Power distribution network fault self-healing time sequence decision-making method and system based on new energy fluctuation

Abstract

The invention discloses a power distribution network fault self-healing time sequence decision-making method and system based on new energy fluctuation, and relates to the technical field of intelligent power distribution networks, and the method comprises the steps: detecting a power distribution network fault in real time, executing an isolation operation, collecting new energy output data, load data and network topology information, constructing a dynamic evolution model, and calculating the power distribution network fault self-healing time sequence decision-making. And generating a self-healing operation sequence based on a decision framework associated with a time sequence, optimizing load recovery and operation cost, adaptively adjusting the operation sequence according to real-time state fluctuation, executing the adjusted self-healing operation, and updating decision parameters. According to the method, the new energy fluctuation model fusing the space-time characteristics is constructed, accurate composite state representation is established, and a method of combining multi-stage optimization and adaptive robust decision is adopted, so that the decision quality in a high-uncertainty environment is effectively improved, and the problem of strategy failure in an extreme fluctuation scene of a traditional method is solved; and a more reliable fault self-healing solution is provided for the high-proportion new energy power distribution network.

Description

Technical Field

[0001] This invention relates to the field of smart distribution network technology, specifically to a method and system for time-series decision-making on self-healing of distribution network faults under new energy fluctuations. Background Art

[0002] Traditional self-healing technologies for distribution networks are mainly based on deterministic network models and static decision-making frameworks, using heuristic rules or deterministic optimization algorithms to generate recovery strategies. In recent years, scholars at home and abroad have made some progress in the field of dynamic decision-making, such as distributed cooperative control based on multi-agent systems and self-healing strategy optimization combined with deep reinforcement learning. However, existing technologies mostly assume that the output of new energy sources can be accurately predicted, or only describe its uncertainty through probability distribution, failing to effectively address the impact of the spatiotemporal correlation fluctuation characteristics on decision-making under high-proportion new energy access. In terms of time-series decision-making, current research mainly adopts staged optimization with a fixed time scale, lacking the ability to dynamically respond to real-time fluctuations, and most methods do not incorporate network reconfiguration, generation scheduling, and energy storage control into a unified optimization framework, resulting in insufficient decision-making coordination.

[0003] Existing technologies suffer from three significant drawbacks: First, the modeling of new energy output is overly simplified, failing to fully consider the spatiotemporal correlation of wind and solar power and their coupling with load status, resulting in insufficient accuracy in state prediction. Second, the processing of temporal correlations in self-healing decision-making is inadequate; traditional dynamic programming algorithms struggle to handle high-dimensional state spaces, while approximation algorithms lack refined optimization of key operational nodes. Finally, existing adaptive mechanisms largely rely on preset rules to adjust strategies, failing to dynamically correct decisions based on real-time risk assessment. In contrast, this invention constructs a new energy fluctuation model that integrates spatiotemporal characteristics, establishes an accurate composite state representation, and employs a method combining multi-stage optimization and adaptive robust decision-making, effectively improving decision-making quality under high uncertainty environments. Summary of the Invention

[0004] In view of the above-mentioned existing problems, the present invention provides a method and system for self-healing timing decision-making of distribution network faults under new energy fluctuations, in order to solve the problems in the prior art that do not fully consider the spatiotemporal correlation of wind power and photovoltaics and their coupling relationship with load status, resulting in insufficient state prediction accuracy, difficulty in dealing with high-dimensional state space, lack of fine optimization of key operation nodes, and inability to dynamically correct decisions based on real-time risk assessment.

[0005] To address the aforementioned technical issues, a method for time-series decision-making based on self-healing of distribution network faults under renewable energy fluctuations is proposed, including:

[0006] Real-time detection of distribution network faults and execution of isolation operations; collection of renewable energy output data, load data and network topology information; construction of a dynamic evolution model; generation of self-healing operation sequences based on a time-series correlation decision framework to optimize load recovery and operation costs; adaptive adjustment of operation sequences according to real-time state fluctuations, execution of adjusted self-healing operations, and updating of decision parameters.

[0007] As a preferred embodiment of the method for time-series decision-making of distribution network fault self-healing under new energy fluctuations described in this invention, the construction of the dynamic evolution model includes: extracting the spatiotemporal correlation characteristics of new energy output, establishing an output fluctuation model, coupling the load state change trend with network topology change, quantifying the uncertainty of system state, predicting new energy output through a multi-model integration framework, and dynamically updating the prediction weights.

