Response prediction-based power distribution network toughness improvement optimization method, system and device, and storage medium

Abstract

The invention relates to the technical field of power distribution network toughness demand response, in particular to a power distribution network toughness improvement optimization method, system and device based on response prediction and a storage medium. Training an integrated decision tree model based on historical data to predict a response intention value of the cooling and heating load user, and determining a temperature regulation boundary and a response frequency upper limit according to the response intention value; generating an initial scene set through source load uncertainty sampling, extracting a typical scene by adopting a transportation distance scene reduction method, and introducing an overall offset constraint and a single-point extreme value constraint to perform two-dimensional limitation on scene probability distribution; a scene probability combination enabling the load recovery value to be minimum is searched in a probability distribution domain, and a scheduling scheme enabling the key load recovery value to be maximum is solved under the condition that the power flow constraint, the cold and heat power balance constraint and the temperature regulation boundary constraint are met; and carrying out probability weighting on the load recovery values of different scenes to obtain a toughness evaluation value, and carrying out sensitivity analysis.

Description

Technical Field

[0001] This invention relates to the field of distribution network resilience demand response technology, and in particular to a distribution network resilience enhancement optimization method, system, device and storage medium based on response prediction. Background Technology

[0002] Traditional power grid dispatching methods mainly rely on generation-side resources for load restoration. Currently, temperature control equipment for heating and cooling loads is gradually becoming an important resource for improving the resilience of power grids due to its high flexibility and fast response speed. Combined cooling, heating and power (CCHP) systems mainly achieve energy conversion and complementarity between electrical energy, heat energy and cooling energy through the coordinated operation of gas turbines, waste heat recovery devices and refrigeration equipment, while providing regulation for the power grid.

[0003] Current methods for improving the resilience of distribution networks still have the following limitations. On the one hand, current temperature control responses generally adopt fixed temperature ranges or static adjustment models based on thermal inertia, but they easily overlook the subjective willingness of users to actually participate in demand response. Furthermore, users' response behavior is affected by multiple factors such as incentive levels, temperature perception, and behavioral habits. Existing simple temperature constraints cannot accurately characterize users' true response capabilities, making it difficult to execute dispatch instructions. On the other hand, the high uncertainty of wind and solar power output affects load recovery dispatch. Traditional optimization methods often use single norm constraints to construct uncertain sets, which can only control the overall shift of the probability distribution or one of the single extreme values, and cannot simultaneously take into account the overall distribution characteristics and extreme scenarios, resulting in overly conservative optimization.

[0004] In addition, existing scheduling models mostly focus on power optimization and lack systematic modeling of energy coupling relationships in combined cooling, heating and power systems; the coordinated regulation among equipment such as gas turbines, waste heat recovery devices, and absorption chillers has not been fully explored, and the regulatory role of cooling and heating loads in emergency recovery has been ignored, reducing the overall resilience of the distribution network. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention is proposed.

[0006] Therefore, the problems to be solved by this invention are: how to accurately predict users' willingness to adjust temperature and transform the uncertainty of user behavior into quantifiable scheduling constraints; how to construct a probabilistic constraint method that takes into account both overall distribution characteristics and extreme scenarios under the condition of high uncertainty of wind and solar power output, so as to avoid overly conservative optimization strategies; and how to make full use of the coordinated regulation capability of the combined cooling, heating and power system to achieve rapid load recovery under multi-energy complementarity during the process of improving the resilience of the distribution network.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a distribution network resilience enhancement optimization method based on response prediction, which includes: acquiring basic data of the distribution network, training an integrated decision tree model based on the basic data, predicting the response willingness values ​​of cold and hot load users, and determining the temperature regulation boundary and the upper limit of the response frequency based on the response willingness values; An initial scene set is generated by sampling the source load uncertainty. Typical scenes and initial probability distributions are extracted by a scene reduction method based on transportation distance. Overall offset constraints and single-point extreme value constraints are introduced to limit the probability distribution of typical scenes in two dimensions to form a probability distribution domain. The overall offset constraints limit the sum of the deviations of all scene probabilities from the initial distribution, and the single-point extreme value constraints limit the maximum fluctuation of the probability of a single scene. Search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained from the search, solve for the scheduling scheme that maximizes the critical load recovery value under the conditions of power flow constraints, hot and cold power balance constraints, and temperature regulation boundary constraints. Based on the load recovery status of each node output by the scheduling scheme, the load recovery value under different scenarios is probabilistically weighted to obtain the distribution network resilience assessment value.

[0008] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, the step of training an ensemble decision tree model based on basic data to predict the response willingness values ​​of cold and hot load users includes: The actual temperature, expected temperature, temperature deviation, and stimulus level of historical periods are used as input features, and whether users participated in the response during the corresponding period is used as a label. Multiple decision trees are trained using the gradient boosting algorithm, with each decision tree trained based on the residuals of its predecessor trees. Input the temperature status and excitation conditions of the current scheduling cycle, and output the response intention value.

[0009] The beneficial effects of this preferred technical solution are as follows: by collecting the user's actual temperature, expected temperature, temperature deviation, and incentive level during historical periods as feature dimensions, a mapping relationship between user response behavior and environmental conditions and incentive factors is established; the gradient boosting algorithm trains multiple decision trees, with each tree optimizing the prediction residual of the preceding tree, gradually capturing the nonlinear patterns and heterogeneous characteristics in user response behavior; after inputting the temperature state and incentive conditions of the current scheduling cycle, the response willingness value output by the model quantifies the user's subjective participation willingness into a specific value, providing personalized constraint boundaries for subsequent scheduling.

[0010] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, wherein: determining the temperature adjustment boundary according to the response intention value includes obtaining the upper and lower limits of the temperature comfort range for users of cold and hot loads; By multiplying the upper and lower limits of the temperature comfort range by the response willingness value, the user's willingness to participate under the scheduling command is predicted, and the actual adjustable upper and lower limits of the temperature are obtained.

[0011] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, the method of introducing overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenarios in two dimensions includes setting an overall offset threshold, requiring that the sum of the absolute values ​​of the differences between the actual probability and the initial probability of all typical scenarios is less than the overall offset threshold. Set a single-point extreme value threshold, requiring that the absolute value of the difference between the actual probability and the initial probability of any typical scenario is less than the single-point extreme value threshold; Based on the total number of wind and solar power output scenarios, the number of typical scenarios, and the target confidence level, the values ​​of the overall offset threshold and the single-point extreme value threshold are inversely derived through probability inequalities.

