: collecting historical wind speed data for a preset duration (no less than one year) in the target area; cleaning and preprocessing the historical wind speed data to eliminate outliers; based on the processed data, fitting the Weibull distribution parameters using the maximum likelihood estimation method; and ensuring the reliability of the distribution through a goodness-of-fit test (using the coefficient of determination R²). This is existing technology and will not be elaborated further. In this embodiment, k∈[1.5, 2.5], c∈[3, 10].

[0044] For modeling light intensity, the beta distribution is typically used, and its probability density function is:

[0045] In the formula, r is the normalized light intensity. For the shape parameters of the illumination, This is the gamma function. The parameters... The following steps are used to determine the target area: Historical light intensity data for a preset duration (e.g., no less than one year) is collected; the historical light intensity data is cleaned and normalized; based on the processed data, the beta distribution parameters are fitted using the maximum likelihood estimation method; and the reliability of the distribution is ensured through a goodness-of-fit test (using the coefficient of determination R²). This is prior art and will not be elaborated further. In this embodiment, .

[0046] For load modeling, the prediction error of the load typically follows a normal distribution, with the following probability density function:

[0047] In the formula, x is the load power. To predict the mean, Let Variance be the variance.

[0048] By analyzing the above variables at each time step probability distribution One independent random sampling can generate There are 1000 original high-dimensional scenes, which can be set according to computing resources and accuracy requirements. Preferably, they can be set to 1000. Each scene n can be represented as a vector. .

[0049] 1.2: In a further implementation, a scenario reduction method weighted by electricity price sensitivity is used to reduce the original scenario set into a typical scenario set. This method introduces electricity price sensitivity as a reduction criterion based on the traditional synchronous backtracking reduction algorithm, in order to retain the key scenarios that are most important for electricity price optimization.

[0050] 1.2.1 Define a price sensitivity index for the scenario to measure the responsiveness of microgrid trading behavior to changes in electricity prices:

[0051] In the formula, N represents the number of microgrids. For the transaction volume of microgrid i, For time-period electricity price scalars, it represents the electricity price strategy vector for a given time period. A pricing strategy vector comprising a specific component or covering the entire dispatch cycle T. This indicator reflects the sensitivity of a microgrid's self-manufacturing versus purchasing decisions to electricity price changes when the scenario is at a critical juncture between renewable energy output and load demand. This price sensitivity is obtained through pre-testing, specifically including: designing M groups of different sample pricing strategies; for each scenario... The lower-level microgrid optimization model was solved using mixed-integer linear programming under each of the M electricity price groups to obtain the corresponding transaction volume data. Finally, the electricity price sensitivity of this scenario is calculated according to the formula based on the rate of change of transaction volume with electricity price.

[0052] 1.2.2 Construct a weighted reduction criterion that combines Wasserstein distance with electricity price sensitivity:

[0053] In the formula, The Wasserstein distance measures the difference in probability distributions; and These are the distribution weights (used to weight the Wasserstein distance, controlling the similarity in probability distribution between the reduced scene set and the original scene set) and the sensitivity weights (used to weight the reciprocal of electricity price sensitivity, controlling the degree of emphasis on the criticality of the optimization objective during the reduction process), satisfying... and This is used to balance the importance of maintaining the probability distribution and maintaining electricity price sensitivity, and can be adjusted according to different requirements for distribution accuracy and price response accuracy in actual application scenarios; the first item ensures the preservation of probability distribution characteristics, and the second item ensures the preservation of high-sensitivity scenarios.

[0054] The algorithm uses Wasserstein distance to quantify the difference between two scene sets and their corresponding probability distributions. For ease of general definition, the following uses... and This refers generally to any scene element supported by two probability distributions, not specifically to the specific scene in the reduction process mentioned above. Two discrete probability distributions are defined as... and , and They are the scenes and The corresponding probability satisfies , ; and Let be the Dirac function, representing the probability distribution in the scene. and The unit probability mass concentrated at a given location. In the scene reduction application of this invention, Corresponding original scene set The probability distribution, The probability distribution after removing a scene from the current scene set The Wasserstein distance between P1 and P2 is defined as:

[0055] In the formula, yes arrive The transportation cost is defined as the Euclidean distance between the vectors of the two scenes. ; It is the joint probability distribution of all possible possibilities. The set whose marginal distribution is .