[0008] As a preferred embodiment of the method for time-series decision-making of distribution network fault self-healing under new energy fluctuations described in this invention, the construction of the dynamic evolution model further includes defining a composite state vector containing network topology, equipment status and load power supply status, generating a state transition probability function based on the dynamic evolution model, and designing a reward function that integrates load recovery amount, operating cost and safety margin.

[0009] As a preferred embodiment of the method for self-healing timing decision-making of distribution network faults under new energy fluctuations described in this invention, the method for generating a self-healing operation sequence includes inputting a composite state vector, a state transition probability function, and a reward function into a decision framework, using a multi-stage optimization algorithm to solve for the optimal operation sequence, and using a priority mechanism to select the operation combination with the highest return and lowest risk.

[0010] As a preferred embodiment of the method for self-healing timing decision-making of distribution network faults under new energy fluctuations described in this invention, the multi-stage optimization algorithm includes: the first stage of optimizing the switching operation sequence and reconstructing the network topology; the second stage of optimizing the distributed generation scheduling strategy under the reconstructed topology; and the final stage of refining the operation parameters of key nodes.

[0011] The multi-stage optimization algorithm also includes: using a value-iteration-based dynamic programming algorithm to solve for the optimal self-healing operation sequence; defining a state-value function; evaluating the long-term reward of taking different actions in each state; obtaining the optimal strategy based on the state-value function, i.e., choosing the action that maximizes the expected reward in each state; introducing an approximate dynamic programming method; using feature functions to extract representative information of the state; approximating the state-value function by learning weight parameters; introducing temporal difference learning to update parameters; and introducing a priority-based experience replay mechanism to prioritize experience samples.

[0012] Define the state-value function and obtain the optimal policy as follows:

[0013]

[0014]

[0015] Where V(s) is the value function of state s, i.e., the maximum expected reward starting from state s, a is the action parameter, R(s,a) is the immediate reward obtained by performing action a in state s, γ is the discount factor, P(s'|s,a) is the probability of transitioning to state s' from state s by performing action a, V(s') is the value function of state s', and π * (s) represents the optimal strategy in state s, i.e., choosing the action with the highest expected reward;

[0016] The approximate dynamic programming method is expressed as:

[0017]

[0018] in, Let m be an approximate function of state s, and φ be the number of features. i (s) is the i-th feature function, where i is the variable index, and θ i The weight parameters are those corresponding to the i-th feature;

[0019] The formula for temporal difference learning is expressed as:

[0020]

[0021] Where, θ t+1 For the updated weight parameter vector, θ t This is the current weight parameter vector. Let r be an approximate function of t+1. t Let α be the instantaneous reward obtained at time step t, and α be the learning rate. Let be an approximate function at time step t. Approximate function Regarding state s t The gradient;

[0022] The formula for the experience replay mechanism is expressed as follows:

[0023] p j =|δ j |+o

[0024] Where, p j Let δ be the priority of the j-th empirical sample. j Let be the temporal difference error of the j-th empirical sample, and o be a small positive constant.

[0025] As a preferred embodiment of the method for self-healing timing decision-making of distribution network faults under new energy fluctuations described in this invention, the adaptive adjustment operation sequence includes: calculating the confidence interval of the operation sequence, combining real-time state fluctuation correction decision, constructing the new energy output prediction error boundary, generating a robust operation strategy, and dynamically adjusting the operation sequence based on a risk assessment model.

[0026] The risk assessment model includes quantifying the impact of new energy fluctuations on the probability of operational failure, integrating equipment status parameters to calculate a risk index, and adjusting the operational threshold based on the risk aversion coefficient.

[0027] The adaptive adjustment operation sequence also includes verifying the robustness of the operation during the topology reconfiguration phase and introducing energy storage system control variables during the power generation scheduling phase, and updating the decision parameters through time-series iteration.

[0028] The formula for calculating the confidence interval of an operation sequence is expressed as:

[0029]

[0030] Among them, s t Let a be the state at time step t. t For the action at time step t, CI(s) t ,a t () represents the confidence interval estimate. σ is the estimated expected reward. R To estimate the standard deviation, κ is the confidence level parameter;

[0031] Operation selection based on confidence intervals is represented as follows:

[0032]

[0033] in, For the optimal action at time step t, Q(s) t Let a) be the value function of action a. Let A be the number of times the state-action pair is accessed, 'a' be the action parameter, 'λ' be the exploration parameter, and 't' be the time step. t Let t be the set of actions at time step t;

[0034] The formula for generating a robust operation strategy is expressed as follows:

[0035]

[0036] Where, π robust (s t ) represents a robust adaptive policy, U(s) t a) is the set of successor states considering uncertainties, s t+1 For the state at time step t+1, R(s) ta) is in state s t The immediate reward gained from performing action a, V(s) t+1 ) represents state s t+1 The value function of , where γ is the discount factor;