[0012] The beneficial effects of this preferred technical solution are as follows: by setting an overall offset threshold to constrain the cumulative deviation of the probability of all scenarios from the initial distribution, the actual probability distribution is kept close to the reference distribution in the whole; by setting a single-point extreme value threshold to limit the maximum fluctuation of the probability of a single scenario, the local extreme situation caused by excessive deviation of the probability of a certain scenario is prevented; based on the total amount of wind and solar power output scenarios, the number of typical scenarios and the target confidence level, the values ​​of the two thresholds are inversely derived through probability inequalities, and a quantitative relationship between the constraint parameters and the confidence level is established.

[0013] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, the step of searching for the scenario probability combination that minimizes the load recovery value within the probability distribution domain includes solving for the load recovery value corresponding to each typical scenario under a given scenario probability combination. With the goal of minimizing the probabilistic weighted sum of load restoration value, the scenario probability combination is solved under the conditions of probability normalization constraint, overall offset constraint and single-point extreme value constraint.

[0014] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, the scheduling scheme includes gas turbine output, waste heat recovery ratio, refrigeration equipment operating status and cold and heat load response status, including determining the power generation output, energy storage scheduling and load response status for each time period, and calculating the input power of waste heat boiler and low temperature waste heat power generation device based on the gas turbine's heat-to-power ratio and waste heat recovery ratio. Calculate the input heat power required to meet the absorption cooling load based on the energy efficiency coefficient of the absorption chiller. Calculate the input power required to meet the electric cooling load based on the coefficient of performance (COP) of the electric chiller. Under the constraints of electric power balance, thermal power balance, and cold power balance, determine the gas turbine output, energy storage charging and discharging power, and cold and hot load response.

[0015] As a preferred embodiment of the distribution network resilience enhancement optimization method based on response prediction described in this invention, the method for obtaining the distribution network resilience assessment value includes statistically analyzing the load recovery status and recovery value under various typical scenarios. The recovery value is weighted based on the scenario probability distribution to obtain the overall resilience assessment value of the distribution network. Analyze the impact of changes in response intention on resilience indicators and generate sensitivity assessment results.

[0016] Secondly, embodiments of the present invention provide a distribution network resilience enhancement optimization system based on response prediction, which includes a response prediction module for acquiring basic data of the distribution network, training an integrated decision tree model based on the basic data, predicting the response intention values ​​of cold and hot load users, and determining the temperature regulation boundary and the upper limit of the response frequency based on the response intention values. The scenario construction module is used to generate an initial scenario set through source load uncertainty sampling. It extracts typical scenarios and initial probability distributions through a scenario reduction method based on transportation distance. It introduces overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenarios in two dimensions, forming a probability distribution domain. The overall offset constraints limit the sum of the deviations of all scenario probabilities from the initial distribution, and the single-point extreme value constraints limit the maximum fluctuation range of the probability of a single scenario. The optimization solution module is used to search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained by the search, under the conditions of satisfying power flow constraints, hot and cold power balance constraints and temperature regulation boundary constraints, the scheduling scheme that maximizes the critical load recovery value is solved. The strategy generation module is used to output the load recovery status of each node according to the scheduling scheme, and to perform probability weighting on the load recovery value under different scenarios to obtain the distribution network resilience assessment value.

[0017] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of the distribution network resilience enhancement optimization method based on response prediction as described in the first aspect of the present invention.

[0018] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the distribution network resilience enhancement optimization method based on response prediction as described in the first aspect of the present invention.

[0019] The beneficial effects of this invention are as follows: By training an ensemble decision tree model, a mapping relationship is established between multi-dimensional factors such as user temperature perception, motivation level, and behavioral habits and historical response behavior. The output response willingness value quantifies the user's subjective participation willingness as a temperature regulation boundary and a response frequency upper limit, solving the problem that traditional fixed temperature constraints cannot reflect the user's true response capability, leading to difficulties in executing scheduling instructions. After generating the initial wind and solar load scenario through Monte Carlo sampling, a scenario reduction method based on transportation distance is used to extract typical scenarios. An overall offset constraint is introduced to control the sum of all scenario probabilities deviating from the initial distribution, and a single-point extreme value constraint is introduced to limit the maximum fluctuation range of a single scenario probability, forming a system that takes into account both the overall distribution characteristics and... The probability distribution domain for extreme scenario control is used. After searching for the scenario probability combination that minimizes the load recovery value within the probability distribution domain, the energy conversion relationship between gas turbine power generation, waste heat boiler heat generation, absorption chiller and electric chiller refrigeration is coordinated under the most unfavorable scenario combination. Under the constraints of power flow, heating and cooling power balance and temperature regulation boundary, the scheduling scheme that maximizes the critical load recovery value is solved, making full use of the coordinated regulation capability of the combined cooling, heating and power system. The load recovery value of different scenarios is probabilistically weighted to obtain the distribution network resilience assessment value. The critical response users are identified through sensitivity analysis, providing a quantitative basis for scheduling strategy and realizing the rapid recovery of the distribution network under extreme scenarios. Attached Figure Description

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

[0021] Figure 1 The flowchart shows the optimization method for improving the resilience of distribution networks based on response prediction. Figure 2 A computer equipment diagram for a distribution network resilience enhancement optimization method based on response prediction; Figure 3 A schematic diagram of the temperature control response prediction process based on XGBoost, which is an optimization method for improving the resilience of distribution networks based on response prediction. Figure 4 This is a schematic diagram illustrating the source-load scenario generation and probability distribution constraints of the distribution network resilience enhancement optimization method based on response prediction. Figure 5 This is a schematic diagram of the two-layer optimization structure of the distribution network resilience enhancement optimization method based on response prediction; Figure 6 A schematic diagram of the energy flow of a combined cooling, heating and power system based on a response prediction-based optimization method for improving the resilience of distribution networks; Figure 7 This is a schematic diagram of the load recovery scheduling results of the distribution network resilience enhancement optimization method based on response prediction; Figure 8 The figure shows the results of resilience assessment and sensitivity analysis of the distribution network resilience enhancement optimization method based on response prediction. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0025] Example 1 Reference Figure 1 - Figure 2 This is the first embodiment of the present invention, which provides a distribution network resilience improvement optimization method based on response prediction, including: S100: Acquire basic data of the distribution network, train an integrated decision tree model based on the basic data, predict the response intention values ​​of cold and heat load users, and determine the temperature regulation boundary and response frequency upper limit based on the response intention values.

