Power supply output coordination control method combined with electricity market transaction decision-making model

Abstract

The invention, which belongs to the technical field of energy output coordination, provides a power supply output coordination control method of a joint power market transaction decision model. A three-layer architecture joint power market transaction decision model comprising a market layer joint transaction framework, a bidding layer game decision module and a scheduling layer execution control module is constructed; establishing a piecewise nonlinear function relationship between the energy storage accelerated aging coefficient and the energy storage charge state, associating the energy storage cycle loss cost with the energy storage charge and discharge power and the charge state and incorporating the energy storage cycle loss cost, the energy storage charge and discharge power and the charge state into an energy storage operator objective function, and adopting a multi-time scale coordinated optimization framework and a hybrid optimization solution strategy. A charge-discharge control scheme which simultaneously meets the requirements of market transaction income optimization and energy storage life protection is obtained, and the technical problem that the charge-discharge cycle life loss of an energy storage system in a multi-market joint transaction environment is difficult to accurately quantify and is brought into an optimization decision is solved.

Description

Technical Field

[0001] This invention belongs to the field of energy output coordination technology, and more specifically, relates to a power output coordination control method for a joint electricity market trading decision model. Background Technology

[0002] Against the backdrop of deepening power market reform, energy storage systems, as an important flexible regulation resource, participate in multiple market transactions, including the electricity market, frequency regulation market, and spinning reserve market. Traditional energy storage operation optimization methods mainly use fixed cost or simple linear depreciation models to estimate energy storage operating costs, and formulate energy storage charging and discharging strategies and capacity configuration schemes by maximizing market transaction benefits. However, the cycle life of energy storage batteries is affected by the nonlinear coupling of various operating parameters such as state of charge, depth of charge and discharge, and charge and discharge rate. Frequent charge and discharge cycles accelerate battery capacity decay and increase internal resistance. Especially when operating in deep charge and discharge and extreme state of charge ranges, the irreversibility of electrochemical reactions leads to accelerated energy storage life decay. Traditional fixed cost models cannot accurately reflect the true life loss characteristics of energy storage under different operating conditions. This may cause energy storage to operate in an accelerated aging range when pursuing short-term market profit maximization, resulting in a serious decline in the economic efficiency of energy storage throughout its entire life cycle. In other words, existing technologies face the technical problem of accurately quantifying the charge and discharge cycle life loss of energy storage systems in a multi-market joint trading environment and incorporating it into optimization decisions. Summary of the Invention

[0003] In view of this, the present invention provides a power output coordination control method for a joint electricity market trading decision model, which can solve the technical problem in the prior art that it is difficult to accurately quantify and incorporate the charge and discharge cycle life loss of energy storage systems in a multi-market joint trading environment into the optimization decision.

[0004] This invention is implemented as follows: It provides a power output coordination control method for a joint power market trading decision model. The method constructs a joint power market trading decision model encompassing the power market, frequency regulation market, and spinning reserve market. A multi-timescale coordination optimization framework is established, and the transmission relationship between long-term decision variables and short-term boundary conditions is established through a cross-scale state transfer matrix. A power node topology weighted graph is constructed, and hierarchical coordination control partitioning is performed. The power node topology weighted graph is divided into multiple weakly coupled control partitions using a spectral clustering algorithm. Objective functions for each market participant are established, and a risk quantification model is introduced. A frequency dynamic stability constraint model is established, and multi-level constraint condition groups are set. A multi-machine system is converted into a single-machine equivalent system using a central inertia equivalent transformation method. The rotor motion frequency response equation describes the dynamic relationship between system frequency and power imbalance. The rotor motion frequency response equation is transformed into a frequency deviation algebraic inequality constraint through frequency response linearization transformation. A hybrid optimization solution strategy is used to solve the joint power market trading decision model, outputting power output coordination control commands and performing hierarchical control.

[0005] It should be noted that this invention establishes a piecewise nonlinear functional relationship between the energy storage accelerated aging coefficient and the energy storage state of charge (SOC). This accurately quantifies the electrochemical aging characteristics of energy storage batteries in different SOC ranges into mathematical expressions, accurately reflecting the electrochemical physical law that aging is significantly accelerated in the deep discharge range and relatively mild in the medium-to-high SOC range. This solves the defect of traditional fixed-cost models that cannot reflect the true lifetime loss characteristics. This invention combines the energy storage accelerated aging coefficient with the cumulative value of the actual number of energy storage cycles to establish a coupling relationship between energy storage cycle loss cost and energy storage charge / discharge power, energy storage SOC, and energy storage charge / discharge depth. This cycle loss cost is incorporated as an important component of the energy storage operator's objective function into the joint electricity market trading decision model. Combined with a multi-timescale coordinated optimization framework and a hybrid optimization solution strategy, this allows energy storage to comprehensively consider the dynamic balance between market trading revenue and lifetime loss cost when formulating charge / discharge strategies, automatically avoiding accelerated aging operating ranges. This solves the defect of traditional methods where energy storage optimization decisions cannot simultaneously consider long-term lifetime protection. In summary, this invention solves the technical problem mentioned in the background art of accurately quantifying and incorporating the charge-discharge cycle life loss of energy storage systems into optimization decisions by establishing a quantitative model of the accelerated aging coefficient of energy storage based on the state of charge and transforming it into a cycle loss cost that can be embedded in the optimization objective function. Attached Figure Description

[0006] Figure 1 Convergence curve of the objective function for improving the whale population optimization algorithm.

[0007] Figure 2This diagram shows the results of the coordinated optimization of the power output of each power source and the power demand of the load over a 24-hour period.

[0008] Figure 3 This diagram illustrates the system's dynamic frequency response and the effect of frequency modulation resource coordination control after power disturbance.

[0009] Figure 4 This is a framework diagram for joint electricity market transactions.

[0010] Figure 5 This is a game structure diagram for a joint electricity market trading model. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0012] The invention provides a power output coordination control method for a joint electricity market trading decision model, comprising: S1. Construct a joint electricity market trading decision model, which includes a market-layer joint trading framework, a bidding-layer game decision module, and a dispatch-layer execution control module. The market-layer joint trading framework includes the electricity market, frequency regulation market, and spinning reserve market. S2. Establish a multi-timescale coordination and optimization framework. Discretize the day-ahead clearing time window with a 1-hour time granularity, discretize the real-time scheduling with a 15-minute time granularity, and discretize the frequency modulation response with a 1-second time granularity. Establish the transmission relationship between long-timescale decision variables and short-timescale boundary conditions through a cross-scale state transfer matrix. S3. Construct a power node topology weighted graph and perform hierarchical coordination control partitioning. Establish a power node topology weighted graph with each power node as the vertex. Use electrical coupling strength as the edge weight of the power node topology weighted graph. Divide the power node topology weighted graph into multiple weakly coupled control partitions through a spectral clustering segmentation algorithm. Each weakly coupled control partition achieves inter-partition coordination through a boundary power consistency protocol. S4. Establish objective functions for each market participant and introduce a risk quantification model. The objective function for energy storage operators is the sum of energy storage electricity market revenue, energy storage frequency regulation market revenue, and energy storage reserve market revenue, minus the energy storage life-cycle amortized cost and energy storage cycle loss cost. The objective function for wind power operators is the sum of wind power electricity revenue, wind power frequency regulation revenue, and wind power reserve revenue, minus the wind curtailment penalty cost. The objective function for grid operators is the sum of thermal power operating costs, energy storage operating costs, wind power operating costs, and photovoltaic operating costs. The Copula-CVaR risk quantification model is used to calculate the risk value of revenue loss caused by the uncertainty of renewable energy output and incorporate it into the objective functions of each market participant. S5. Establish a frequency dynamic stability constraint model and set a multi-level constraint condition group. Convert the multi-machine system into a single-machine equivalent system through the center inertia equivalent transformation method. Use the rotor motion frequency response equation to describe the dynamic relationship between the system frequency and the power imbalance. Transform the rotor motion frequency response equation into a frequency deviation algebraic inequality constraint through frequency response linearization transformation. The multi-level constraint condition group includes energy storage charging and discharging power constraints, power supply frequency regulation capacity constraints, power supply reserve capacity constraints, power balance constraints, frequency regulation demand constraints, and reserve demand constraints. S6. A hybrid optimization solution strategy is adopted to solve the joint electricity market transaction decision model. The non-convex quadratic constraint is transformed into a second-order conical convex constraint by using a hybrid integer second-order cone relaxation transformation. The branch and bound search algorithm is used to reduce the search space of integer variables. A high-quality initial solution is generated by an improved whale population optimization algorithm. The upper-level Cplex solver is combined with the lower-level improved whale population optimization algorithm to obtain the transaction price of each market, the output power of each power source, the charging and discharging power of energy storage, the frequency regulation capacity configuration value and the reserve capacity configuration value. S7. Output power supply coordination control command and execute hierarchical control. Send the output power of each power supply as the active power setpoint to the corresponding power controller. Convert the energy storage charging and discharging power into energy storage converter power command and send it to the energy storage converter. Convert the frequency regulation capacity configuration value into frequency regulation controller capacity parameter setting and send it to the frequency regulation controller. Convert the standby capacity configuration value into standby response controller capacity parameter setting and send it to the standby response controller. Each power controller adjusts the active power output of the corresponding power supply to the target power value according to the active power setpoint of each power supply. The deviation between the target power value and the active power setpoint of each power supply is less than the power control accuracy threshold.