[0037] The formula for constructing the error boundary of new energy power output prediction is expressed as follows:

[0038]

[0039] Among them, U(s) t a) represents the set of successor states after considering uncertainties, f(s) t ,a,H RE H is the state transition function. RE Actual contribution to new energy sources For the predicted output of new energy sources, ΔH is the prediction error boundary, and ∈ indicates a membership relationship;

[0040] The risk assessment model for self-healing operations is expressed as follows:

[0041]

[0042] Prob(e|s t ,a t ) = F risk (s t ,a t ,ΔH RE )

[0043] Among them, Risk(s) t ,a t ) represents state s t Take action a t The risk value, E, is the set of potential risk events, Prob(e|s) t ,a t ) is in state s t Take action a t The conditional probability of risk event e occurring, Impact(e) is the degree of influence of event e, e is the index of the risk event variable, and F risk (s t ,a t ,ΔH RE Let ΔH be the risk function. RE Fluctuations in the output of new energy sources;

[0044] Decision adjustments based on risk assessment are expressed as follows:

[0045]

[0046] Where Q(s) tLet a) be the value function of action a. The optimal action selected after risk adjustment under state s, A t Let be the set of actions at time step t, μ be the risk aversion coefficient, and Risk(s) be the risk aversion coefficient. t a) represents state s t The risk value of taking action 'a'.

[0047] As a preferred embodiment of the method for self-healing timing decision-making of distribution network faults under new energy fluctuations described in this invention, the updated decision parameters include: collecting system state data after self-healing operation, correcting the state transition probability function and reward function weights, iteratively optimizing the characteristic function parameters, incorporating the energy storage system control vector into the self-healing operation sequence, restoring power supply in stages according to load importance weights, and pre-verifying the feasibility of the operation through Monte Carlo simulation.

[0048] As a preferred embodiment of the system for self-healing timing decision-making of distribution network faults under new energy fluctuations as described in this invention, it is characterized by including a state assessment and prediction module, a self-healing decision optimization module, an adaptive adjustment module, and an execution and update module.

[0049] The state assessment and prediction module is used to establish a new energy output fluctuation model, quantify the impact of random disturbances, predict load change trends and network state evolution, and generate a composite state vector containing voltage amplitude, phase angle and switching state.

[0050] The self-healing decision optimization module is used to construct a decision framework that includes network topology, device status and load power supply, design a reward mechanism that integrates load recovery, operating costs and system security, and generate operation sequences through multi-stage optimization.

[0051] The adaptive adjustment module is used to calculate the confidence interval of the operation sequence, respond to real-time fluctuations in new energy, construct the prediction error boundary, generate a robust operation strategy, quantify the operation risk index, and dynamically adjust the execution threshold.

[0052] The execution and update module is used to execute the final self-healing operation sequence and monitor the effect, and verify the robustness of the scheme through Monte Carlo simulation, and update the core parameters in a closed loop.

[0053] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method for making self-healing timing decisions for distribution network faults based on new energy fluctuations.

[0054] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the method for making a timing decision on self-healing faults in a distribution network under new energy fluctuations are implemented.

[0055] The beneficial effects of this invention are as follows: By detecting and isolating distribution network faults in real time, and combining this with the dynamic acquisition of renewable energy output, load, and topology data, a precise dynamic evolution model is constructed, enabling rapid fault location and isolation, effectively preventing fault propagation. Simultaneously, spatiotemporal correlation modeling significantly improves the prediction accuracy of renewable energy output. Based on composite state vectors, state transition probabilities, and reward functions, a multi-stage optimization algorithm generates a self-healing operation sequence, handling topology reconfiguration, generation scheduling, and energy storage control step by step. A priority mechanism is used to select high-return, low-risk operation combinations, significantly optimizing decision-making efficiency and economy. By dynamically adjusting the operation sequence through confidence interval estimation, prediction error boundaries, and risk assessment models, the robustness and stability of the system under renewable energy fluctuations are enhanced, ensuring the effectiveness of the strategy under extreme conditions. Through closed-loop updating of model parameters and graded power restoration according to load priority, continuous parameter optimization and efficient resource allocation are achieved, and energy storage collaborative control further improves power supply continuity. Attached Figure Description

[0056] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart illustrating the overall process of a method for self-healing timing decision-making of distribution network faults under new energy fluctuations, as provided in one embodiment of the present invention.

[0058] Figure 2 The flowchart of a system scheme for a distribution network fault self-healing timing decision based on new energy fluctuations is provided as an embodiment of the present invention. Detailed Implementation

[0059] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0060] 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.