[0026] S200: An initial scene set is generated by sampling the source load uncertainty. Typical scenes and initial probability distributions are extracted by a scene reduction method based on transportation distance. Overall offset constraints and single-point extreme value constraints are introduced to limit the probability distribution of typical scenes in two dimensions, forming a probability distribution domain. Overall offset constraints limit the sum of the deviations of all scene probabilities from the initial distribution, while single-point extreme value constraints limit the maximum fluctuation of the probability of a single scene.

[0027] S300: Search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained from the search, solve for the scheduling scheme that maximizes the critical load recovery value under the conditions of power flow constraints, hot and cold power balance constraints, and temperature regulation boundary constraints.

[0028] S400: Output the load recovery status of each node according to the scheduling scheme, and perform probability weighting on the load recovery value under different scenarios to obtain the distribution network resilience assessment value.

[0029] It should be noted that after the distribution network suffers from extreme weather or external shocks, users' willingness to respond to temperature control commands is affected by temperature perception, incentive level and behavioral habits. Traditional fixed temperature constraints cannot accurately characterize users' actual participation ability; random fluctuations in wind and solar power output make source-load matching difficult.

[0030] Therefore, the complete technical solution is achieved through steps S100-S400. S100 predicts user response intentions using an integrated decision tree model, transforming the uncertainty of subjective behavior into temperature regulation boundaries and response frequency upper limits, providing personalized constraints for scheduling. S200 uses the Monte Carlo method to generate initial wind and solar load scenarios, extracts typical scenarios by reducing transportation distances, and introduces dual-dimensional constraints of overall offset and single-point extrema to limit the scenario probability distribution domain, controlling the overall distribution characteristics while suppressing extreme fluctuations. S300 searches for the most unfavorable scenario combination within the probability distribution domain, and solves the load recovery scheduling scheme for this combination under the constraints of power flow, cooling and heating power balance, and temperature boundaries, maximizing the recovery value of critical loads through coordinated cooling, heating, and power regulation. S400 performs probability weighting on the recovery value of different scenarios, outputting a distribution network resilience assessment value, quantifying the system's recovery capability under extreme scenarios.

[0031] Example 2 Reference Figure 2 - Figure 8 This is the second embodiment of the present invention.

[0032] In this embodiment, step S100 involves acquiring basic data from the distribution network, training an ensemble decision tree model based on the basic data, predicting the response intention values ​​of cooling and heating load users, and determining the temperature regulation boundary and response frequency upper limit based on the response intention values. This includes the following steps A1-A2: A1: Based on basic data, train an ensemble decision tree model to predict the response intentions of users with heating and cooling loads, including... The actual temperature, expected temperature, temperature deviation, and stimulus level of historical periods are used as input features, and whether users participated in the response during the corresponding period is used as a label. Multiple decision trees are trained using the gradient boosting algorithm, with each decision tree trained based on the residuals of its predecessor trees. Input the temperature status and excitation conditions of the current scheduling cycle, and output the response intention value.

[0033] Specifically, the basic data for the distribution network includes historical temperature response data from the user side, economic incentive records, renewable energy forecast data, power load forecast data, and parameters of the combined cooling, heating, and power (CCHP) system. Historical temperature response data from the user side records the actual indoor temperature, desired temperature, response behavior, and corresponding environmental conditions of users at different times. This data, after preprocessing, forms a user behavior feature database. Economic incentive records contain the correspondence between incentive levels, incentive methods, and actual user responses within historical dispatch cycles. Renewable energy forecast data covers historical output curves and forecast error distributions for wind and solar power. Power load forecast data includes the time-series characteristics of load demand at each node. CCHP system parameters include equipment parameters such as the rated power of gas turbines, the efficiency of waste heat recovery devices, the coefficient of performance (COP) of absorption chillers, the coefficient of performance (COP) of electric chillers, and the capacity and charge / discharge efficiency of energy storage devices.

[0034] Temperature range that the temperature control load can withstand Primarily determined by the individual's comfort level when using energy, at their desired optimal temperature. There exists a comfort zone in the vicinity where users perceive a high level of comfort. When the user's actual temperature deviates from this comfort zone, their overall comfort satisfaction decreases, and their desire to improve comfort increases, thus motivating them to improve the temperature through load adjustment. Because the user's current actual temperature... Often different from the expected optimal temperature Combined with user's temperature adjustment response Its actual usable temperature range as follows: (1) The willingness to respond in the formula reflects the differences in users' enthusiasm for participating in demand-side response according to dispatch instructions. Its magnitude is influenced by both objective and subjective factors. Objective factors are determined by the deviation of the user's current temperature from the aforementioned temperature comfort zone, while subjective factors are mainly related to the level of economic incentives for demand response and the degree of matching between user expectations. It is evident that users' willingness to respond has significant uncertainty, and its value will profoundly affect the dispatchable potential of temperature-controlled loads, thereby impacting the final effect of improving the resilience of the distribution network.

[0035] The willingness to respond value reflects the differences in users' willingness to participate in demand-side response according to dispatch instructions. Its magnitude is influenced by both objective and subjective factors. Objective factors are determined by the deviation of the user's current temperature from the comfort zone, while subjective factors are mainly related to the economic incentive level for demand response and the degree of matching between user expectations. The willingness to respond value is uncertain, and its value will affect the dispatchable potential of temperature-controlled loads, thereby affecting the effectiveness of distribution network resilience improvement.

[0036] This reveals the differences in user response intentions under different feature value ranges, significantly improving prediction accuracy while also possessing efficient parallel processing capabilities. The model can be expressed as equation (2). Where K is the number of decision trees, Let F be an independent function of the function space F consisting of decision trees.

[0037] (2) The objective function of the XGBoost algorithm can be divided into two parts: the loss error function and the regularization term. The loss error function... Describes the difference between the predicted target value and the actual target value; regularization term Control the complexity of the tree to prevent overfitting.