[0013] The market-layer joint trading framework includes a power trading center, which publishes power market transaction prices, frequency regulation market transaction prices, and spinning reserve market transaction prices to the bidding-layer game decision-making module based on the supply and demand balance principle. The bidding-layer game decision-making module includes decision-making units for grid operators, energy storage operators, wind power operators, and photovoltaic operators. The market-layer joint trading framework and the bidding-layer game decision-making module form a Stackelberg game relationship, while the energy storage operator, wind power operator, and photovoltaic operator decision-making units within the bidding-layer game decision-making module form a Nash game relationship. The dispatch-layer execution control module includes thermal power unit controllers, energy storage device controllers, wind farm controllers, and photovoltaic power plant controllers. The dispatch-layer execution control module reports operating status parameters, power generation cost parameters, and power generation plan parameters to the bidding-layer game decision-making module.

[0014] The cross-scale state transfer matrix is ​​used to describe the transfer relationship of system state variables between different time granularities. When the long-time scale decision variables are transferred as short-time scale boundary conditions, the power output obtained from the day-ahead clearing is used as the upper limit constraint and lower limit constraint of the power output for real-time scheduling, and the power setting benchmark value obtained from real-time scheduling is used as the power benchmark reference value for frequency regulation response. By updating the values ​​of decision variables for future periods within each optimization time window through a rolling time-domain optimization strategy, the temporal consistency of decision results at different time granularities is ensured.

[0015] The edge weights of the power node topology weighted graph represent the electrical coupling strength between power nodes, which is determined by the electrical distance between power nodes and the power transmission sensitivity. The spectral clustering segmentation algorithm calculates the eigenvalues ​​and eigenvectors of the Laplacian matrix of the power node topology weighted graph, clustering power nodes with high similarity index values ​​into the same weakly coupled control partition. A weakly coupled control partition refers to a control partition where the electrical coupling strength between power nodes within the partition is greater than the electrical coupling strength between power nodes between partitions. Within each weakly coupled control partition, a centralized model predictive control algorithm is used to optimize the power output within the partition. The prediction time domain length of the centralized model predictive control algorithm is 15 minutes, and the control time domain length is 5 minutes. The boundary power consistency protocol uses iterative calculation to converge the boundary power exchange information of each weakly coupled control partition to a globally consistent state, ensuring that the global power balance constraint is satisfied. The system observability and controllability indices are evaluated using the Laplacian matrix eigenvalues. When the change in the distribution of the Laplacian matrix eigenvalues ​​exceeds the eigenvalue change threshold, the spectral clustering segmentation algorithm is re-executed to dynamically adjust the boundaries of the weakly coupled control partitions. The hierarchical coordinated control method decomposes the high-dimensional optimization problem into multiple low-dimensional sub-problems that are solved in parallel, reducing computational complexity, improving the real-time performance of control response, and enhancing the adaptability to changes in the power grid topology.

[0016] The revenue from the energy storage electricity market is the cumulative sum over time of the products of electricity market transaction prices for each period and the energy storage charging and discharging power. The revenue from the energy storage frequency regulation market is the cumulative sum over time of the products of frequency regulation market transaction prices for each period and the energy storage frequency regulation capacity. The revenue from the energy storage standby market is the cumulative sum over time of the products of spinning standby market transaction prices for each period and the energy storage standby capacity. The energy storage lifecycle amortized cost includes energy storage power investment cost, energy storage capacity investment cost, and energy storage operation and maintenance cost, which is converted to a daily amortized cost based on the number of days in the energy storage lifecycle. The energy storage cycle loss cost is related to the energy storage state of charge and the depth of charge and discharge, and the impact of cycle charge and discharge on energy storage lifespan degradation is quantified using an accelerated aging coefficient. The accelerated aging coefficient of energy storage is the ratio of the cumulative actual number of energy storage cycles to the rated number of energy storage cycles. When the energy storage state of charge is less than or equal to 0.4, the accelerated aging coefficient is calculated as -0.75 times the energy storage state of charge plus 1.3. When the energy storage state of charge is greater than 0.4, the accelerated aging coefficient is calculated as -0.25 times the energy storage state of charge plus 1.1.

[0017] The wind power revenue is the product of the peak-hour electricity market transaction price and the actual wind power output. The wind power frequency regulation revenue is the cumulative sum over time of the product of the frequency regulation market transaction price and the wind power frequency regulation capacity. The wind power reserve revenue is the cumulative sum over time of the product of the spinning reserve market transaction price and the wind power reserve capacity. The wind curtailment penalty cost is the product of the wind curtailment penalty coefficient and the amount of wind power curtailed, and the wind curtailment penalty coefficient is dynamically adjusted and determined by the power trading center based on the system's peak-shaving demand.

[0018] The operating costs of thermal power plants include the electricity market costs, frequency regulation market costs, and spinning reserve market costs. The operating costs of energy storage are the negative of the objective function of the energy storage operator. The operating costs of wind power and photovoltaic power include the electricity market costs, frequency regulation market costs, and reserve market costs for wind and photovoltaic power, respectively.

[0019] The Copula-CVaR risk quantification model uses kernel density estimation to fit the probability distributions of wind power output and photovoltaic output. It estimates the probability density functions of historical wind power output and historical photovoltaic output data sequences to obtain the marginal probability distribution functions of wind power output and photovoltaic output. An optimal Copula function family is selected using an information entropy minimization criterion. This criterion calculates the information entropy values ​​of the Gaussian Copula function, t-Copula function, and Clayton-Copula function, selecting the Copula function with the smallest information entropy value as the optimal Copula function family. The Copula function parameters of the optimal Copula function family are determined using a maximum likelihood estimation algorithm. The conditional value-at-risk (VAT) method is used to calculate the expected loss value at a given confidence level. This expected loss value is the conditional expectation of the loss exceeding the VAT threshold. The return loss risk value is calculated by integration, where the return loss exceeds the risk threshold, which is determined by the risk aversion coefficient and the confidence level parameter.

[0020] The central inertia equivalent transformation method equates a multi-generator system containing multiple generator sets to a single-generator infinite-scale equivalent system. The equivalent inertia constant of the single-generator infinite-scale equivalent system is the capacity-weighted average of the inertia constants of each generator set, with the weight of the capacity-weighted average being the proportion of the rated capacity of each generator set to the total installed capacity of the system. The rotor motion frequency response equation describes the dynamic response relationship between the system frequency deviation and the power imbalance, where the power imbalance is equal to the difference between the total generated power, the total load power, and the system loss power. The frequency response linearization transformation performs a Taylor series expansion of the rotor motion frequency response equation near the system's operating point, retaining the first derivative term of the Taylor series expansion to obtain a linearized frequency response algebraic equation, transforming the differential equation form constraints into algebraic inequality form constraints. The frequency deviation algebraic inequality constraint includes an upper limit constraint on the rate of change of frequency and a dead zone constraint on frequency deviation. The upper limit constraint on the rate of change of frequency is set according to the power grid frequency safety operation standard. The dead zone constraint on frequency deviation is a frequency deviation range threshold for which frequency regulation resources do not respond. When the absolute value of the system frequency deviation exceeds the frequency deviation range threshold, the frequency regulation resources respond and adjust according to the frequency deviation proportional coefficient.

[0021] The energy storage charging and discharging power constraints include upper limit constraints on energy storage charging power, upper limit constraints on energy storage discharging power, mutual exclusion constraints on energy storage charging and discharging states, and discrete variable constraints on energy storage charging and discharging states. The upper limit constraint on energy storage charging power states means that the energy storage charging power is no greater than the product of the energy storage charging state variable and the maximum energy storage charging power. The upper limit constraint on energy storage discharging power states means that the energy storage discharging power is no greater than the product of the energy storage discharging state variable and the maximum energy storage discharging power. The mutual exclusion constraint on energy storage charging and discharging states means that the sum of the energy storage charging state variable and the energy storage discharging state variable is no greater than 1. The discrete variable constraints on energy storage charging and discharging states mean that the energy storage charging state variable and the energy storage discharging state variable take values ​​of 0 or 1.

[0022] The power frequency regulation capacity constraint means that the power frequency regulation capacity is not greater than the power maximum frequency regulation capacity, and the power reserve capacity constraint means that the power reserve capacity is not greater than the power maximum reserve capacity. The power sources include energy storage, wind power, and photovoltaic power.

[0023] The power balance constraint means that the sum of thermal power output, net energy storage output, and wind power output in each time period equals the load demand power, and the net energy storage output is the difference between energy storage discharge power and energy storage charging power.

[0024] The frequency regulation demand constraint means that the sum of the frequency regulation capacity of thermal power, energy storage, wind power and photovoltaic power in each time period is not less than the frequency regulation demand capacity. The frequency regulation demand capacity is the product of the frequency regulation demand coefficient and the net load power. The net load power is the difference between the load demand power and the wind power output power and the photovoltaic power output power.

[0025] The reserve demand constraint means that the sum of the frequency regulation capacity and the total reserve capacity in each time period is not less than the total reserve demand capacity. The total reserve demand capacity includes the reserve demand capacity of renewable energy and the reserve demand capacity of load. The reserve demand capacity of renewable energy is the product of the reserve demand coefficient and the total output power of renewable energy, and the reserve demand capacity of load is the product of the frequency regulation demand coefficient and the net load power.

[0026] The mixed-integer second-order cone relaxation transformation transforms non-convex quadratic constraints into second-order cone convex constraints by introducing auxiliary variables and rotating cone constraints. The branch-and-bound search algorithm divides the feasible region of integer variables into multiple sub-feasible regions through branching operations, calculates the upper and lower bounds of the objective function for each sub-feasible region through bounding operations, and gradually narrows the search space until the optimal solution is found. The improved whale population optimization algorithm uses an average entropy population initialization method to improve population diversity, employs a dynamic relaxation constraint handling method to address constraint violations, and applies chaotic mapping perturbation to the optimal candidate solution to avoid getting trapped in local optima. The improved whale population optimization algorithm generates random values ​​between 0 and 1. When the random value is greater than or equal to 0.5, it updates offspring individuals based on the random population solution; when the random value is less than 0.5, it updates offspring individuals based on the optimal population solution. By adaptively adjusting the balance between global exploration capability and local exploitation capability, it improves the algorithm's convergence speed and solution accuracy.