[0061] 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 an embodiment that is mutually exclusive, either alone or selectively, with other embodiments.

[0062] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0063] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0064] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0065] Example 1, referring to Figure 1 This is the first embodiment of the present invention, which provides a method for time-series decision-making on self-healing faults in distribution networks under new energy fluctuations, including:

[0066] S1: Real-time detection of power distribution network faults and execution of isolation operations, collection of new energy output data, load data and network topology information, and construction of dynamic evolution model.

[0067] Furthermore, the construction of the dynamic evolution model includes extracting the spatiotemporal correlation characteristics of new energy output, establishing an output fluctuation model, coupling the load state change trend with network topology changes, quantifying system state uncertainty, predicting new energy output through a multi-model integration framework, and dynamically updating the prediction weights.

[0068] The power output fluctuation model includes establishing a photovoltaic power output model (including irradiance and temperature coefficient) and a wind power output model (including cut-in / rated / cut-out wind speeds), quantifying the impact of random disturbances, defining a composite state vector that includes voltage amplitude, phase angle and switching state, using a time-series correlation model to predict future states, calculating the covariance matrix of state prediction, and integrating the new energy power output prediction error and process noise.

[0069] The power output fluctuation model of new energy sources is expressed as follows:

[0070] H RE (t)=P rated ·u(w(t),s(t),ε(t))

[0071] Among them, H RE (t) represents the actual output of the new energy source at time t, P rated Let u be the rated capacity, w(t) be the output characteristic function, s(t) be the meteorological parameter vector at time step t, s(t) be the system state vector, and ε(t) be the random disturbance term.

[0072] The photovoltaic power output fluctuation model is expressed as:

[0073] P PV (t)=η·S·G'(t)·[1-α1(T'(t)-T ref )]·ξ(t)

[0074] Among them, P PV (t) represents photovoltaic output, S represents panel area, η represents photoelectric conversion efficiency, G'(t) represents irradiance, α1 represents temperature coefficient, T'(t) represents panel temperature, and T ref The reference temperature is ξ(t), and the random fluctuation factor is ξ(t).

[0075] The wind power output fluctuation model is represented as follows:

[0076]

[0077] Among them, P WT v represents the output of wind power, v represents the wind speed, v in To cut off the wind speed, v rated For the rated wind speed, v out To cut off the wind speed, P rated ζ(t) represents the rated power, and ζ(t) represents the random fluctuation factor.

[0078] To improve the accuracy of new energy power output prediction, an ensemble learning framework is used to integrate prediction methods for power output prediction. The formula is as follows:

[0079]

[0080] in, This represents the predicted output value of new energy sources after time step Δt. For the output of the k-th prediction model, ω k Here, n represents the total number of prediction models, k represents the variable index, and g represents the weights of the prediction models. k (t) represents the prediction error of the k-th model, and α represents the learning rate.

[0081] Furthermore, the construction of the dynamic evolution model also includes defining a composite state vector that includes network topology, device status, and load power supply status, generating a state transition probability function based on the dynamic evolution model, and designing a reward function that integrates load recovery amount, operating cost, and safety margin.

[0082] In the embodiments of this application, the construction of the dynamic evolution model includes: adopting an improved spatiotemporal correlation model, considering cloud movement and irradiance changes, establishing a photoelectric conversion efficiency and temperature correction model, modeling wind speed characteristics based on Weibull distribution, establishing an output model in combination with power curves, integrating three prediction methods: ARIMA, LSTM and physical model, and dynamically adjusting model weights using an adaptive weight update mechanism.

[0083] In one optional implementation, the construction of the dynamic evolution model includes constructing a spatiotemporal graph neural network, using the geographical location of new energy power stations and meteorological monitoring data as node features, calculating the spatiotemporal correlation strength based on the Pearson correlation coefficient, and the prediction framework including a graph attention network to extract spatial features, a TCN network to capture temporal dependencies, incorporating physical constraint corrections, and using a sliding window-updated generative adversarial network (GAN) to compensate for prediction biases.

[0084] In another optional implementation, the construction of the dynamic evolution model includes establishing a wind-solar resource field-grid coupling model, outputting high-precision wind speed and irradiance data in the meteorological WRF model, considering the dynamic characteristics of wind turbine pitch angle and photovoltaic panel tilt angle in the equipment layer, and performing dynamic topology modeling based on the admittance matrix in the grid layer. The polynomial chaotic expansion (PCE) method is adopted, and a joint probability density function is constructed for non-qualitative quantification.