[0038] (3) Here, N represents the dimensionality of the prediction data, and K represents the number of decision trees. The loss function ensures that the model continuously approximates the true response intention values ​​of the training data, while the regularization term constrains model complexity to avoid overly fitting the training samples and lacking generalization ability. The error term focuses on reducing the difference between the predicted values ​​and actual user responses to improve prediction accuracy; the regularization term suppresses excessive model complexity to achieve better generalization performance. These two parts work together to enable the XGBoost model to maintain appropriate simplicity while pursuing accurate fitting, thus accurately reflecting historical patterns and handling potential new situations in response intention prediction tasks.

[0039] Accurately predicting user responses to temperature changes helps dispatchers optimize the prioritization of cold and hot users and rationally allocate power resources, thereby reducing energy waste and improving the overall resilience of the system. Temperature regulation response prediction analyzes historical temperature and load data, combines user behavior and external variables, and uses the XGBoost algorithm to build a model to predict user temperature regulation behavior in future periods.

[0040] The temperature response prediction process based on XGBoost is as follows: Figure 2 As shown, historical temperature data, temperature control response data, and stimulus data are first used as input data to capture the correlation between user behavior and the environment and stimulus strategies.

[0041] Subsequently, the XGBoost algorithm was modeled, with hyperparameters including the number of iterations, tree depth, and shrinkage step size. The number of iterations controls the number of ensemble trees; tree depth limits the complexity of a single decision tree to prevent overfitting; and the learning rate adjusts the magnitude of weight updates during gradient descent. The model is iteratively trained and optimized to obtain the optimal parameters.

[0042] Finally, the trained model is used to predict user response intentions. The predicted temperature control response is then fed into the resilience enhancement model to help optimize load management and improve the grid's responsiveness and system resilience.

[0043] A2: Determine the temperature adjustment boundary based on the response intention value, including obtaining the upper and lower limits of the temperature comfort range for users with heating and cooling loads; By multiplying the upper and lower limits of the temperature comfort range by the response willingness value, the user's willingness to participate under the scheduling command is predicted, and the actual adjustable upper and lower limits of the temperature are obtained.

[0044] Specifically, the upper and lower limits of the temperature comfort range for users with both heating and cooling loads are obtained. The temperature comfort range reflects the temperature range that users can accept without affecting normal production and daily life. For users with cooling loads, the temperature comfort range is typically set as follows: ℃; for users with high heat loads, the temperature comfort range is typically set to ℃. ℃. These comfort zone parameters are determined based on users' energy consumption characteristics and historical scheduling experience, and are adjusted in actual applications according to building type, user preferences, and regional climate conditions.

[0045] Multiplying the upper and lower limits of the temperature comfort range by the willingness to respond, respectively, predicts the user's willingness to participate under scheduling instructions, thus obtaining the actual adjustable temperature upper and lower limits. The actual adjustable temperature boundary for user i in time period t is calculated as follows: In the formula, and For user i, the actual adjustable temperature lower and upper limits during time period t; The optimal temperature for user i; This represents the predicted willingness to respond. When the willingness to respond... When the value is close to 1, it indicates that the user's willingness to respond is high and the actual adjustable temperature range is close to the comfort range; when the willingness to respond is close to 0, it indicates that the user's willingness to respond is low and the actual adjustable temperature range shrinks to near the desired temperature.

[0046] The upper limit of response frequency is determined based on the response willingness value. To avoid disrupting users' normal work and daily life, the maximum number of responses a user can make within the entire scheduling cycle is limited. The upper limit of response frequency is positively correlated with the response willingness value, and is calculated as follows: in, The maximum number of responses for user i within the scheduling period; The baseline maximum number of responses is typically set to 30% to 50% of the total number of scheduling period segments; This represents the average response willingness value of user i across all time periods within the scheduling period; ⌊⋅⌋ represents the floor function. Users with high response willingness values ​​have a higher upper limit on response frequency and can participate in scheduling for more time periods, while users with low response willingness values ​​have their response frequency limited to avoid excessive interference with the user's energy experience.

[0047] In this embodiment, step S200 generates an initial scene set through source load uncertainty sampling, extracts typical scenes and initial probability distributions through a scene reduction method based on transportation distance, and introduces overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenes in two dimensions to form a probability distribution domain, including the following step B1: B1: An initial scene set is generated by sampling the source load uncertainty, and typical scenes and initial probability distributions are extracted by a scene reduction method based on transportation distance.

[0048] Set an overall offset threshold, requiring that the sum of the absolute values ​​of the differences between the actual probabilities and the initial probabilities of all typical scenarios be less than the overall offset threshold; Set a single-point extreme value threshold, requiring that the absolute value of the difference between the actual probability and the initial probability of any typical scenario is less than the single-point extreme value threshold; Based on the total number of wind and solar power output scenarios, the number of typical scenarios, and the target confidence level, the values ​​of the overall offset threshold and the single-point extreme value threshold are inversely derived through probability inequalities.

[0049] Specifically, the actual output of wind power, photovoltaic power and electrical load is determined by the predicted output and the prediction error. Existing studies often use a normal distribution model to fit their prediction errors, and the error generally does not exceed 15%.

[0050] The sampling strategy for generating source-load scenarios using the Monte Carlo method is as follows: a normal distribution is used to simulate wind and solar prediction errors and load fluctuations, and the probability distribution function is determined using the mean and standard deviation. The number of sampling points is determined based on the convergence of the optimization objective.

[0051] The linear reduction method based on optimal transportation distance employs an iterative two-step strategy. First, with a fixed subset of scenarios, its probability distribution is recalculated using linear programming. Then, with a fixed probability, the scenarios requiring updating are replaced from the original subset through transportation cost distance assessment. These two steps are executed alternately until the Wasserstein distance change falls below a threshold.

[0052] The initial probability distribution obtained for each typical scenario is biased. Due to the uncontrollable power output of wind and solar power, there must be a certain deviation between the actual probability distribution and the initial distribution. Therefore, it is necessary to introduce a comprehensive norm condition to constrain the power output scenarios of wind and solar power.

[0053] First, assuming there are A scenarios with wind and solar power output, clustering is used to filter them, resulting in K discrete scenarios, typical wind and solar power output scenarios, and the initial probability distribution of each scenario. Then, since wind and solar power output is uncontrollable, there will be a certain deviation between the actual probability distribution and the initial probability distribution. Therefore, in addition to the constraints mentioned above, a comprehensive norm condition is introduced to constrain the wind and solar power output scenarios: In the formula: An uncertain set of probability distributions with the 1-norm; This represents the permissible deviation value for the probability of landscape scenes under the 1-norm constraint.