[0027] The active power setpoints for each power source include active power setpoints for thermal power units, energy storage devices, wind farms, and photovoltaic power plants. The power source controllers include thermal power unit controllers, energy storage device controllers, wind farm controllers, and photovoltaic power plant controllers. The thermal power unit controller adjusts the opening of the turbine speed control valve of the thermal power unit according to the active power setpoint, thereby adjusting the active power output of the generator by changing the steam flow rate, so that the generator's active power output tracks the active power setpoint. The energy storage converter power command includes active power command and reactive power command. The energy storage device controller adjusts the active power output of the energy storage converter according to the active power command, and regulates the charging and discharging power of the energy storage battery by controlling the conduction angle of the power switch tube of the energy storage converter. The wind farm controller adjusts the pitch angle and generator torque of the wind turbine generator set according to the active power setpoint of the wind farm, and regulates the active power output of the wind farm by changing the wind turbine capture power and the generator electromagnetic torque. The photovoltaic power station controller adjusts the maximum power point tracking reference value of the photovoltaic inverter according to the active power setpoint of the photovoltaic power station, and regulates the active power output of the photovoltaic power station by changing the operating point of the photovoltaic array. The frequency regulation controller capacity parameters include the frequency regulation controller up-adjustment capacity parameters and the frequency regulation controller down-adjustment capacity parameters. The frequency regulation controller generates frequency regulation power adjustment commands based on the system frequency deviation and the frequency regulation controller capacity parameters. The standby response controller capacity parameter is the maximum available standby power after receiving a standby start signal. When the standby response controller receives a standby start signal, it rapidly increases the power output according to the standby response controller capacity parameters. The power control accuracy threshold is the maximum allowable deviation value of the power supply active power output. The power control accuracy threshold ranges from 0.5% to 2% of the power supply rated power.

[0028] As an optional implementation, the present invention also provides a method for forming a power output coordination control system for a joint electricity market transaction decision model by means of a computer, wherein the computer is provided with a readable storage medium, the readable storage medium stores program instructions, and the program instructions execute the above-mentioned power output coordination control method for the joint electricity market transaction decision model when the computer is run.

[0029] The specific implementation methods of the above steps are described in detail below.

[0030] The specific implementation of step S1 is as follows: First, a joint trading framework is established at the market layer. This framework integrates the three trading markets: the electricity market, the frequency regulation market, and the spinning reserve market. The electricity trading center achieves centralized management of market information and price discovery functions, providing a unified trading platform for different types of electricity ancillary services. Then, a game theory decision-making module is constructed at the bidding layer. This module includes decision-making units for grid operators, energy storage operators, wind power operators, and photovoltaic operators. Stackelberg game theory is used to model the relationship between the market layer and the bidding layer. The electricity trading center, as the leader, first releases market price signals, and each operator, as… Followers make bidding decisions based on price signals. Within the bidding layer, energy storage operators, wind power operators, and photovoltaic operators are modeled using Nash game theory. Each operator makes decisions simultaneously and influences the others. Finally, an execution control module is established at the scheduling layer. This module includes thermal power unit controllers, energy storage device controllers, wind farm controllers, and photovoltaic power station controllers. By collecting real-time operating status parameters, power generation cost parameters, and power generation plan parameters and reporting them to the bidding layer, a closed-loop feedback of information between the decision-making layer and the execution layer is achieved. The purpose of the three-layer architecture is to decouple market transaction decisions, game optimization decisions, and physical execution control hierarchically, reduce system modeling complexity, and improve solution efficiency.

[0031] The specific implementation of step S2 is as follows: First, the time axis is divided into three scales according to the characteristics of different operating stages of the electricity market. The day-ahead clearing stage uses a 1-hour time granularity for discretization, which meets the accuracy requirements for day-ahead market clearing calculations. The real-time dispatching stage uses a 15-minute time granularity for discretization, which matches the dispatching cycle of the current electricity spot market. The frequency regulation response stage uses a 1-second time granularity for discretization, which captures the dynamic changes in system frequency. Then, a cross-scale state transfer matrix is ​​constructed, which describes the mapping relationship between system state variables at different time granularities. The system uses the power output obtained from the day-ahead clearing as the power boundary constraint for real-time scheduling, and the power setting benchmark value obtained from real-time scheduling as the power benchmark reference value for frequency regulation response. Through this transfer relationship, a coupling constraint between long-term decision variables and short-term boundary conditions is established. Finally, a rolling time-domain optimization strategy is adopted, in which the values ​​of decision variables for future periods are updated according to the latest system state and prediction information within each optimization time window. Dynamic optimization is achieved by continuously rolling the optimization window forward. The purpose of this strategy is to ensure the temporal consistency of decision results at different time granularities and avoid conflicts between long-term decision-making and short-term execution.

[0032] The specific implementation of step S3 is as follows: First, a power node topology weighted graph is established with each power node as a vertex. The electrical distance and power transmission sensitivity between power nodes are calculated, and the two are combined to determine the edge weight of the graph. The larger the edge weight value, the stronger the electrical coupling strength between nodes. Then, a spectral clustering segmentation algorithm is used to partition the power node topology weighted graph. This algorithm first calculates the Laplacian matrix of the graph, which is defined as the difference between the degree matrix and the adjacency matrix. Then, the Laplacian matrix is ​​decomposed into eigenvalues ​​to obtain several minimum eigenvalues ​​and their corresponding eigenvectors. The first k eigenvectors are selected to form a feature matrix. Each row of the feature matrix is ​​regarded as a sample point. The sample points are divided into k clusters using the k-means clustering algorithm. Each cluster corresponds to a weak coupling control partition. This partition satisfies the characteristic that the electrical coupling strength between nodes within the partition is greater than the electrical coupling strength between nodes in the interval. Then, in each weak... Within the coupled control zone, a centralized model predictive control algorithm is used to optimize the power output of the power source within the zone. The prediction time domain length of this algorithm is set to 15 minutes, and the control time domain length is set to 5 minutes. The optimal control sequence within the zone is solved through rolling optimization. The zones are coordinated through a boundary power consistency protocol. This protocol uses an iterative calculation method, in which each zone exchanges boundary power information in each iteration. The boundary power exchange information is gradually converged to a globally consistent state through the Lagrange multiplier update method, thereby ensuring that the global power balance constraint is satisfied. Finally, the degree of change in the system topology is evaluated by monitoring the change in the distribution of the Laplace matrix eigenvalues. When the change in the distribution of eigenvalues ​​exceeds a set threshold, the threshold reference value can be set to 0.15, and the spectral clustering segmentation algorithm is re-executed to dynamically adjust the boundary of the weakly coupled control zone. The purpose of this dynamic adjustment mechanism is to enhance the system's adaptability to changes in the grid topology.

[0033] The specific implementation of step S4 is as follows: First, an objective function for the energy storage operator is established. This function consists of the sum of three revenues: energy storage electricity market revenue, energy storage frequency regulation market revenue, and energy storage reserve market revenue, minus two cost components: the energy storage lifecycle amortized cost and the energy storage cycle loss cost. Specifically, the energy storage electricity market revenue is calculated by summing the products of the electricity market transaction price for each time period and the corresponding energy storage charging and discharging power. The energy storage frequency regulation market revenue is calculated by summing the products of the frequency regulation market transaction price for each time period and the energy storage frequency regulation capacity. The energy storage reserve market revenue is calculated by summing the products of the spinning reserve market transaction prices for each time period. The total lifecycle cost of energy storage, obtained by multiplying and summing the energy storage reserve capacity, includes the investment cost of energy storage power, the investment cost of energy storage capacity, and the operation and maintenance cost of energy storage. This cost is converted to a daily cost allocation based on the number of days in the energy storage lifecycle. The energy storage cycle loss cost is quantified using an accelerated aging factor, defined as the ratio of the cumulative actual number of cycles to the rated number of cycles. When the energy storage state of charge (SOC) is less than or equal to 0.4, the accelerated aging factor is calculated as -0.75 times the SOC plus 1.3 (empirical value). When the SOC is greater than 0.4, the accelerated aging factor is -0.25 times the SOC plus 1.1. (Empirical value) Calculation: This piecewise function design can reflect the accelerated impact of deep discharge on energy storage lifespan degradation. Then, a wind power operator objective function is established, consisting of the sum of wind power revenue, wind power frequency regulation revenue, and wind power reserve revenue minus the curtailment penalty cost. Wind power revenue is the product of the peak-hour electricity market transaction price and the actual wind power output. The curtailment penalty cost is the product of the curtailment penalty coefficient and the curtailed wind power volume. The curtailment penalty coefficient is dynamically adjusted by the power trading center based on system peak-shaving demand. Next, a grid operator objective function is established, which is the sum of thermal power operating costs, energy storage operating costs, wind power operating costs, and photovoltaic operating costs. The energy storage operating cost is the negative of the energy storage operator's objective function. Finally, the Copula-CVaR risk quantification model is introduced. This model first uses kernel density estimation to estimate the probability density function of historical wind power output data sequences and historical photovoltaic output data sequences, obtaining the marginal probability distribution function of wind power output. The model first calculates the marginal probability distribution function of photovoltaic output, then selects the optimal Copula function family using the information entropy minimization criterion. The information entropy values ​​of the Gaussian Copula function, t-Copula function, and Clayton-Copula function are calculated respectively, and the function with the smallest information entropy value is selected as the optimal Copula function family. The parameters of this function family are determined using the maximum likelihood estimation algorithm. The expected loss value at a given confidence level is calculated using the conditional value-at-risk (VAT) method. This value represents the conditional expectation of the loss exceeding the VAT threshold. The expected loss when revenue exceeds the risk threshold is calculated through integration and incorporated as the revenue loss risk value into the objective function of each market participant. The risk threshold is jointly determined by the risk aversion coefficient and the confidence level parameter, with a reference value of 0.95. The purpose of this risk quantification model is to quantitatively assess the revenue loss risk caused by the uncertainty of renewable energy output and reflect it in optimization decision-making.