[0085] Distribution network condition assessment is based on generalized condition estimation theory, taking into account the uncertainty of measurement data and the impact of topology changes. A state vector is defined, which includes the voltage magnitude and phase angle of all nodes, as well as the state variables of all switches. A condition estimation model is established to process the information, linking the actual measured data with the theoretical values ​​calculated through the state vector, and introducing a measurement error vector to quantify the inaccuracy of observations. To obtain the optimal condition estimation result, an objective function is constructed to minimize the error between the measured and theoretical values, and a regularization term is used to smooth the condition estimation result. A time-series correlated model is used for condition prediction. To comprehensively evaluate the reliability of the prediction results, it is also necessary to quantify the uncertainty of the prediction.

[0086] The state estimation model considering the fluctuations of new energy sources is expressed as follows:

[0087] z = h(X) + τ

[0088] Where z is the measurement vector, h is the measurement equation, X is the input vector, and τ is the measurement error vector;

[0089] The objective function for state estimation is expressed as:

[0090]

[0091] in, Let Ω(X) be the objective function for state estimation, Z be the regularization term, and Z be the measurement error covariance matrix. Here, T is the regularization parameter, and T is the transpose operation;

[0092] The uncertainty of state prediction is quantitatively represented as:

[0093]

[0094] Where, Σ X (t+Δt) is the state covariance matrix at time step t+Δt, J X Let Σ be the Jacobian matrix for the state transition. X J(t) is the state covariance matrix at time step t. P Σ is the Jacobian matrix of the power output prediction error. P (t+Δt) is the output prediction error covariance matrix at time step t+Δt, and O is the process noise covariance matrix.

[0095] S2: A decision framework based on temporal correlation generates a self-healing operation sequence to optimize load recovery and operation costs.

[0096] In this embodiment of the application, generating the self-healing operation sequence includes inputting a composite state vector, a state transition probability function, and a reward function into a decision framework, using a multi-stage optimization algorithm to solve for the optimal operation sequence, and using a priority mechanism to select the operation combination with the highest return and lowest risk.

[0097] In one optional implementation, the generation of the self-healing operation sequence includes dividing the system into three levels that work collaboratively. The high-level strategy focuses on the topology reconfiguration of the distribution network, with the action space consisting of switch combinations, and guides decisions through a reward function that combines the load recovery degree and the number of operations. The middle-level strategy is responsible for the scheduling of distributed generation (DG), adjusting the output of DG within the action space, and ensuring that all scheduling decisions meet the actual constraints by introducing a projection layer to obtain a feasible solution. At the lowest level, the strategy is implemented based on the TD3 (Dual Delay Deep Deterministic Policy Gradient) algorithm, performing millisecond-level fine-tuning of parameters.

[0098] In another alternative implementation, the generation of the self-healing operation sequence includes establishing a two-layer optimization model, the objective function of which aims to minimize the active power loss of the power grid (measured by the L1 norm) and reduce the number of switching operations (measured by the L0 norm, where λ is a weighting coefficient balancing the relationship between the two), and an event triggering mechanism is introduced to improve efficiency and adapt to dynamic changes. At the same time, the powerful parallel computing capabilities of the GPU are used to process the optimization part involving continuous variables, so as to achieve fast response and efficient solution.

[0099] Furthermore, in the self-healing decision-making problem of distribution network faults, a Markov Decision Process (MDP) is used for modeling, including a state space, action space, state transition probability function, reward function, and discount factor. The design of the state space comprehensively considers the network topology state, equipment operating state vector, load power supply state vector, and renewable energy output state, fully reflecting the real-time operation of the system. The action space includes all self-healing operations, specifically including switch operation vectors, distributed generation adjustment vectors, and energy storage system control vectors, providing diverse choices for decision-making. The state transition probability function specifically considers the volatility of renewable energy output, more accurately predicting changes in the system state.

[0100] The reward function formula is expressed as:

[0101] R(s t ,a t )=b1·R load (s t ,a t )-b2·C op (a t )+b3·R rel (s t ,a t )

[0102]

[0103] Wherein, R(s) t ,a t R is the total reward function. load As a reward for load recovery, C op For operating costs, s t Let a be the state at time step t. t For the action at time step t, R rel (s t ,a t Reliability rewards, where b1, b2, and b3 are weighting coefficients, ρ d (s t ) represents the load point d in state s t Under the power supply state, ρ d (s t+1 ) represents the load point d in state s t+1 Under the power supply state, ω d n represents the importance weight of the load. L c is the total number of composite points. sw,x Let n be the operating cost coefficient for switch x. sv Let a be the number of switches. sw,x,t For the operation of switch x at time step t, n DG c represents the number of distributed generation units. DG,y Let be the generation dispatch cost coefficient of distributed generation unit y, be the generation power of distributed generation unit y at time step t, β1 and β2 are the weighting coefficients of reliability bonus, and I' q Let I' be the actual current in line q. q,mex V' is the maximum allowable current of line q. ξ V' is the actual voltage at node ξ. ξ,nem V' is the rated voltage at node ξ. nem Let d be the system's rated voltage, and q, ξ, and y be variable indices.