[0054] In the formula: An uncertain set of probability distributions with the ∞-norm; Let be the permissible deviation value of the probability of a landscape scene under the ∞-norm constraint.

[0055] along with and As the value increases, the model's solution becomes overly conservative, resulting in a significant discrepancy with reality. Therefore, it is necessary to find a method to determine a reasonable level of conservatism.

[0056] The confidence level can be expressed as: In the formula: Let be the probability that the inequality holds true; and Let represent the confidence levels for the two inequalities to hold under the constraints of the 1-norm and ∞-norm, respectively, for discrete scenarios. Therefore, the relationship between uncertainty, the total number of wind and solar power output scenarios, the number of typical wind and solar power output scenarios, and the confidence levels can be obtained as follows: In the formula: ; According to the above formula, as the cumulative amount of data increases, the uncertainty approaches 0 infinitely. At this point, the comprehensive norm probability distribution of the wind and solar power output scenario approaches the actual probability distribution.

[0057] In the distribution network resilience enhancement model, the source-load scenario generation method based on the comprehensive norm provides more diverse and accurate inputs for the system optimization process. These scenarios not only better reflect the uncertainty of renewable energy output such as wind and solar power, but also improve the model's adaptability to actual operating conditions. By balancing overall and extreme uncertainties while reducing overly conservative strategies, this approach enhances the grid's resilience in the face of complex and extreme events.

[0058] In this embodiment, step S300, which involves searching for the scenario probability combination that minimizes the load recovery value within the probability distribution domain and solving for the scheduling scheme that maximizes the critical load recovery value, includes the following steps C1-C2: C1: Search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain, including solving the load recovery value corresponding to each typical scenario under a given scenario probability combination; With the goal of minimizing the probabilistic weighted sum of load restoration value, the scenario probability combination is solved under the conditions of probability normalization constraint, overall offset constraint and single-point extreme value constraint.

[0059] Specifically, the probability combination of scenarios that minimizes load restoration value within the probability distribution domain adopts a two-layer optimization structure: the upper layer aims to maximize the minimum load restoration value and searches for the probability distribution of scenarios that minimizes system resilience under the constraints of the comprehensive norm; the lower layer solves the load restoration scheduling problem in each typical scenario, with the goal of maximizing the restoration value of critical loads while satisfying the operating constraints of electrical and thermal energy systems and user behavior restrictions.

[0060] The upper-layer model focuses on finding the most unfavorable power-load scenario probability distribution within a given uncertain confidence domain to examine extreme supply-demand imbalances that the system may encounter during the post-disaster emergency recovery phase. The lower-layer model, on the other hand, optimizes the scheduling decisions for power load recovery after determining the scenario probabilities for specific source-load scenarios, maximizing the recovery value of critical loads during the outage period. Although the model is two-layered, the upper and lower layer problems can actually be decoupled. That is, the lower-layer model can be solved independently first, and then the upper-layer model can be solved based on the solution results of the lower-layer model.

[0061] The lower-level problem corresponds to the scheduling optimization problem given a specific scenario, and can be expressed as: The aforementioned lower-level problem is a typical mixed-integer linear programming problem with clearly defined constraints and well-defined decision variables. In practice, the YALMIP optimization modeling tool within the MATLAB environment can be directly invoked, and the optimal solution can be obtained quickly and efficiently through commercial solvers such as CPLEX. During this process, because this paper accurately predicts the temperature adjustment response intentions of both hot and cold users, the obtained temperature adjustment response parameters (including upper and lower temperature limits and response count limits) can be directly converted into specific numerical values ​​of the constraints in the model, greatly improving the accuracy and efficiency of the model solution.

[0062] The upper-level problem can be represented as searching for the probability distribution of the scenario that results in the worst system resilience within the probability confidence region, based on the solution results of the lower-level problem. This problem can be expressed as: Since the upper-level problem is a simple linear optimization problem, the solution size is finite and well-defined. The recovered value obtained from the lower-level problem... By inputting known constants into the upper-level model, the worst-case probability distribution can be quickly obtained through linear programming. Similarly, it can be solved efficiently by calling existing linear programming solvers such as CPLEX using YALMIP.

[0063] Therefore, the actual solution steps are as follows: For each given scenario, first solve the lower-level load recovery scheduling problem separately to obtain the optimal solution for recovery value for all scenarios. Using the solution results of the lower-level problem as input, construct and solve the upper-level linear optimization problem to obtain the most unfavorable probability distribution and the corresponding worst-case scenario for the distribution network resilience.

[0064] The optimization variables include: load recovery state decision; energy storage charging and discharging power; gas turbine power generation output; heat exchange ratio and operating status of refrigeration equipment; and start-up and shutdown status of heat and cold users.

[0065] The constraints include: node power balance and voltage operating boundaries; energy balance between heat load and cooling load; natural gas pipeline flow constraints; upper limit of user temperature adjustment response times, comfort maintenance constraints, and adjustable temperature range boundaries, etc.

[0066] C2: The dispatching scheme includes gas turbine output, waste heat recovery ratio, refrigeration equipment operation status and cold and heat load response status, including determining the power generation output, energy storage dispatch and load response status for each time period, and calculating the input power of waste heat boiler and low temperature waste heat power generation device based on the gas turbine's heat-to-power ratio and waste heat recovery ratio. Calculate the input heat power required to meet the absorption cooling load based on the energy efficiency coefficient of the absorption chiller. Calculate the input power required to meet the electric cooling load based on the coefficient of performance (COP) of the electric chiller. Under the constraints of electric power balance, thermal power balance, and cold power balance, determine the gas turbine output, energy storage charging and discharging power, and cold and hot load response.