[0034] The specific implementation of step S5 is as follows: First, the multi-machine system containing multiple generator sets is equivalent to a single-machine infinite equivalent system using the central inertia equivalent transformation method. The inertia constant of the equivalent system is obtained by calculating the capacity-weighted average of the inertia constants of each generator set, with the weight being the proportion of the rated capacity of each generator set to the total installed capacity of the system. The purpose of this equivalent transformation is to simplify the complex multi-machine system into a single-machine system to reduce modeling complexity. Then, a rotor motion frequency response equation is established to describe the dynamic response relationship between the system frequency deviation and the power imbalance. The power imbalance is equal to the difference between the total generated power, the total load power, and the system loss power. This equation is in the form of a second-order differential equation. The frequency response equation of rotor motion is linearized by performing a frequency response linearization transformation. A Taylor series expansion is then performed on the equation near the system's operating point, retaining the first derivative term to obtain the linearized frequency response algebraic equation. The differential equation constraints are then transformed into algebraic inequality constraints. These transformed frequency deviation algebraic inequality constraints include an upper limit constraint on the rate of change of frequency and a dead zone constraint on frequency deviation. The upper limit constraint on the rate of change of frequency is set according to the power grid frequency safety operation standard, with a reference value of 0.5 Hz / s. The dead zone constraint on frequency deviation is the threshold value for the frequency deviation range from which frequency regulation resources do not respond, with a reference value of 0.03 Hz. When the absolute value of the system frequency deviation exceeds this threshold, the frequency regulation resources... The frequency deviation proportional coefficient is adjusted accordingly. Finally, a multi-level constraint set is established, including energy storage charging and discharging power constraints, power supply frequency regulation capacity constraints, power supply reserve capacity constraints, power balance constraints, frequency regulation demand constraints, and reserve demand constraints. Energy storage charging and discharging power constraints include upper limits for energy storage charging power, upper limits for energy storage discharging power, mutual exclusion constraints for energy storage charging and discharging states, and discrete variable constraints for energy storage charging and discharging states. The mutual exclusion constraint for charging and discharging states indicates that the sum of the energy storage charging state variables and the energy storage discharging state variables is no greater than 1. The discrete variable constraint indicates that the state variables take values ​​of 0 or 1. The power supply frequency regulation capacity constraints and power supply reserve capacity constraints respectively limit the frequency regulation capacity and reserve capacity of each power supply to no more than [a certain value]. The power balance constraint means that the sum of thermal power output, net energy storage output, and wind power output in each time period equals the load demand power. The frequency regulation demand constraint means that the sum of the frequency regulation capacity of all power sources in each time period is not less than the frequency regulation demand capacity, which is the product of the frequency regulation demand coefficient and the net load power. The reference value of the frequency regulation demand coefficient can be set to 0.01. The reserve demand constraint means that the sum of the frequency regulation capacity and the sum of the reserve capacity in each time period is not less than the total reserve demand capacity, which includes the reserve demand capacity of renewable energy and the reserve demand capacity of the load. The purpose of these constraints is to ensure that the system meets the requirements of electricity market transactions while also meeting the requirements of frequency stability and supply and demand balance.

[0035] The specific implementation of step S6 is as follows: First, the non-convex quadratic constraints in the model are handled using a mixed integer second-order cone relaxation transformation. By introducing auxiliary variables and rotating cone constraints, the non-convex quadratic constraints are transformed into second-order cone convex constraints. This transformation method is based on convex optimization theory and can transform nonlinear optimization problems into convex optimization problems, thereby improving solution efficiency. Then, a branch and bound search algorithm is used to narrow the search space for integer variables. This algorithm divides the feasible region of integer variables into multiple sub-feasible regions through branch operations. The upper and lower bounds of the objective function for each sub-feasible region are calculated through bound operations, gradually narrowing the search space until the optimal solution is found. Next, an improved whale population optimization algorithm is used to generate a high-quality initial solution. This algorithm uses the average entropy population initialization method to improve the population diversity index. The uniform distribution of individuals in the population in the solution space can enhance the global search capability. Finally, a dynamic relaxation constraint handling method is used to handle constraint violations. To avoid getting trapped in local optima, the algorithm performs chaotic mapping perturbation on the optimal candidate solution. During execution, it generates random values ​​between 0 and 1. When the random value is greater than or equal to 0.5, it updates the offspring individuals based on the random solution of the population to enhance global exploration capabilities. When the random value is less than 0.5, it updates the offspring individuals based on the optimal solution of the population to enhance local exploitation capabilities. By adaptively adjusting the balance between global exploration and local exploitation, the algorithm's convergence speed and solution accuracy are improved. Finally, the upper-level Cplex solver is combined with the lower-level improved whale population optimization algorithm. The Cplex solver is used to solve the transformed convex optimization problem, while the improved whale population optimization algorithm is used to optimize discrete variables and generate initial solutions. Through two-level collaborative optimization, the market transaction prices, power output of each power source, energy storage charging and discharging power, frequency regulation capacity configuration value, and reserve capacity configuration value are obtained. The purpose of this hybrid optimization solution strategy is to balance solution accuracy and computational efficiency.

[0036] The specific implementation of step S7 is as follows: First, the output power of each power source obtained from the solution is used as the active power setpoint for each power source. These setpoints include the active power setpoints for thermal power units, energy storage devices, wind farms, and photovoltaic power plants. These setpoints are then sent to the corresponding power controllers via a communication network. Next, the energy storage charging and discharging power is converted into energy storage converter power commands. These commands include active power commands and reactive power commands for the energy storage converter, and are sent to the energy storage converter. Finally, the frequency regulation capacity configuration value is converted into frequency regulation controller capacity parameters. These parameters include… This includes setting the capacity parameters for both the frequency regulation controller (FM controller) and the standby response controller (SRC controller). The standby capacity configuration value is then converted into a standby response controller capacity parameter, representing the maximum available standby power after receiving a standby start signal. This parameter is also set to the SRC controller. Next, each power controller adjusts the active power output of its corresponding power source based on the received active power setpoint. The thermal power unit controller adjusts the turbine speed control valve opening based on the thermal power unit's active power setpoint, thereby adjusting the generator's active power output by changing the steam flow rate. Energy storage equipment... The controller adjusts the active power output of the energy storage converter according to the active power command of the energy storage converter, and regulates the charging and discharging power of the energy storage battery by controlling the conduction angle of the power switch tube of the energy storage converter. The wind farm controller adjusts the pitch angle and generator torque of the wind turbine according to the active power setpoint of the wind farm, and regulates the active power output of the wind farm by changing the wind turbine capture power and generator electromagnetic torque. The photovoltaic power station controller adjusts the maximum power point tracking reference value of the photovoltaic inverter according to the active power setpoint of the photovoltaic power station, and regulates the active power output of the photovoltaic power station by changing the operating point of the photovoltaic array. The frequency modulation controller generates frequency modulation power adjustment commands according to the system frequency deviation and the frequency modulation controller capacity parameters. When the standby response controller receives the standby start signal, it quickly increases the power output according to the standby response controller capacity parameters. Finally, each power controller monitors the deviation between the active power output of the power supply and the set value in real time. When the deviation is less than the power control accuracy threshold, the control process ends. This threshold is the maximum allowable deviation value of the active power output of the power supply. The reference value can be set to 1% of the rated power of the power supply. The function of this hierarchical control mechanism is to transform the optimization decision results into control commands for physical equipment and achieve precise tracking control.

[0037] It should be noted that this invention also solves the following technical problem: the difficulty in ensuring the timing consistency between multi-timescale market trading decisions and power supply execution control. Power market trading involves decision-making processes at different timescales, such as day-ahead clearing, real-time dispatch, and frequency regulation response. Traditional methods optimize decisions at each timescale independently, lacking an effective state transfer mechanism. This results in long-term market trading results failing to provide reasonable boundary constraints for short-term execution control, and the actual operating status at the short-term timescale failing to be promptly fed back to long-term decision adjustments, causing significant deviations between market trading plans and actual execution control. This invention establishes a multi-timescale coordinated optimization framework, employs a cross-scale state transfer matrix to describe the transfer relationship of system state variables at different time granularities, uses the power output obtained from day-ahead clearing as the constraint boundary for real-time dispatch, and uses the power setting benchmark value obtained from real-time dispatch as the reference benchmark for frequency regulation response. Through a rolling time-domain optimization strategy, the decision variable values ​​for future periods are updated within each optimization time window, forming a complete closed-loop transfer link from market trading to execution control, effectively solving the technical problem of inconsistent timing of multi-timescale decision results.