[0104] The reward function is designed to comprehensively balance the load restoration, operating costs, and system reliability. It is implemented through three parts: load restoration reward, operating cost reward, and reliability reward, with weighting coefficients used to adjust the importance of each part. The load restoration reward is calculated based on the power supply status and importance weight of the load point, the operating cost covers the switching operation cost and generation dispatch cost, and the reliability reward is measured by the safety margin of the system state, including deviations in line current and node voltage. Through the MDP model, the self-healing decision-making process of the distribution network can be effectively optimized, improving the stability and economy of the system.

[0105] Furthermore, the multi-stage optimization algorithm includes: a first stage of optimizing the switching operation sequence and reconstructing the network topology; a second stage of optimizing the distributed generation scheduling strategy under the reconstructed topology; and a final stage of refining the operation parameters of key nodes.

[0106] The multi-stage optimization algorithm also includes: using a value-iteration-based dynamic programming algorithm to solve for the optimal self-healing operation sequence; defining a state-value function; evaluating the long-term reward of taking different actions in each state; obtaining the optimal strategy based on the state-value function, i.e., choosing the action that maximizes the expected reward in each state; introducing an approximate dynamic programming method; using feature functions to extract representative information of the state; approximating the state-value function by learning weight parameters; introducing temporal difference learning to update parameters; and introducing a priority-based experience replay mechanism to prioritize experience samples.

[0107] Define the state-value function and obtain the optimal policy as follows:

[0108]

[0109]

[0110] Where V(s) is the value function of state s, i.e., the maximum expected reward starting from state s, a is the action parameter, R(s,a) is the immediate reward obtained by performing action a in state s, γ is the discount factor, P(s'|s,a) is the probability of transitioning to state s' from state s by performing action a, V(s') is the value function of state s', and π * (s) represents the optimal strategy in state s, i.e., choosing the action with the highest expected reward;

[0111] The approximate dynamic programming method is expressed as:

[0112]

[0113] in, Let m be an approximate function of state s, and φ be the number of features. i (s) is the i-th feature function, where i is the variable index, and θ i The weight parameters are those corresponding to the i-th feature;

[0114] The formula for temporal difference learning is expressed as:

[0115]

[0116] Where, θ t+1 For the updated weight parameter vector, θ t This is the current weight parameter vector. Let r be an approximate function of t+1. tLet α be the instantaneous reward obtained at time step t, and α be the learning rate. Let be an approximate function at time step t. Approximate function Regarding state s t The gradient;

[0117] The formula for the experience replay mechanism is expressed as follows:

[0118] p j =|δ j |+o

[0119] Where, p j Let δ be the priority of the j-th empirical sample. j Let be the temporal difference error of the j-th empirical sample, and o be a small positive constant.

[0120] S3: Adaptively adjust the operation sequence based on real-time state fluctuations, execute the adjusted self-healing operation, and update the decision parameters.

[0121] In this embodiment of the application, the adaptive adjustment of the operation sequence includes calculating the confidence interval of the operation sequence, combining real-time state fluctuation correction decisions, constructing the new energy output prediction error boundary, generating a robust operation strategy, and dynamically adjusting the operation sequence based on a risk assessment model.

[0122] In one optional implementation, the adaptive adjustment operation sequence includes defining the neighborhood and fuzzy set form of the new energy output distribution using Wasserstein distance, using sparse bar decision-making, updating the fuzzy set radius every 5 minutes, and automatically adjusting parameters according to the risk exposure.

[0123] In another optional implementation, the adaptive adjustment operation sequence includes constructing a digital twin model and performing a forward-looking strategy evaluation, and simultaneously using the TOPSIS method for multi-objective decision-making when the recovery rate is >95% and the risk is <0.3.

[0124] Furthermore, the risk assessment model includes quantifying the impact of new energy fluctuations on the probability of operational failure, integrating equipment status parameters to calculate a risk index, and adjusting the operational threshold based on the risk aversion coefficient.

[0125] The adaptive adjustment operation sequence also includes verifying the robustness of the operation during the topology reconfiguration phase and introducing energy storage system control variables during the power generation scheduling phase, and updating the decision parameters through time-series iteration.