[0067] Specifically, the constraints of the distribution network resilience enhancement model include: integrated energy system equipment operation constraints, the specific formulas are as follows; cooling and heating user dispatch constraints, including user temperature availability adjustment range constraints, temperature deviation constraints, and participation response frequency constraints, the specific formulas are also as follows; in addition, system safety operation constraints must be met, including comfort constraints, distribution network operation constraints, natural gas network operation constraints, and cooling, heating, and power balance constraints, as shown below. Gas turbines and gas boilers: In the formula: For gas turbines during time periods Power generation capacity, ; The power generated by the natural gas consumed by the gas turbine. ; The power generation efficiency of the gas turbine; This represents the minimum power output of the gas turbine. ; This represents the maximum power output of the gas turbine. ; This represents the lower limit of the gas turbine's ramp power. ; This represents the upper limit of the gas turbine's ramp power. ; For gas boilers during time periods Heating power, ; The power generated by the natural gas consumed by the gas-fired boiler. ; Thermal efficiency of gas-fired boilers; This represents the minimum heating power of the gas-fired boiler. ; This represents the maximum heating power of the gas-fired boiler. .

[0068] Waste heat boilers and low-temperature waste heat power generation units: In the formula: The heat-to-electric ratio of a gas turbine; For time period The proportion of waste heat from gas turbines that is recovered by waste heat boilers; For time period The proportion of waste heat from gas turbines that is recovered by low-temperature waste heat power generation devices; The heat recovery efficiency of the waste heat boiler; The heat-to-electricity conversion efficiency of a low-temperature waste heat power generation device.

[0069] Absorption chillers and electric chillers: In the formula: For time period Input heat power of absorption chiller ; For absorption refrigeration, the refrigeration power is... ; The coefficient of performance (COP) of an absorption chiller; This represents the minimum input thermal power. ; This represents the maximum value of the input thermal power. ; For time period Input electrical power of the electric refrigeration unit ; For the refrigeration power of the electric refrigeration unit, ; The coefficient of performance (COP) of an electric refrigeration system; This represents the minimum input electrical power. ; This represents the maximum value of the input electrical power. .

[0070] Heat exchanger: In the formula: For time period The input heat power of the heat exchanger ; To output thermal power, ; This represents the energy efficiency ratio. The thermal power input to the absorption chiller and heat exchanger satisfies the above formula.

[0071] Energy storage devices: The power output constraints, charge / discharge power constraints, energy storage charge / discharge state constraints, and energy storage capacity constraints for energy storage i in time period t are as follows: In the formula: , Let i be the discharge power and charging power of energy storage in time period t, in MW; The maximum charging power is expressed in MW. In charging state; The maximum discharge power is expressed in MW. It is in a discharge state; For discharge efficiency; Let i be the capacity of energy storage in time period t, in MWh; For charging efficiency; , The upper and lower limits of energy storage capacity are in MWh.

[0072] Photovoltaic power generation: In the formula: This is the power factor, typically taken as 0.9; Rated power, in MW; The actual irradiance of the photovoltaic module is expressed in W / m². The power temperature coefficient; The surface temperature of the solar panel is ℃; The temperature is the standard test temperature, typically 25℃.

[0073] Wind power generation: In the formula: Rated power, in MW; Let be the wind speed during time period t, in m / s; Cut-in wind speed, m / s; Rated wind speed, m / s; To cut off the wind speed, m / s.

[0074] The comfort model is designed to enhance the resilience of distribution networks. The dispatching process for clusters of cooling and heating users should consider the users' energy consumption goals; in other words, the dispatching strategy should take into account the users' temperature comfort requirements, and after dispatching, the temperature where users are located should be as far away as possible from the boundary of the available temperature control range. To evaluate the relationship between temperature and the controllable domain boundary after cooling / heating loads participate in demand response, a temperature deviation index is constructed. As shown below: In the formula: For the first The median of the available temperature control range for each cluster, in °C; For the first The number of cold / hot load users contained in each cluster; Total number of users with cooling / heating load; For 0 / 1 variables, cooling / heating load During the period When responding Select 1 if the value is 1, otherwise select 0. The total number of clusters representing cooling / heating loads; This represents the total number of time periods during the scheduling cycle. The temperature deviation index formula constructed above indicates the degree to which the temperature of all cold / heat load clusters deviates from the median of their available temperature range throughout the entire scheduling cycle. Considering the different number of users in each cluster, [the following is used...] As its weight, The smaller the value, the greater the user's temperature comfort.

[0075] Based on this metric, during the entire scheduling process, each user cluster must satisfy the following two formulas: The temperature shown can be constrained by the adjustable range, and also needs to meet the temperature deviation constraint, as expressed in the following formula: In the formula: For 0 / 1 variables, cold / hot users The value is 1 if the user does not participate in the response, and 0 otherwise. This is the temperature deviation threshold for cold / heat loads.

[0076] When implementing demand response, in order not to disrupt users' normal production and daily life, the maximum number of responses a user can make within the entire scheduling cycle is usually limited, as shown in the following formula. Wherein, This represents the maximum number of times the load participates in the response.

[0077] Distribution network model: (1) Active and reactive power balance at distribution network nodes: In the formula: , For the scene Down Time period from node Flow to Node The active power and reactive power are respectively, with units of active power and reactive power. and ; , For the scene Down Time period from node Flow to Node The active power and reactive power are respectively, with units of active power and reactive power. and ; For nodes A set of lines for terminal nodes; For nodes The set of lines originating from the node; , For the scene Down Time period nodes The raw active power and raw reactive power, respectively, are in units of and .

[0078] (2) Relationship between voltage and power at both ends of the line: In the formula: , For the scene Down Time-of-day branch Voltage at both ends of the node ; Rated voltage, ; branch road The resistance, ; branch road Reactance, .

[0079] (3) Branch flow constraints: In the formula: branch road The maximum active power limit that can be tolerated. ; For reactive power limits, .

[0080] (4) Node voltage constraints: In the formula: For nodes The upper limit of voltage, ; For nodes The lower limit of voltage, .

[0081] The natural gas network model assumes that natural gas is in a steady state, the natural gas flow direction in the pipeline is fixed, and ignores pipeline inventory and other related constraints. Pipelines without compressors are considered passive pressure pipelines, while those with compressors are considered active pressure pipelines. The main operational constraints of the natural gas network include: node flow balance constraints, compressor operation constraints, natural gas pipeline gas flow constraints, node voltage constraints, and constraints on the relationship between gas flow and voltage at both ends in passive pipelines. See the following formulas for specific details: Power balance constraints, (1) Electric power balance: In the formula: The number of gas turbines; The number of fans; The number of photovoltaic power generation devices; The number of interruptible loads; The number of low-temperature waste heat power generation devices; The number of energy storage devices; The number of electric chillers; This represents the quantity of electrical loads.