[0038] This invention addresses the technical challenge of excessive computational complexity in distributed coordinated control of large-scale power systems. With the continuous expansion of power system scale and the increasing proportion of renewable energy integration, the number of power nodes participating in market transactions has increased significantly. Traditional centralized optimization methods require simultaneous handling of the coupling constraints of all power nodes, leading to an exponential increase in the dimensionality of decision variables and the number of constraints, resulting in excessive computational complexity and difficulty in meeting real-time control requirements. Furthermore, it lacks adaptive adjustment capabilities when the power grid topology changes. This invention constructs a weighted graph of power node topology and uses electrical coupling strength as edge weights. It utilizes a spectral clustering segmentation algorithm to calculate the eigenvalues ​​and eigenvectors of the Laplace matrix, clustering power nodes with high similarity indices into the same weakly coupled control partition. Within each partition, a centralized model predictive control algorithm is used to optimize power output in parallel. Coordination between partitions is achieved through a boundary power consistency protocol. This decomposes the high-dimensional optimization problem into multiple low-dimensional subproblems for parallel solution, significantly reducing computational complexity and improving control response real-time performance while ensuring global power balance constraints are met. This effectively solves the technical problem of computational difficulties in coordinated control of large-scale power systems.

[0039] Specifically, the principle of this invention is as follows: This invention defines the accelerated aging coefficient of energy storage as the ratio of the cumulative actual number of energy storage cycles to the rated number of energy storage cycles. This ratio reflects the proportion of the energy storage's consumed cycle life. Combined with the energy storage's life-cycle allocated cost, the incremental cost of cycle loss caused by each charge-discharge operation is calculated. This cost quantification method transforms the physical process of energy storage electrochemical aging into an economically meaningful cost metric, allowing the energy storage life protection objective to be naturally integrated into the market transaction optimization framework through cost minimization. This invention incorporates the energy storage cycle loss cost as a negative term into the energy storage operator's objective function. During the optimization process, this cost term forms an inherent constraint with the market transaction revenue term. When energy storage attempts to obtain higher market revenue through frequent charge-discharge or deep cycling, the rapid increase in cycle loss cost automatically suppresses this trend. The optimization algorithm finds the optimal balance between revenue and cost, naturally avoiding the accelerated aging operating range. This mechanism of transforming physical constraints into economic constraints conforms to the mathematical principles of multi-objective optimization, ensuring that energy storage extends its actual lifespan while pursuing economic benefits.

[0040] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0041] The specific implementation method of step S1 is the same as described above, and will not be repeated here.

[0042] The specific implementation of step S2 involves establishing a multi-timescale coordination and optimization framework. The day-ahead clearing time window is discretized using a 1-hour time granularity, real-time scheduling is discretized using a 15-minute time granularity, and the frequency modulation response is discretized using a 1-second time granularity. Cross-scale state transfer matrix. The expression is as follows: ; In the formula, The dimensionless cross-scale state transfer matrix is ​​the one that extends from the day-to-day to the real-time point. For the first The day before the period to the The state transfer coefficient for each real-time period has a value range of 0 to 1 and is dimensionless. This represents the total number of days in advance, typically set to 24. This represents the total number of real-time time periods, typically set to 96. When long-term decision variables are passed as short-term boundary conditions, this represents the upper limit constraint on power output. and power output lower limit constraint The calculation formula is as follows: ; ; In the formula, For the first The power supply is in the first The upper limit of output power for each real-time period, in units of ; For the first The power supply is in the first The lower limit of output power for each real-time period, in units of ; For the first The power supply is in the first The power output for the period of the day before yesterday, in units of ; This is a power reference value, in units of The value is usually 100. ; To adjust the margin factor upwards, the empirical value is 0.1, which is dimensionless; To adjust the margin factor, an empirical value of 0.1 is used; this value is dimensionless. Power reference value. The calculation formula is as follows: ; In the formula, For the first The power supply is in the first The power reference value for each frequency modulation period, in units of ; For the first The real-time period is directed to the first The time weighting coefficient for each frequency modulation period ranges from 0 to 1, is dimensionless, and satisfies the following conditions: ; For the first The power supply is in the first The power setting reference value for each real-time period, in units of ; This is the frequency modulation time period number.

[0043] The specific implementation of step S3 is to construct a power node topology weighted graph and perform hierarchical coordination control partitioning, establishing a power node topology weighted graph with each power node as a vertex. ,in For the set of vertices, Let be the set of edges. This is the edge weight matrix. Edge weights Electrical coupling strength is used as the weight. The calculation formula is as follows: ; In the formula, For power nodes With power nodes The electrical coupling strength between them is dimensionless; This is the electrical distance normalization coefficient, which defaults to 1 and is dimensionless. For power nodes With power nodes The electrical distance between them, in ohms; For reference electrical distance, the unit is ohms, usually taken as 1 ohm; For power nodes Active power to power nodes The partial derivative of the phase angle, in units of radian; The reference phase angle is expressed in radians, and is typically set to 1 radian. For power nodes Rated capacity, in units of ; For power nodes Rated capacity, in units of ; This is the capacity baseline value, in units of The value is usually 100. Laplace matrix The calculation formula is as follows: ; In the formula, It is a Laplace matrix, dimensionless; A dimensionless degree matrix. diagonal elements Off-diagonal elements are 0; The edge weight matrix is ​​dimensionless. This represents the total number of power nodes. The Laplace matrix is ​​calculated by solving... eigenvalues and the corresponding feature vector , will go The eigenvectors corresponding to the smallest non-zero eigenvalues ​​constitute the feature matrix. , characteristic matrix The expression is as follows: ; In the formula, The characteristic matrix is ​​dimensionless. For the first The th eigenvector of the th feature vector One element, dimensionless; The number of partitions is controlled for weak coupling. For the feature matrix... After normalizing each row, the power nodes are clustered into groups using the K-means clustering algorithm. A loosely coupled control partition. The iterative formula for the boundary power consistency protocol is as follows: ; In the formula, For the first The weakly coupled control partition in the first The first boundary node The power exchange value of the next iteration, in units of ; This is the convergence coefficient of the consensus protocol, ranging from 0.01 to 0.1, and is dimensionless. In order to be with the first A set of weakly coupled control partitions adjacent to each other; For adjacent weakly coupled control partitions In the The first boundary node The power exchange value of the next iteration, in units of ; This is the sequence number of the weakly coupled control partition; The boundary node number; This represents the number of iterations. It also represents the change in the eigenvalue distribution. Exceeding the threshold of feature value change When the spectral clustering segmentation algorithm is re-executed, the change in the eigenvalue distribution is... The calculation formula is as follows: ; In the formula, The change in the distribution of eigenvalues ​​is dimensionless. For the current moment 1 eigenvalue, dimensionless; For the previous moment 1 eigenvalue, dimensionless; This is the threshold for the change in eigenvalues, typically set to 0.05, and is dimensionless.

[0044] The specific implementation of step S4 involves establishing objective functions for each market participant and introducing a risk quantification model, as well as the objective function for energy storage operators. The expression is as follows: ; In the formula, The objective function for energy storage operators is expressed in yuan. This is a cost reference value, in yuan, and is usually taken as 10,000 yuan; This represents the total number of time periods in the scheduling cycle. For the first Electricity market transaction prices for different time periods, in yuan. ; This is a price reference value, in yuan. The value is usually 100 yuan. ; The number of energy storage power stations; For the first The energy storage power station in the first Charging power during a period of time, in units of ; For the first The energy storage power station in the first Discharge power over a period of time, in units of ; The duration is expressed in hours. This is an energy reference value, in units of The value is usually 1. ; For the first Time-of-use frequency (TOU) market transaction prices, in yuan. ; For the first The energy storage power station in the first Frequency modulation capacity for a given time period, in units of ; For the first The trading price of the rotating reserve market during the time period is in yuan. ; For the first The energy storage power station in the first Reserve capacity for a given period, in units of ; For the first The total life-cycle cost of an energy storage power station, in yuan; For the first The energy storage power station in the first Cyclic loss cost for a given period, in yuan; This refers to the serial number of the energy storage power station; This refers to the time period number. The cost is amortized over the entire lifecycle of energy storage. The calculation formula is as follows: ; In the formula, The unit investment cost per unit of energy storage capacity is expressed in yuan. ; For the first The rated power of an energy storage power station, in units of ; The unit investment cost per unit of energy storage capacity is expressed in yuan. ; For the first The rated capacity of an energy storage power station, in units of ; This refers to the total lifespan of the energy storage system in years, typically ranging from 10 to 20 years. The unit cost for energy storage operation and maintenance is in yuan. Year ; The inflation rate, typically ranging from 0.02 to 0.05, is dimensionless. This is the discount rate, typically ranging from 0.05 to 0.08, and is dimensionless. For the first The total lifecycle of an energy storage power station, in days. Energy storage cycle loss cost. The calculation formula is as follows: ; ; In the formula, For the first The energy storage power station in the first The cyclic loss coefficient for a given time period is dimensionless. For the first The energy storage power station in the first The accelerated aging coefficient over a period of time is dimensionless. The rated number of energy storage cycles is typically between 3,000 and 8,000. The rated depth of discharge for energy storage, typically ranging from 0.8 to 0.9, is dimensionless. Accelerated aging factor for energy storage. The calculation formula is as follows: ; ; In the formula, For the first The energy storage power station in the first The state of charge during a given time period is dimensionless. For the first The energy storage power station in the first Energy storage capacity during a given period, in units of Objective function of wind power operators The expression is as follows: ; In the formula, The objective function for wind power operators is expressed in yuan. Peak electricity market transaction price, in yuan. ; For the first Actual wind power output during the period, in units of ; For the first Wind power frequency regulation capacity during specific time periods, in units of ; For the first wind power reserve capacity for a given period, in units of ; This is the wind curtailment penalty coefficient, in yuan. ; For the first Time-based wind curtailment power, in units of Objective function of power grid operators The expression is as follows: ; In the formula, This is a function representing the total cost of power grid operation, expressed in yuan. This refers to the number of thermal power units. For the first The thermal power unit was in the first Electricity market cost for the specified time period, in yuan; For the first The thermal power unit was in the first Output power over a given period, in units of ; For the first The thermal power unit was in the first The market cost of FM radio for a given time period, in yuan; For the first The thermal power unit was in the first Frequency modulation capacity for a given time period, in units of ; For the first The thermal power unit was in the first The market cost of rotating reserve during a given period, in yuan; For the first The thermal power unit was in the first Reserve capacity for a given period, in units of ; For the first Frequency regulation market transaction price for wind or solar power during specific time periods, in yuan. ; For the first Frequency regulation capacity of wind or solar power during a given time period, in units of ; For the first The market transaction price of spinning reserve for wind or solar power during a given period, in yuan. ; For the first Reserve capacity of wind or solar power during a given period, in units of ; This refers to the serial number of the thermal power unit. Conditional Value at Risk (VaR) in the Copula-CVaR risk quantification model. The calculation formula is as follows: ; In the formula, Conditional risk value, in yuan; Value at Risk (VaR) threshold, in yuan; This is a confidence level parameter, ranging from 0.9 to 0.99, and is dimensionless. For wind power output, the unit is [missing information]. ; The output of photovoltaic power is a random variable, and the unit is . ; For the set of uncertainties in wind power; For the set of uncertainties in photovoltaics; This is the profit / loss function, expressed in yuan. ; Let be the probability density function of wind power output, in units of . ; Let be the probability density function of photovoltaic power output, in units of . ; A Copula function, dimensionless; Let be the marginal probability distribution function of wind power output, which is dimensionless; Let be the edge probability distribution function of photovoltaic power output, which is dimensionless; These are decision variables.