[0126] The formula for calculating the confidence interval of an operation sequence is expressed as:

[0127]

[0128] Among them, st Let a be the state at time step t. t For the action at time step t, CI(s) t ,a t () represents the confidence interval estimate. σ is the estimated expected reward. R To estimate the standard deviation, κ is the confidence level parameter;

[0129] Operation selection based on confidence intervals is represented as follows:

[0130]

[0131] in, For the optimal action at time step t, Q(s) t Let a) be the value function of action a. Let A be the number of times the state-action pair is accessed, 'a' be the action parameter, 'λ' be the exploration parameter, and 't' be the time step. t Let t be the set of actions at time step t;

[0132] The formula for generating a robust operation strategy is expressed as follows:

[0133]

[0134] Where, π robust (s t ) represents a robust adaptive policy, U(s) t a) is the set of successor states considering uncertainties, s t+1 For the state at time step t+1, R(s) t a) is in state s t The immediate reward gained from performing action a, V(s) t+1 ) represents state s t+1 The value function of , where γ is the discount factor;

[0135] The formula for constructing the error boundary of new energy power output prediction is expressed as follows:

[0136]

[0137] Among them, U(s) t a) represents the set of successor states after considering uncertainties, f(s) t ,a,H RE H is the state transition function. RE Actual contribution to new energy sources For the predicted output of new energy sources, ΔH is the prediction error boundary, and ∈ indicates a membership relationship;

[0138] The risk assessment model for self-healing operations is expressed as follows:

[0139]

[0140] Prob(e|s t ,a t ) = F risk (s t ,a t ,ΔH RE )

[0141] Among them, Risk(s) t ,a t ) represents state s t Take action a t The risk value, E, is the set of potential risk events, Prob(e|s) t ,a t ) is in state s t Take action a t The conditional probability of risk event e occurring, Impact(e) is the degree of influence of event e, e is the index of the risk event variable, and F risk (s t ,a t ,ΔH RE Let ΔH be the risk function. RE Fluctuations in the output of new energy sources;

[0142] Decision adjustments based on risk assessment are expressed as follows:

[0143]

[0144] Where Q(s) t Let a) be the value function of action a. The optimal action selected after risk adjustment under state s, A t Let be the set of actions at time step t, μ be the risk aversion coefficient, and Risk(s) be the risk aversion coefficient. t a) represents state s t The risk value of taking action 'a'.

[0145] Furthermore, the updated decision parameters include collecting system state data after the self-healing operation, correcting the state transition probability function and reward function weights, iteratively optimizing the characteristic function parameters, incorporating the energy storage system control vector into the self-healing operation sequence, restoring power supply in stages according to the load importance weights, and pre-verifying the feasibility of the operation through Monte Carlo simulation.

[0146] The corrected formulas for the state transition probability function and reward function weights are expressed as follows:

[0147]

[0148] Where B(s'|s,a) is the probability of transitioning to state s' after performing action a in the current state s, L is the number of samples, I is the indicator function, and s l The current state at the l-th sampling time is s. l,next For the next state after the l-th sampling, a l For the action during the l-th sampling, For the updated parameter vector, Let α be the current parameter vector, and α be the learning rate. For the objective function For parameters The gradient, M is the number of trajectories, T t For time steps, For state s t,t Next, execute action a t,t The probability is the discounted reward from time step t to the end. For action probability pairs of parameters The gradient.

[0149] 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.

[0150] Example 2, refer to Figure 2 This is the second embodiment of the present invention. This embodiment provides a system for self-healing timing decision-making of distribution network faults under new energy fluctuations, including a state assessment and prediction module, a self-healing decision optimization module, an adaptive adjustment module, and an execution and update module.

[0151] The state assessment and prediction module is used to establish a new energy output fluctuation model, quantify the impact of random disturbances, predict load change trends and network state evolution, and generate a composite state vector containing voltage amplitude, phase angle and switching state.

[0152] The self-healing decision optimization module is used to construct a decision framework that includes network topology, device status and load power supply, design a reward mechanism that integrates load recovery, operating costs and system security, and generate operation sequences through multi-stage optimization.

[0153] The adaptive adjustment module is used to calculate the confidence interval of the operation sequence, respond to real-time fluctuations in new energy, construct the prediction error boundary, generate a robust operation strategy, quantify the operation risk index, and dynamically adjust the execution threshold.

[0154] The execution and update module is used to execute the final self-healing operation sequence and monitor the effect, and verify the robustness of the scheme through Monte Carlo simulation, and update the core parameters in a closed loop.

[0155] 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.