[0056] 1.2.3 The iterative reduction process is as follows: Note: To include all A collection of original scenes generated by Monte Carlo remained unchanged throughout the cut process; The set of scenes to be reduced in the current iteration is initially equal to As the reduction process gradually shrinks.

[0057] Step 1: Initialize the scene set To include all A collection of original scenes.

[0058] Step 2: In each iteration, for Each scene in Calculate the new scene set formed after removing it. With the original scene set The weighted reduction criterion values ​​between them.

[0059] Step 3: Select the scenario that minimizes the criterion value. Remove the scene and update the probability of the remaining scene, where This represents the probability distribution corresponding to the scene set, and probability redistribution is required when calculating distance.

[0060] Step 4: Repeat steps 2 and 3 until... The number of scenes equals the number of targets. The target number can be set according to the need for a balance between computational complexity and scenario representativeness.

[0061] Step 5: The final result The scenario is a typical scenario, and its corresponding probability is the final probability after being updated during the iteration process.

[0062] S2: Construction of a Master-Slave Game Market Framework This step establishes the hierarchical architecture of the system, defining the optimization objectives and decision variables for upper-level market operators and lower-level microgrids. This architecture decomposes the complex regional energy management problem into two sub-problems: upper-level pricing and lower-level response. It also establishes a mechanism for information exchange between the upper and lower levels through "electricity price-electricity consumption," thus defining the boundaries for subsequent distributed solutions.

[0063] 2.1: In a further implementation, the upper-level leader model is as follows: The goal of the market operator is to minimize the expected total comprehensive operating cost of the entire regional power grid under various uncertain scenarios. This cost consists of conventional operating costs, and its decision variables are the internal transaction electricity prices (buying price and selling price) formulated and issued to the lower level.

[0064] 2.2: In a further implementation, the lower-level follower model is as follows: After receiving the electricity price signal given by the upper level, each microgrid aims to minimize its own overall operating cost, which consists of the internal equipment operating cost and energy trading cost. It optimizes the operating strategies of its internal equipment (energy storage, gas turbine, interruptible load, etc.) and decides on the amount of electricity to be traded with the market operator.

[0065] 2.3: In a further implementation, the system model of the lower-level microgrid may include an aggregated electric vehicle charging and discharging (V2G) model, which is used as a flexibly dispatchable mobile energy storage resource for optimization.

[0066] Operational constraints of aggregated electric vehicle fleets: The aggregator (which may also be a microgrid operator) aggregates a large number of spatiotemporally discrete electric vehicles (EVs) into a unified, schedulable virtual battery. For any EV e in the aggregator, it must satisfy the following constraints during the scheduling period: (1) State transition constraints:

[0067] In the formula, Let e ​​be the state of charge of the electric vehicle e at time t; and These are its charging and discharging power, respectively; and For charge and discharge efficiency; This refers to the battery capacity.

[0068] (2) Charging and discharging power constraints:

[0069]

[0070] In the formula, and This is the maximum charge / discharge power; It is a binary variable to ensure that charging and discharging behaviors do not occur simultaneously.

[0071] (3) Off-grid demand constraints:

[0072] In the formula, The preset off-grid time for electric vehicle e. This refers to the minimum state of charge required for the electric vehicle (e) to meet its next trip. This constraint is crucial to ensuring that V2G services do not impact users' basic travel needs.

[0073] In the lower-level optimization, the aggregator needs to ensure that all vehicles in the aggregator meet the above constraints, and at the same time, adjust the total adjustable power of the aggregator. The overall power balance of the microgrid is optimized in a unified manner.

[0074] S3: Adaptive PSO Algorithm Based on Feedback Quality This step employs an adaptive PSO algorithm based on feedback quality to solve the upper-level optimization model. A feedback quality index, Q, is defined using transaction volume statistics from the lower-level S4 layer, and the algorithm's inertia weights and learning factors are dynamically adjusted based on this index. This mechanism allows the upper-level model to adjust its search direction based on system response, achieving optimal electricity pricing strategy, even with only interactive electricity data. The electricity pricing strategy generated in this step will be passed as input to the lower-level S4 layer.