[0045] The specific implementation of step S5 involves establishing a frequency dynamic stability constraint model and setting a multi-level constraint condition group. The multi-machine system is then converted into a single-machine equivalent system using the central inertia equivalent transformation method, with the equivalent inertia constant... The calculation formula is as follows: ; In the formula, is the equivalent inertia constant of a single-machine equivalent system, in seconds; For the first The rated capacity of each generator set, in units of ; Total installed capacity of the system, in units of , ; For the first The inertia constant of each generator set is expressed in seconds. The expression for the rotor motion frequency response equation is as follows: ; In the formula, The system frequency deviation is expressed in units of 1000 ppm. ; Time, in seconds; For reference time, the unit is seconds, and the value is usually 1 second; This represents the change in total power generation, in units of... ; This represents the change in total load power, in units of... ; The load damping coefficient, typically ranging from 1 to 2, is dimensionless. Through frequency response linearization transformation, a Taylor series expansion of the rotor motion frequency response equation is performed near the system's operating point. Retaining the first derivative term, the linearized frequency response algebraic equation is obtained as follows: ; In the formula, For the first The system frequency deviation at time t, in units of ; For the first The system frequency deviation at time t, in units of ; The time step is in seconds. For the first The change in total power generation at any given time, in units of ; For the first The change in total load power at time t, in units of The frequency deviation algebraic inequality constraint includes the upper limit constraint on the rate of change of frequency and the dead zone constraint on frequency deviation. The expression for the upper limit constraint on the rate of change of frequency is as follows: ; In the formula, The maximum rate of change of frequency, in units of According to the power grid frequency safety operation standards, the value is typically set between 0.5 and 2. ; The system's rated frequency, in units of The value is usually 50 or 60. The expression for the frequency deviation dead zone constraint is as follows: ; ; In the formula, Frequency deviation dead zone threshold, unit: The value is typically between 0.03 and 0.05. ; For the first Frequency modulation power at any given time, in units of ; This is the frequency deviation proportionality coefficient, which is dimensionless. This is a sign function; it takes a value of 1 when the parameter is greater than 0, -1 when it is less than 0, and 0 when it is equal to 0. Energy storage charging and discharging power constraints include upper limit constraints on energy storage charging power, upper limit constraints on energy storage discharging power, mutual exclusion constraints on energy storage charging and discharging states, and discrete variable constraints on energy storage charging and discharging states. The specific expressions are as follows: ; ; ; ; In the formula, For the first The energy storage power station in the first The charging state variable for a given period takes the value of 0 or 1 and is dimensionless. This represents the maximum charging power of the energy storage system, measured in units of... ; For the first The energy storage power station in the first The discharge state variable for a given period takes the value of 0 or 1 and is dimensionless. The maximum discharge power of the energy storage is expressed in units of 1. The expressions for the power supply frequency regulation capacity constraint and the power supply reserve capacity constraint are as follows: ; ; In the formula, For the first indivual Type of power supply in the first Frequency modulation capacity for a given time period, in units of ; for The maximum frequency regulation capacity of this type of power supply, in units of ; For the first indivual Type of power supply in the first Reserve capacity for a given period, in units of ; for Maximum standby capacity of this type of power supply, in units of ; This represents energy storage, wind power, and photovoltaics. The expression for the power balance constraint is as follows: ; In the formula, For the first Load demand power for a given period, in units of The expression for the frequency modulation demand constraint is as follows: ; In the formula, For the first Photovoltaic frequency regulation capacity during the time period, in units of ; This is the frequency regulation demand factor, typically ranging from 0.01 to 0.03, and is dimensionless. For the first Photovoltaic power output per time period, in units of The expression for the spare requirement constraint is as follows: ; In the formula, This is the reserve demand factor, typically ranging from 0.05 to 0.15, and is dimensionless.

[0046] The specific implementation of step S6 involves using a hybrid optimization solution strategy to solve the joint electricity market trading decision model, and transforming the non-convex quadratic constraints into second-order conical convex constraints using a mixed integer second-order cone relaxation transformation. For the non-convex quadratic constraints... Introducing auxiliary variables and The rotational cone constraint condition is transformed into a second-order cone convex constraint form as follows: ; In the formula, This is the first decision variable; This is the second decision variable; To constrain the upper limit variable; for Reference value; for Reference value; for Reference value; It is a norm 2. The branch and bound search algorithm divides the feasible region of the integer variable into multiple sub-feasible regions through branching operations, and calculates the upper bound of the objective function for each sub-feasible region through bounding operations. and the lower bound of the objective function The update formulas for the upper and lower bounds are as follows: ; ; In the formula, Let be the upper bound of the objective function of the subfeasible region; The objective function value of the current optimal solution; Let be the lower bound of the objective function for the subfeasible region; For the feasible region; The value of the objective function after relaxation; For decision variable vectors; The reference value for the objective function is used. The improved whale population optimization algorithm employs an average entropy population initialization method to enhance population diversity. The average entropy of the population... The calculation formula is as follows: ; In the formula, The population average entropy is dimensionless. This represents the population size, typically ranging from 50 to 100. For decision variables; For the first The first individual Dimensional decision variables; For the first Lower bound of the decision variable; For the first Upper bound of the decision variable; For individual serial numbers; This refers to the dimension index of the decision variable. The dynamic relaxation constraint handling method uses the constraint violation degree... Quantify constraint violations and the degree of constraint violation. The calculation formula is as follows: ; In the formula, To constrain the degree of violation, it is dimensionless; The number of equality constraints; For the first One equality constraint function; For the first Reference values ​​for each equality constraint; The number of inequality constraints; For the first Inequality constraint functions; For the first Reference values ​​for each inequality constraint; The number represents the equality constraint. Inequality constraint numbers are used. To avoid getting trapped in local optima, a chaotic mapping perturbation is applied to the optimal candidate solution. The chaotic mapping uses a Logistic mapping, the expression of which is as follows: ; In the formula, For the first The chaotic variable in the next iteration has a value range of 0 to 1 and is dimensionless. This is a chaos parameter, typically taken as 4, and is dimensionless. The iteration count is used to generate random values ​​between 0 and 1 in an improved whale population optimization algorithm. When random values When the value is greater than or equal to 0.5, the offspring individuals are updated based on the random solution of the population. The update formula is as follows: ; In the formula, For the first The offspring solutions of each individual; An individual randomly selected from the population; This serves as the lower bound for the decision variable. This serves as the upper bound for the decision variable; The global exploration coefficient is dimensionless. ; The parameter is linearly decreasing and dimensionless. ; A dimensionless random number between 0 and 1; These are weighting coefficients, dimensionless. ; A dimensionless random number between 0 and 1; For the first The current solution for each individual; This represents the current iteration number; This represents the maximum number of iterations. When the random value... When the value is less than 0.5, the offspring individuals are updated based on the population optimal solution. The update formula is as follows: ; In the formula, The normalized distance between the current individual and the optimal solution is dimensionless. The optimal solution for the population; is the helical coefficient, usually taken as 1, and is dimensionless; A dimensionless random number between -1 and 1. The algorithm's convergence speed and solution accuracy are improved by adaptively adjusting the balance between global exploration capability and local exploitation capability. The global exploration capability index... and local development capability indicators The calculation formula is as follows: ; ; In the formula, This is a dimensionless indicator of overall exploration capability. The location of the population center ; This is a dimensionless indicator of local development capability. For the first The optimal solution in the next iteration; For the first The optimal solution in the next iteration.