[0082] (2) Thermal power balance: In the formula: The number of heat exchangers; The number of gas-fired boilers; The number of heat storage devices, For the scene Down Periodic heat load power, .

[0083] (3) Cooling power balance: Cooling power balance includes absorption cooling load power balance and electric cooling load power balance, as shown in the following two formulas: In the formula: For the scene Down Time-of-use absorption cooling load power, ; For the scene Down Electric cooling load power during the period ; The number of absorption chillers; This refers to the number of electric chillers.

[0084] In this embodiment, step S400 outputs the load recovery status of each node according to the scheduling scheme, performs probability weighting on the load recovery value under different scenarios, and obtains the distribution network resilience assessment value, including the following step D1: D1: Obtaining the distribution network resilience assessment value includes statistical analysis of load recovery status and recovery value under various typical scenarios; The recovery value is weighted based on the scenario probability distribution to obtain the overall resilience assessment value of the distribution network. Analyze the impact of changes in response intention on resilience indicators and generate sensitivity assessment results.

[0085] Specifically, the load recovery status and recovery value are statistically analyzed for each typical scenario. For the k-th typical scenario, based on the scheduling scheme obtained in step C2, the load recovery status and recovery power of each node in each time period are extracted. The load restoration status variable takes a value of 0 or 1, where 1 indicates that node n has restored power during time period t, and 0 indicates that it has not been restored. Calculate the total load restoration value under scenario k: In the formula, The load recovery value for scenario k is expressed in kWh. Let n be the load importance coefficient for node n. The duration is defined as the time period. The recovery time, recovery power ratio, and energy supply structure of key nodes in each scenario are statistically analyzed to form scenario-level resilience indicators.

[0086] The recovery value is weighted based on the scenario probability distribution to obtain the overall resilience assessment value of the distribution network. The worst-case scenario probability combination obtained in step C1 is then used. Calculate the expected load recovery value: In the formula, This is a resilience assessment value for the distribution network, representing the expected load recovery value that the system can achieve under the most unfavorable source-load distribution conditions. This indicator reflects the lower limit of the distribution network's recovery capability in extreme scenarios; a higher value indicates stronger system resilience. Furthermore, the standard deviation of the recovery value for each scenario is calculated: The standard deviation reflects the degree of fluctuation in recovery capability across different scenarios; the smaller the value, the more stable the system's performance under different scenarios.

[0087] Analyze the impact of changes in response intention on resilience indicators to generate sensitivity assessment results. This is done by changing user response intention values. The temperature regulation boundary and response frequency upper limit are recalculated, and the scheduling problem is solved again to obtain a new load recovery value. The range of variation for the response willingness value is set as follows: ,in =0.1. For each level of change in willingness to respond, repeat steps C1 and C2 to calculate the corresponding resilience assessment value. The sensitivity of the response intention to resilience is calculated as follows: In the formula, Let represent the sensitivity coefficient of user i's willingness to respond. A positive sensitivity coefficient indicates that increasing the user's willingness to respond helps improve system resilience; a larger value indicates a more critical contribution of the user to system resilience. Based on the sensitivity analysis results, key response users are identified, and incentive measures are prioritized for these users to improve the distribution network resilience in a cost-effective manner. Simultaneously, the resilience change trend under different uncertainty levels is analyzed by adjusting the overall offset threshold and the single-point extreme value threshold, as shown in the table below: Table 1. Comparison of resilience values ​​between the ∞-norm and the composite norm.

[0088] Table 2 Comparison of resilience values ​​between 1-norm and composite norm

[0089] Assessing the impact of robustness constraints on resilience indices provides a basis for conservative adjustments to scheduling strategies.

[0090] In summary, through steps S100 to S400, this method achieves a complete process from predicting user response behavior, constructing source-load uncertainty scenarios, robust optimization scheduling, to resilience assessment. S100 accurately predicts user response intentions based on the XGBoost model, transforming subjective behavioral uncertainty into temperature regulation boundaries and response frequency upper limits, providing personalized constraints for subsequent scheduling. S200 generates typical wind and solar load scenarios through Monte Carlo sampling and scenario reduction methods, introducing dual-dimensional constraints of overall offset and single-point extrema to construct a probability distribution domain, controlling the overall distribution characteristics while suppressing extreme fluctuations. S300 searches for the most unfavorable scenario probability combination within the probability distribution domain, and under this combination, solves the load recovery scheme through coordinated scheduling of cooling, heating, and power, fully utilizing the coordinated regulation capabilities of gas turbines, waste heat recovery, refrigeration equipment, and energy storage devices. S400 performs probability weighting on the recovery value of different scenarios, outputs a resilience assessment value, and performs sensitivity analysis to quantify the system's recovery capability under extreme scenarios and identify key response users.

[0091] Example 3 The above is a schematic scheme of a distribution network resilience enhancement optimization method based on response prediction. It should be noted that the technical solution of this distribution network resilience enhancement optimization system based on response prediction belongs to the same concept as the technical solution of the aforementioned distribution network resilience enhancement optimization method based on response prediction. Details not described in detail in the technical solution of the distribution network resilience enhancement optimization system based on response prediction in this embodiment can be found in the description of the technical solution of the aforementioned distribution network resilience enhancement optimization method based on response prediction.

[0092] This embodiment also provides a distribution network resilience enhancement optimization system based on response prediction, including: The response prediction module is used to acquire basic data of the distribution network, train an integrated decision tree model based on the basic data, predict the response intention values ​​of cold and heat load users, and determine the temperature regulation boundary and response frequency upper limit based on the response intention values. The scenario construction module is used to generate an initial scenario set through source load uncertainty sampling. It extracts typical scenarios and initial probability distributions through a scenario reduction method based on transportation distance. It introduces overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenarios in two dimensions, forming a probability distribution domain. The overall offset constraints limit the sum of the deviations of all scenario probabilities from the initial distribution, and the single-point extreme value constraints limit the maximum fluctuation range of the probability of a single scenario. The optimization solution module is used to search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained by the search, under the conditions of satisfying power flow constraints, hot and cold power balance constraints and temperature regulation boundary constraints, the scheduling scheme that maximizes the critical load recovery value is solved. The strategy generation module is used to output the load recovery status of each node according to the scheduling scheme, and to perform probability weighting on the load recovery value under different scenarios to obtain the distribution network resilience assessment value.