[0075] 3.1: In a further implementation, an adaptive mechanism based on feedback quality is established. This mechanism infers the system state from the transaction volume distribution and cost improvement of lower-level feedback, and then dynamically adjusts the PSO search strategy.

[0076] First, define the particle and feedback quality metrics. The m-th particle... This represents a complete set of electricity pricing strategies, namely It can be regarded as a A dimensional vector, where and These represent the buy price and sell price at time interval t, respectively. Correspondingly, the particle's velocity... Individual historical best position and the global optimal position All are vectors of the same dimension, and their updates follow the vector operation rules of the standard particle swarm optimization algorithm. Define the feedback quality index for the k-th iteration. :

[0077] In the formula, This represents the cost improvement after the k-th iteration relative to the (k-1)-th iteration. Let $\frac{ ... This represents the standard deviation of the transaction volume data for each microgrid. Within this index, the cost improvement rate measures the effectiveness of the search direction, while transaction stability measures the system's equilibrium state. When the feedback quality index... When the value is negative (i.e., the cost increases), it indicates that the system cost has increased. The algorithm automatically increases the inertia weight to strengthen global exploration, which helps to quickly escape non-optimal regions.

[0078] Secondly, based on the quality of feedback Dynamically adjust inertia weight :

[0079] In the formula, These are the maximum and minimum inertia weights, used to control the global and local search capabilities of the particles, and can be adjusted according to the complexity of the optimization problem; The parameters used to adjust sensitivity.

[0080] Secondly, based on the quality of feedback Dynamically adjust learning factors and :

[0081] In the formula, It serves as the baseline value for the learning factor, used to balance the learning tendency of particles towards their individual optimal and global optimal positions; The adjustment coefficient controls the magnitude of the learning factor's adjustment in response to feedback quality.

[0082] Finally, in the In the next iteration, the particles speed and location Update using the following formula:

[0083]

[0084] In the formula, A random number within the interval [0, 1]; Let m be the individual historical best position of particle m at the k-th iteration; This is the globally optimal position at the k-th iteration.

[0085] S4: Lower-level microgrid optimization and feedback This step enables the lower-level microgrid to respond to the upper-level electricity price signal. Each microgrid solves its own cost minimization model based on the electricity price issued by S3, derives its optimal operating plan, and calculates the planned interaction volume with the upper level. This interaction volume is transmitted as feedback information to the upper-level S3 to evaluate the suitability of the current electricity price strategy. This process does not require uploading the microgrid's internal equipment parameters and detailed operating data.

[0086] 4.1: In a further implementation, the lower-level optimization model is a mixed-integer linear programming (MILP) problem. Its optimization objective is to minimize the overall operating cost of the microgrid itself.

[0087] In the formula, This includes the operating costs of internal equipment, such as gas turbine fuel costs and energy storage equipment depreciation costs. Energy trading costs with market operators; The carbon emission costs generated by the internal conventional generating units are considered. Constraints include power balance constraints, energy storage operation constraints, conventional unit output constraints, V2G aggregation constraints, interruptible load constraints, and network security constraints. Of these constraints, except for the V2G aggregation constraint, which is detailed in Section 1.3, the remaining constraints are standard constraints for microgrid optimization models known to those skilled in the art.

[0088] 4.2: In a further implementation, the optimal operating strategy is obtained by solving the above model, and the interaction power consumption is then compared with the operator's planned power consumption. Feedback is sent to the upper layer, where the subscript i represents the i-th microgrid, s represents the s-th typical scenario, and t represents the t-th time period. This process only reports the net exchanged electricity and does not upload detailed data from within the microgrid.

[0089] S5: Master-Slave Game Iteration and Convergence This step involves iterating and determining the convergence of the game. The upper-level operator collects the interactive electricity feedback from S4, calculates the fitness and feedback quality indicators of the current electricity pricing strategy, and updates the particle states of the PSO algorithm accordingly to generate a new generation of electricity pricing strategies. This process is repeated until the preset convergence conditions are met, outputting the final internal trading electricity price and the operation plan for each microgrid.