[0047] The specific implementation of step S7 involves outputting power supply output coordination control commands and executing hierarchical control. The output power of each power source is sent as its active power setpoint to the corresponding power controller. The energy storage charging and discharging power is converted into an energy storage converter power command and sent to the energy storage converter. The frequency regulation capacity configuration value is converted into a frequency regulation controller capacity parameter setting and sent to the frequency regulation controller. The standby capacity configuration value is converted into a standby response controller capacity parameter setting and sent to the standby response controller. Each power controller adjusts the active power output of its corresponding power source to the target power value based on its active power setpoint. The deviation between the target power value and the active power setpoint of each power source is less than the power control accuracy threshold. Power control accuracy threshold The value ranges from 0.5% to 2% of the rated power of the power supply. The thermal power unit controller is based on the active power setpoint of the thermal power unit. Adjusting the opening of the turbine speed control valve of the thermal power unit Speed ​​control valve opening The adjustment formula is as follows: ; In the formula, For the first The opening degree of the speed control valve at any given time, expressed as a percentage; For the first The opening degree of the speed control valve at any given time, expressed as a percentage; This is the speed control valve opening adjustment coefficient, typically ranging from 0.01 to 0.05, and is dimensionless. For the first The thermal power unit was in the first The active power setpoint at any given time, in units of ; For the first The thermal power unit was in the first Actual active power output at any given time, in units of ; For the first The rated power of each thermal power unit, in units of The energy storage device controller operates based on the active power command from the energy storage converter. Adjusting the active power output of the energy storage converter is achieved by controlling the conduction angle of the power switch transistors in the energy storage converter. To achieve regulation of the charging and discharging power of the energy storage battery, and the conduction angle The calculation formula is as follows: ; In the formula, For the first The conduction angle of the power switch at any given moment, in radians; DC side voltage, unit: ; Rated current, unit is ; The power factor is typically between 0.95 and 1 and is dimensionless. Rated power of the energy storage converter, in units of ; Pi, approximately 3.14159. The wind farm controller uses the active power setpoint of the wind farm. Adjusting the pitch angle of the wind turbine generator set and generator torque Pitch angle The adjustment formula is as follows: ; In the formula, For the first The pitch angle at any given moment, in degrees; For the first The pitch angle at any given moment, in degrees; This is the pitch angle adjustment coefficient, typically ranging from 0.1 to 0.5, and is dimensionless. For the first The actual active power output of the wind farm at any given time, in units of ; Rated power of the wind farm, in units of The photovoltaic power plant controller is based on the active power setpoint of the photovoltaic power plant. Adjusting the maximum power point tracking reference value of the photovoltaic inverter Maximum power point tracking reference value The calculation formula is as follows: ; In the formula, For the first Maximum power point tracking reference value at time , in units of ; This is the open-circuit voltage of the photovoltaic array, in units of... ; For the first Maximum available power of photovoltaic power station at any time, in units of ; This is the power point tracking factor, typically ranging from 0.8 to 1.2, and is dimensionless. Frequency modulation controller capacity parameters include the frequency modulation controller's up-regulation capacity parameters. And frequency modulation controller to adjust capacity parameters The frequency modulation controller adjusts according to the system frequency deviation. Generate frequency modulation power adjustment commands based on the frequency modulation controller capacity parameters. Frequency modulation power adjustment command The calculation formula is as follows: ; In the formula, For the first Frequency modulation power adjustment command at any time, in units of ; Adjust the capacity parameter of the frequency modulation controller, in units of ; Adjust the capacity parameter of the frequency modulation controller, in units of Backup response controller capacity parameters The standby response controller represents the maximum available standby power after receiving a standby start signal. Upon receiving the standby start signal, the standby response controller rapidly increases its power output according to its capacity parameters. The rate of increase in power output is... The calculation formula is as follows: ; In the formula, Rate of increase in backup power, in units of ; This refers to the standby response controller capacity parameter, in units of... ; This is the backup response ramp-up time, measured in seconds, typically ranging from 10 to 30 seconds.

[0048] To better understand and implement this invention, a specific application scenario of the invention is provided below as Example 2: To verify the effectiveness of the invention, technicians built a numerical simulation analysis environment. By constructing a regional power grid model including thermal power units, energy storage systems, wind farms, and photovoltaic power plants, simulation tests were conducted on the power output coordination control method of the joint electricity market trading decision model. The simulation system sets the total installed capacity to be 850MW, of which thermal power capacity is 400MW, energy storage system capacity is 100MW / 200MWh, wind power capacity is 250MW, and photovoltaic capacity is 100MW. The system's day-ahead load demand curve exhibits a typical bimodal characteristic, with peak loads occurring at 10:00 AM and 7:00 PM, reaching 682MW and 715MW respectively, while the nighttime off-peak load drops to 328MW.

[0049] First, technicians construct a joint electricity market trading decision-making model based on step S1, establishing a market-level joint trading framework and setting trading rules for the electricity market, frequency regulation market, and spinning reserve market. The electricity trading center sets the day-ahead electricity market price benchmark at 0.52 yuan / kWh, the frequency regulation market price at 12.8 yuan / MW, and the spinning reserve market price at 8.6 yuan / MW based on historical trading data. The bidding layer game decision-making module includes decision-making units for grid operators, energy storage operators, wind power operators, and photovoltaic operators, forming a Stackelberg-Nash hybrid game relationship among these units. The dispatch layer execution control module reports operating status parameters to the bidding layer: the unit generation cost of thermal power units is 0.38 yuan / kWh, the energy storage system's charge / discharge efficiency is 92%, and the marginal cost of wind farms and photovoltaic power plants is close to zero.

[0050] Following step S2, a multi-timescale coordinated optimization framework is established. The 24-hour day-ahead clearing time window is discretized using a 1-hour time granularity, resulting in 24 decision variables for different time periods. Real-time scheduling uses a 15-minute time granularity, subdividing each hour into 4 scheduling periods, for a total of 96 real-time scheduling periods. Frequency regulation response uses a 1-second time granularity, performing dynamic power adjustment within a 900-second frequency regulation cycle. A cross-scale state transfer matrix establishes the transfer relationship between day-ahead clearing power as a boundary condition for real-time scheduling and the transfer path between the real-time scheduling power benchmark value and the frequency regulation response reference value. A rolling time-domain optimization strategy updates the decision variables for the next 4 hours within each optimization window, ensuring a 98.7% satisfaction rate of time-series consistency constraints.

[0051] Technicians performed step S3 to construct a topology weighted graph of the power nodes, establishing a topology structure with the 12 main power nodes in the system as vertices. The edge weight distribution of the topology weighted graph was determined by calculating the electrical distance and power transmission sensitivity between the power nodes, as shown in Table 1. A spectral clustering segmentation algorithm was used to calculate the eigenvalues ​​and eigenvectors of the Laplacian matrix, dividing the 12 power nodes into three weakly coupled control partitions. The first control partition includes nodes 1, 2, 3, and 4, mainly composed of thermal power units and energy storage systems. The second control partition includes nodes 5, 6, 7, and 8, with wind farms as the main power source. The third control partition includes nodes 9, 10, 11, and 12, consisting of photovoltaic power plants and some energy storage units. The average electrical coupling strength between power nodes within each partition was 0.78, and the average electrical coupling strength at the boundary nodes between partitions decreased to 0.23, achieving the goal of weakly coupled partitioning.

[0052] Table 1. Electrical Coupling Strength Matrix Between Power Nodes

[0053] Each weakly coupled control zone employs a centralized model predictive control algorithm, with a prediction time domain length of 15 minutes and a control time domain length of 5 minutes. The boundary power consistency protocol, through eight iterations, converges the boundary power exchange information of each zone to a globally consistent state, reducing the boundary power deviation to within 0.8 MW. When the change in the Laplacian matrix eigenvalue distribution exceeds the eigenvalue change threshold of 0.15, the system automatically re-executes the spectral clustering segmentation algorithm to dynamically adjust the zone boundaries.

[0054] Based on step S4, objective functions are established for each market participant. Energy storage operators generate revenue from energy storage electricity during peak hours by discharging electricity and during off-peak hours by charging and storing energy. The entire lifespan of energy storage is set at 15 years, with a daily average cost of 1820 yuan. The energy storage state of charge (SOC) is maintained between 0.25 and 0.85 during operation, and an accelerated aging coefficient is calculated based on the SOC. When the SOC is 0.35, the accelerated aging coefficient is calculated to be 1.0375. Wind power operators' wind power revenue peaks during high wind speed periods, and the actual wind power output fluctuates with wind speed. The wind curtailment penalty coefficient is set by the power trading center at 0.18 yuan / kWh based on system peak-shaving demand. Technical personnel use the Copula-CVaR risk quantification model to calculate the uncertainty risk of renewable energy output, fitting the probability distribution of 365 days of historical wind and photovoltaic output data using kernel density estimation. The information entropy minimization criterion shows that the information entropy value of the t-Copula function is 2.47, lower than that of the Gaussian Copula function (2.83) and the Clayton-Copula function (2.69). Therefore, the t-Copula function is selected as the optimal Copula function family. The maximum likelihood estimation algorithm determines the Copula function parameter to be 0.63, the confidence level parameter to be set at 95%, and the risk aversion coefficient to be set at 0.12. The conditional value at risk calculation method yields an expected loss of 286,000 yuan / day due to fluctuations in renewable energy output.

[0055] Following step S5, a frequency dynamic stability constraint model is established, and the multi-unit system is converted into a single-unit equivalent system using the central inertia equivalent transformation method. The inertia constants of the four thermal power units in the system are 6.8 seconds, 7.2 seconds, 6.5 seconds, and 7.5 seconds, respectively, with rated capacities of 120MW, 100MW, 90MW, and 90MW, respectively. The calculated equivalent inertia constant is 6.95 seconds. The rotor motion frequency response equation describes the dynamic relationship between the system frequency and the power imbalance. When the power imbalance is 15MW, the system frequency deviation reaches 0.18Hz. The frequency response linearization transformation is performed using a Taylor series expansion near the rated frequency of 50Hz, retaining the first derivative term to obtain the linearized frequency response algebraic equation. The upper limit constraint of the frequency change rate is set to 0.5Hz / s, and the frequency deviation dead zone constraint range is ±0.03Hz. When the absolute value of the system frequency deviation exceeds 0.03Hz, the frequency regulation resources are adjusted according to the frequency deviation proportional coefficient of 0.06.