[0093] This embodiment also provides an electronic device suitable for distribution network resilience enhancement optimization based on response prediction, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the distribution network resilience enhancement optimization method based on response prediction as proposed in the above embodiment.

[0094] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the distribution network resilience enhancement optimization method based on response prediction as proposed in the above embodiments.

[0095] The storage medium proposed in this embodiment and the method for improving the resilience of distribution networks based on response prediction proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0096] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distribution network resilience enhancement optimization method based on response prediction, characterized in that: This includes acquiring basic data of the distribution network, training an integrated decision tree model based on the basic data, predicting the response willingness values ​​of users of cooling and heating loads, and determining the temperature regulation boundary and the upper limit of the response frequency based on the response willingness values; An initial scene set is generated by sampling the source load uncertainty. Typical scenes and initial probability distributions are extracted by a scene reduction method based on transportation distance. Overall offset constraints and single-point extreme value constraints are introduced to limit the probability distribution of typical scenes in two dimensions to form a probability distribution domain. The overall offset constraints limit the sum of the deviations of all scene probabilities from the initial distribution, and the single-point extreme value constraints limit the maximum fluctuation of the probability of a single scene. Search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained from the search, solve for the scheduling scheme that maximizes the critical load recovery value under the conditions of power flow constraints, hot and cold power balance constraints, and temperature regulation boundary constraints. Based on the load recovery status of each node output by the scheduling scheme, the load recovery value under different scenarios is probabilistically weighted to obtain the distribution network resilience assessment value.

2. The distribution network resilience enhancement optimization method based on response prediction as described in claim 1, characterized in that: The ensemble decision tree model trained based on basic data predicts the response intention values ​​of users with heating and cooling loads, including: The actual temperature, expected temperature, temperature deviation, and stimulus level of historical periods are used as input features, and whether users participated in the response during the corresponding period is used as a label. Multiple decision trees are trained using the gradient boosting algorithm, with each decision tree trained based on the residuals of its predecessor trees. Input the temperature status and excitation conditions of the current scheduling cycle, and output the response intention value.

3. The distribution network resilience enhancement optimization method based on response prediction as described in claim 2, characterized in that: The step of determining the temperature adjustment boundary based on the response intention value includes obtaining the upper and lower limits of the temperature comfort range for users with heating and cooling loads; By multiplying the upper and lower limits of the temperature comfort range by the response willingness value, the user's willingness to participate under the scheduling command is predicted, and the actual adjustable upper and lower limits of the temperature are obtained.

4. The distribution network resilience enhancement optimization method based on response prediction as described in claim 3, characterized in that: The introduction of overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenarios in two dimensions includes setting an overall offset threshold, requiring that the sum of the absolute values ​​of the differences between the actual probability and the initial probability of all typical scenarios be less than the overall offset threshold. Set a single-point extreme value threshold, requiring that the absolute value of the difference between the actual probability and the initial probability of any typical scenario is less than the single-point extreme value threshold; Based on the total number of wind and solar power output scenarios, the number of typical scenarios, and the target confidence level, the values ​​of the overall offset threshold and the single-point extreme value threshold are inversely derived through probability inequalities.

5. The distribution network resilience enhancement optimization method based on response prediction as described in claim 4, characterized in that: The search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain includes solving for the load recovery value corresponding to each typical scenario under a given scenario probability combination. With the goal of minimizing the probabilistic weighted sum of load restoration value, the scenario probability combination is solved under the conditions of probability normalization constraint, overall offset constraint and single-point extreme value constraint.

6. The distribution network resilience enhancement optimization method based on response prediction as described in claim 5, characterized in that: The scheduling scheme includes gas turbine output, waste heat recovery ratio, refrigeration equipment operating status and cold and heat load response status, including determining the power generation output, energy storage scheduling and load response status for each time period, and calculating the input power of waste heat boiler and low temperature waste heat power generation device based on the gas turbine's heat-to-power ratio and waste heat recovery ratio. Calculate the input heat power required to meet the absorption cooling load based on the energy efficiency coefficient of the absorption chiller. Calculate the input power required to meet the electric cooling load based on the coefficient of performance (COP) of the electric chiller. Under the constraints of electric power balance, thermal power balance, and cold power balance, determine the gas turbine output, energy storage charging and discharging power, and cold and hot load response.

7. The distribution network resilience enhancement optimization method based on response prediction as described in claim 6, characterized in that: The obtained distribution network resilience assessment value includes statistical analysis of load recovery status and recovery value under various typical scenarios; The recovery value is weighted based on the scenario probability distribution to obtain the overall resilience assessment value of the distribution network. Analyze the impact of changes in response intention on resilience indicators and generate sensitivity assessment results.

8. A distribution network resilience enhancement optimization system based on response prediction, based on the distribution network resilience enhancement optimization method based on response prediction as described in any one of claims 1 to 7, characterized in that: It also includes a response prediction module, which is used to acquire basic data of the distribution network, train an integrated decision tree model based on the basic data, predict the response intention values ​​of cold and heat load users, and determine the temperature regulation boundary and response frequency upper limit based on the response intention values; The scenario construction module is used to generate an initial scenario set through source load uncertainty sampling. It extracts typical scenarios and initial probability distributions through a scenario reduction method based on transportation distance. It introduces overall offset constraints and single-point extreme value constraints to limit the probability distribution of typical scenarios in two dimensions, forming a probability distribution domain. The overall offset constraints limit the sum of the deviations of all scenario probabilities from the initial distribution, and the single-point extreme value constraints limit the maximum fluctuation range of the probability of a single scenario. The optimization solution module is used to search for the scenario probability combination that minimizes the load recovery value within the probability distribution domain. For the scenario probability combination obtained by the search, under the conditions of satisfying power flow constraints, hot and cold power balance constraints and temperature regulation boundary constraints, the scheduling scheme that maximizes the critical load recovery value is solved. The strategy generation module is used to output the load recovery status of each node according to the scheduling scheme, and to perform probability weighting on the load recovery value under different scenarios to obtain the distribution network resilience assessment value.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the distribution network resilience enhancement optimization method based on response prediction as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the distribution network resilience enhancement optimization method based on response prediction as described in any one of claims 1 to 7.

Images