[0090] 5.1: In a further implementation, fitness and feedback quality are calculated. The fitness function is the upper-level optimization objective, i.e., the expected total cost across all scenarios:

[0091] in, The overall operating cost of the power grid; The number of typical scenarios; Let be the probability of scenario s; The position of the m-th particle in the k-th iteration represents the current electricity pricing strategy; The interaction cost between the regional power grid and the upstream main grid during time period t under scenario s; Let be the carbon emission cost generated by conventional generating units during time period t under scenario s; simultaneously calculate the feedback quality index for the k-th iteration. .

[0092] 5.2: In a further implementation, update and convergence determination are performed. First, the individual optimal position of each particle is updated. and global optimal position Secondly, based on the quality of feedback Calculate the new PSO parameters. Next, update the velocity and position of all particles for the next iteration. Finally, determine if the convergence conditions are met: reaching the maximum number of iterations, the globally optimal fitness value showing no significant improvement over multiple generations, or the feedback quality index remaining stable at a high level. If satisfied, output the optimal internal trading electricity price strategy.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

[0094] Example 2 This invention also proposes a multi-microgrid operation system based on electricity price sensitivity weighted scenario reduction, including a typical scenario generation module, a master-slave game iteration module, and an optimal internal transaction price strategy output module: A master-slave game market framework construction module, wherein the master-slave game market framework includes an upper-level leader model and a lower-level follower model; The typical scenario generation module uses a scenario reduction method weighted by electricity price sensitivity to generate several typical scenarios; In the master-slave game iterative module, based on the typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost; combining the transaction volume data, the upper-level leader model calculates the fitness and feedback quality index of the current internal transaction price based on the adaptive mechanism of feedback quality, and generates a new generation of internal transaction prices based on the feedback quality index; The optimal internal transaction electricity price strategy output module repeats the above iterative process until the fitness reaches the iteration condition, and then outputs the optimal internal transaction electricity price strategy.

[0095] Example 3 The present invention also proposes a terminal, including a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0096] Example 4 The present invention also proposes a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements the steps of the above-described method.

[0097] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction, characterized in that, include: Construct a master-slave game market framework, including an upper-level leader model and a lower-level follower model; Several typical scenarios were obtained by using a scenario reduction method that weights scenarios based on electricity price sensitivity; Based on the aforementioned typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost. Combining transaction volume data, the upper-level leader model calculates the fitness and feedback quality indicators of the current internal transaction electricity price based on an adaptive mechanism of feedback quality, and generates a new generation of internal transaction electricity prices based on the feedback quality indicators. Repeat the above iterative process until the fitness meets the iteration condition, and output the optimal internal trading electricity price strategy.

2. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 1, characterized in that: The specific methods for obtaining several typical scenarios using the electricity price sensitivity weighted scenario reduction method include: Based on the probability distribution of uncertain variables, Monte Carlo simulation is used to generate the original operating scenario; the uncertain variables include wind speed, light intensity, and load. For the uncertain variable at each time step probability distribution Sub-independent random sampling, generating A primitive high-dimensional scene; set up To include all A collection of original scenes generated by Monte Carlo remained unchanged throughout the cut process; The set of scenes to be reduced in the current iteration is initially equal to As the reduction process gradually shrinks; Initialize scene set To include all A collection of original scenes; for Each scene in Calculate the new scene set formed after removing it. With the original scene set The weighted reduction criterion values ​​between; For all calculated weighted reduction criterion values, select the scenario corresponding to the smallest weighted reduction criterion value. Remove; Repeat the above steps until... The number of scenes equals the number of targets. The target quantity is set according to requirements; The final result The scenario described above is a typical scenario.

3. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 2, characterized in that: The method for calculating the weighted reduction criterion value is as follows: for Each scene in ,calculate and remove scene The new scene set that followed Calculate the Wasserstein distance between them to obtain the distance result; Electricity price sensitivity index for computing scenarios; The scenario is obtained by weighting and summing the distance result and the reciprocal of the electricity price sensitivity index. The weighted reduction criterion value.

4. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 3, characterized in that: The calculation method for the electricity price sensitivity index is as follows: In the scene Below, calculate the transaction volume of microgrid i in terms of electricity price over time. The absolute value of the rate of change; The absolute values ​​corresponding to the N microgrids are added together to calculate the scenario. Electricity price sensitivity index.

5. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 1, characterized in that: Each microgrid in the lower-level follower model receives an internal transaction price from the upper-level leader model, aiming to minimize its own overall operating costs and obtain transaction volume data; the specific process is as follows: The overall operating cost of the company includes the operating cost of internal equipment, the energy trading cost with market operators, and the carbon emission cost generated by the internal conventional generating units. The constraints include power balance constraints, energy storage operation constraints, conventional unit output constraints, V2G aggregation constraints, interruptible load constraints, and network security constraints. Based on the stated objective and constraints, the lower-level follower model is solved to obtain trading volume data, including buy volume. and sales volume , where the subscript i represents the i-th microgrid, s represents the s-th typical scenario, and t represents the t-th time period.

6. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 5, characterized in that: The V2G aggregation constraints refer to the operational constraints of the aggregated electric vehicle charging and discharging V2G model, including state transition constraints, charging and discharging power constraints, and off-grid demand constraints. In the optimization process of the lower-level follower model, the total adjustable power of the aggregated electric vehicle charging and discharging V2G model is incorporated into the overall power balance of the microgrid for unified optimization.

7. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 1, characterized in that: The specific steps of the adaptive mechanism based on feedback quality are as follows: Define the m-th particle This represents a complete set of internal trading electricity pricing strategies, in which and These represent the buy price and sell price at time interval t, respectively. Calculate the feedback quality index for the k-th iteration. : Calculate the cost improvement after the k-th iteration relative to the (k-1)-th iteration; Using the total system cost of the k-th iteration as the denominator and the cost improvement amount as the numerator, the first ratio result is obtained; Calculate the standard deviation of the transaction volume data for each microgrid, and use the reciprocal of the standard deviation as the second ratio result; multiply the first ratio result and the second ratio result to obtain the feedback quality index for the k-th iteration. ; According to the feedback quality indicators Adjust the inertial weights that control the global and local search capabilities of particles. and learning factors and According to the adjusted inertia weights and learning factors and For particles speed and location Update; obtain the first The internal transaction price for the next iteration.

8. The multi-microgrid operation method based on electricity price sensitivity weighted scenario reduction according to claim 7, characterized in that: The inertial weight The adjustment method is as follows: Set the maximum and minimum inertia weights according to the actual situation, and adjust the sensitivity parameters. ; Calculate the first difference between the maximum and minimum inertia weights, and calculate the parameters. With the aforementioned feedback quality indicators The opposite of the product; Multiply the first difference by the function value with base e and the opposite of e as the exponent, and add the multiplication result to the minimum inertia weight to obtain the inertia weight. ; The learning factor and The adjustment method is as follows: For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient and the feedback quality indicators Add the product of 1 and 1, then add the result to the baseline value of the learning factor. Multiply by the product to obtain the learning factor. The adjusted result; For learning factors Set the adjustment coefficient according to the actual situation. The adjustment coefficient is calculated. and the feedback quality indicators The product result is then subtracted from 1, and the result of the subtraction is compared with the learning factor baseline value. Multiply by the product to obtain the learning factor. The adjusted result.

9. A multi-microgrid operation system based on electricity price sensitivity weighted scenario reduction using the method described in any one of claims 1-8, comprising a typical scenario generation module, a master-slave game iteration module, and an optimal internal transaction electricity price strategy output module, characterized in that: A master-slave game market framework construction module, wherein the master-slave game market framework includes an upper-level leader model and a lower-level follower model; The typical scenario generation module uses a scenario reduction method weighted by electricity price sensitivity to generate several typical scenarios; In the master-slave game iterative module, based on the typical scenario, the upper-level leader model takes minimizing the expected total comprehensive operating cost of the entire regional power grid as its objective function; each microgrid in the lower-level follower model receives the internal transaction price given by the upper-level leader model and obtains transaction volume data with the objective of minimizing its own comprehensive operating cost; combining the transaction volume data, the upper-level leader model calculates the fitness and feedback quality index of the current internal transaction price based on the adaptive mechanism of feedback quality, and generates a new generation of internal transaction prices based on the feedback quality index; The optimal internal transaction electricity price strategy output module repeats the above iterative process until the fitness reaches the iteration condition, and then outputs the optimal internal transaction electricity price strategy.

10. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-8.

Images