[0056] The multi-level constraint group includes constraints on energy storage charging and discharging power, with a maximum charging power of 80MW and a maximum discharging power of 80MW. The charging and discharging states of energy storage are described using binary variables to represent mutual exclusion relationships. The power frequency regulation capacity constraint sets the maximum frequency regulation capacity for energy storage at 25MW, for wind power at 35MW, and for photovoltaic power at 15MW. The power reserve capacity constraint sets the maximum reserve capacity for thermal power at 60MW and for energy storage at 20MW. The power balance constraint requires that the total power generation in each time period equal the load demand power. The frequency regulation demand capacity is configured at 2.5% of the net load power. The total reserve demand capacity includes renewable energy reserve demand configured at 8% of the total renewable energy output, and load reserve demand configured at 1.5% of the net load power.

[0057] Technicians executed step S6, employing a hybrid optimization solution strategy to solve the joint electricity market trading decision model. They utilized a hybrid integer second-order cone relaxation transformation to convert the non-convex quadratic constraint of energy storage charging and discharging power into a second-order cone convex constraint form. The branch-and-bound search algorithm, through 28 branch and bound operations, expanded the search space of integer variables from... The feasible region was narrowed down to 147 sub-feasible regions. The improved whale population optimization algorithm adopted an average entropy population initialization method, setting the population size to 80 individuals and the maximum number of iterations to 150. A dynamic constraint relaxation method penalized constraint violations, with an initial penalty factor set at 1000, increasing by a factor of 1.05 with each iteration. A chaotic mapping perturbation was applied using a Logistic mapping to perturb the optimal candidate solution, preventing the algorithm from getting trapped in local optima. Figure 1As shown, the convergence curve of the improved whale population optimization algorithm indicates that the objective function value tends to stabilize after 82 iterations. The upper-level Cplex solver, combined with the lower-level improved whale population optimization algorithm, obtained the optimal solution after 217 seconds of solution time. The market transaction prices, power output of each power source, energy storage charging and discharging power, frequency regulation capacity configuration value, and reserve capacity configuration value all meet the constraints.

[0058] The results show that after the market clearing, the output power of thermal power units during peak load periods is 385MW. Energy storage systems are in a discharging state from 10:00 to 15:00, with an average discharging power of 62MW, and in a charging state from 23:00 to 6:00 the next day, with an average charging power of 58MW. Wind farms achieve an output power of 218MW during high wind speed periods, accounting for 87.2% of the installed capacity, with wind curtailment controlled within 6.8MWh. Photovoltaic power plants reach a peak output power of 89MW at noon. Regarding frequency regulation capacity configuration, energy storage has a frequency regulation capacity of 18MW, wind power 26MW, photovoltaic 11MW, and thermal power 32MW, for a total frequency regulation capacity of 87MW, meeting the frequency regulation demand of 83MW. Regarding reserve capacity configuration, thermal power reserve capacity is 48MW, energy storage reserve capacity is 15MW, for a total reserve capacity of 63MW, meeting the total reserve demand of 59MW.

[0059] Following the power output coordination control command output in step S7, technicians send the power output of each power source as the active power setpoint to the corresponding controller. Upon receiving the active power setpoint, the thermal power unit controller adjusts the turbine speed control valve opening from 68% to 74%, thereby changing the steam flow to increase the active power output of the thermal power unit generator from 362MW to 385MW, with a power regulation response time of 180 seconds. Upon receiving the energy storage device power command, the energy storage device controller adjusts the active power output of the energy storage converter, increasing the energy storage battery discharge power from 56MW to 62MW by controlling the conduction angle of the IGBT power switches, with a power regulation response time of 3 seconds. Upon receiving the active power setpoint, the wind farm controller adjusts the wind turbine generator pitch angle from 5.2 degrees to 3.8 degrees, while simultaneously increasing the generator torque from 1850kNm to 2120kNm, thereby increasing the wind farm's active power output from 195MW to 218MW, with a power regulation response time of 12 seconds. After receiving the active power setpoint, the photovoltaic power station controller adjusts the maximum power point tracking reference voltage of the photovoltaic inverter from 685V to 712V, thereby increasing the active power output of the photovoltaic power station from 83MW to 89MW. The power adjustment response time is 1.5 seconds.

[0060] like Figure 2As shown, the power balance curve of the system within a 15-minute real-time dispatch cycle shows that the deviation between the sum of the output power of each power source and the load demand power remains within 2.3MW, and the power balance constraint satisfaction rate is 99.6%. After receiving the frequency regulation capacity parameters, the frequency regulation controller generates a frequency regulation power adjustment command based on the system frequency deviation. When the system frequency deviation is +0.08Hz, the frequency regulation controller issues an increase in power generation command of 5.2MW according to the upward adjustment capacity parameters, so that the system frequency recovers to +0.02Hz within 8 seconds. After receiving the standby start signal, the standby response controller rapidly increases the power generation output according to the standby capacity parameters. The thermal power unit increases from 385MW to 428MW within 3 minutes, and the energy storage system increases from 62MW to 75MW within 15 seconds, increasing the total power generation of the system by 48MW to meet the emergency load demand. The deviations between the active power output and the active power setpoint of each power source are all less than the power control accuracy threshold. The power deviation of the thermal power unit is 0.7MW, accounting for 0.18% of the rated power; the power deviation of the energy storage system is 0.3MW, accounting for 0.3% of the rated power; the power deviation of the wind farm is 1.2MW, accounting for 0.48% of the rated power; and the power deviation of the photovoltaic power station is 0.5MW, accounting for 0.5% of the rated power. All of these meet the power control accuracy requirements.

[0061] like Figure 3 As shown, the system frequency dynamic response curve indicates that after a power disturbance, the system frequency deviation is effectively suppressed by the rapid response of frequency regulation resources. The maximum frequency change rate is 0.42 Hz / s, which does not exceed the upper limit constraint of 0.5 Hz / s. The maximum frequency deviation is 0.15 Hz, and after 6 seconds of frequency regulation resource response, the frequency deviation drops to within 0.05 Hz. After 18 seconds, the frequency recovers to within ±0.03 Hz of the frequency deviation dead zone. The energy storage system provides primary support in the initial stage of frequency regulation due to its rapid response capability. Wind farms and photovoltaic power plants participate in frequency regulation through frequency regulation capacity configuration, while thermal power units provide continuous power support through primary and secondary frequency regulation. The coordinated cooperation of multiple power sources achieves rapid stabilization of the system frequency. It should be noted that the joint electricity market transaction and its game structure involved in this embodiment are as follows: Figure 4 and 5 As shown.

[0062] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A power output coordination control method for a joint electricity market trading decision model, characterized in that, A joint power market trading decision model incorporating the electricity market, frequency regulation market, and spinning reserve market is constructed. A multi-timescale coordinated optimization framework is established, and the transfer relationship between long-term decision variables and short-term boundary conditions is established through a cross-scale state transfer matrix. A power node topology weighted graph is constructed, and hierarchical coordinated control partitioning is implemented. The power node topology weighted graph is divided into multiple weakly coupled control partitions using a spectral clustering segmentation algorithm. Objective functions for each market participant are established, and a risk quantification model is introduced. A frequency dynamic stability constraint model is established, and a multi-level constraint condition group is set. The multi-machine system is converted into a single-machine equivalent system using the central inertia equivalent transformation method. The rotor motion frequency response equation is used to describe the dynamic relationship between system frequency and power imbalance. The rotor motion frequency response equation is transformed into a frequency deviation algebraic inequality constraint through frequency response linearization transformation. A hybrid optimization solution strategy is used to solve the joint power market trading decision model, output power output coordinated control commands, and implement hierarchical control.

2. The method according to claim 1, characterized in that, The joint power market trading decision model includes a market-level joint trading framework, a bidding-level game-theoretic decision-making module, and a dispatch-level execution control module.

3. The method according to claim 2, characterized in that, The market-level joint trading framework and the bidding-level game decision-making module form a Stackelberg game relationship, while the energy storage operator decision-making unit, wind power operator decision-making unit, and photovoltaic operator decision-making unit within the bidding-level game decision-making module form a Nash game relationship.

4. The method according to claim 3, characterized in that, The establishment of a multi-timescale coordination and optimization framework discretizes the day-ahead clearing time window with a 1-hour time granularity, the real-time scheduling with a 15-minute time granularity, and the frequency modulation response with a 1-second time granularity.

5. The method according to claim 4, characterized in that, The topology weighted graph of power nodes is constructed by establishing a topology weighted graph of power nodes with each power node as a vertex, and the electrical coupling strength is used as the edge weight of the topology weighted graph of power nodes.

6. The method according to claim 5, characterized in that, The execution of the spectral clustering segmentation algorithm involves calculating the eigenvalues ​​and eigenvectors of the Laplacian matrix of the power node topology weighted graph, and clustering power nodes with high similarity index values ​​into the same weakly coupled control partition.

7. The method according to claim 6, characterized in that, Each weakly coupled control zone achieves inter-zone coordination through a boundary power consistency protocol, and a centralized model predictive control algorithm is used within each weakly coupled control zone to optimize the power output of the power source within the zone.

8. The method according to claim 7, characterized in that, The objective function of an energy storage operator is the sum of the revenue from the energy storage electricity market, the revenue from the energy storage frequency regulation market, and the revenue from the energy storage reserve market, minus the energy storage life-cycle amortized cost and the energy storage cycle loss cost.

9. The method according to claim 8, characterized in that, The objective function of wind power operators is the sum of wind power revenue, wind power frequency regulation revenue, and wind power reserve revenue, minus the cost of wind curtailment penalties.

10. The method according to claim 9, characterized in that, The introduction of the risk quantification model uses the Copula-CVaR risk quantification model to calculate the risk value of revenue loss caused by the uncertainty of renewable energy output and incorporates it into the objective function of each market participant.

Images