[0156] Example 3, the third embodiment of the present invention, differs from the previous two embodiments in that:

[0157] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0158] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0159] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0160] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

Claims

1. A method for time-series decision-making on self-healing faults in distribution networks under renewable energy fluctuations, characterized in that: include, Real-time detection of power distribution network faults and execution of isolation operations; collection of new energy output data, load data and network topology information; and construction of dynamic evolution models. A decision framework based on temporal correlation generates a self-healing operation sequence to optimize load recovery and operating costs. The operation sequence is adaptively adjusted based on real-time state fluctuations, the adjusted self-healing operation is executed, and the decision parameters are updated.

2. The method for time-series decision-making of distribution network fault self-healing under new energy fluctuations as described in claim 1, characterized in that: The construction of the dynamic evolution model includes extracting the spatiotemporal correlation characteristics of new energy output, establishing an output fluctuation model, coupling the load state change trend with network topology changes, quantifying system state uncertainty, predicting new energy output through a multi-model integration framework, and dynamically updating the prediction weights.

3. The method for time-series decision-making on self-healing of distribution network faults under new energy fluctuations as described in claim 2, characterized in that: The construction of the dynamic evolution model also includes defining a composite state vector that includes network topology, device status and load power supply status, generating a state transition probability function based on the dynamic evolution model, and designing a reward function that integrates load recovery amount, operating cost and safety margin.

4. The method for time-series decision-making of distribution network fault self-healing under new energy fluctuations as described in claim 3, characterized in that: The process of generating a self-healing operation sequence involves inputting a composite state vector, a state transition probability function, and a reward function into a decision framework, using a multi-stage optimization algorithm to solve for the optimal operation sequence, and using a priority mechanism to select the operation combination with the highest return and lowest risk.

5. The method for time-series decision-making of distribution network fault self-healing under new energy fluctuations as described in claim 4, characterized in that: The multi-stage optimization algorithm includes: the first stage optimizes the switching operation sequence and reconstructs the network topology; the second stage optimizes the distributed generation scheduling strategy under the reconstructed topology; and the final stage refines the operation parameters of key nodes. The multi-stage optimization algorithm also includes: using a value-iteration-based dynamic programming algorithm to solve for the optimal self-healing operation sequence; defining a state-value function; evaluating the long-term reward of taking different actions in each state; obtaining the optimal strategy based on the state-value function, i.e., choosing the action that maximizes the expected reward in each state; introducing an approximate dynamic programming method; using feature functions to extract representative information of the state; approximating the state-value function by learning weight parameters; introducing temporal difference learning to update parameters; and introducing a priority-based experience replay mechanism to prioritize experience samples. Define the state-value function and obtain the optimal policy as follows: Where V(s) is the value function of state s, i.e., the maximum expected reward starting from state s, a is the action parameter, R(s,a) is the immediate reward obtained by performing action a in state s, γ is the discount factor, P(s'|s,a) is the probability of transitioning to state s' from state s by performing action a, V(s') is the value function of state s', and π * (s) represents the optimal strategy in state s, i.e., choosing the action with the highest expected reward; The approximate dynamic programming method is expressed as: in, Let m be an approximate function of state s, and φ be the number of features. i (s) is the i-th feature function, where i is the variable index, and θ i The weight parameters are those corresponding to the i-th feature; The formula for temporal difference learning is expressed as: Where, θ t+1 For the updated weight parameter vector, θ t This is the current weight parameter vector. Let r be an approximate function of t+1. t Let α be the instantaneous reward obtained at time step t, and α be the learning rate. Let be an approximate function at time step t. Approximate function Regarding state s t The gradient; The formula for the experience replay mechanism is expressed as follows: p j =|δ j |+o Where, p j Let δ be the priority of the j-th empirical sample. j Let be the temporal difference error of the j-th empirical sample, and o be a small positive constant.

6. The method for time-series decision-making of distribution network fault self-healing under new energy fluctuations as described in claim 5, characterized in that: The adaptive adjustment operation sequence includes calculating the confidence interval of the operation sequence, making correction decisions based on real-time state fluctuations, constructing the error boundary for new energy output prediction, generating a robust operation strategy, and dynamically adjusting the operation sequence based on a risk assessment model. The risk assessment model includes quantifying the impact of new energy fluctuations on the probability of operational failure, integrating equipment status parameters to calculate a risk index, and adjusting the operational threshold based on the risk aversion coefficient. The adaptive adjustment operation sequence also includes verifying the robustness of the operation during the topology reconfiguration phase and introducing energy storage system control variables during the power generation scheduling phase, and updating the decision parameters through time-series iteration. The formula for calculating the confidence interval of an operation sequence is expressed as: Among them, s t Let a be the state at time step t. t For the action at time step t, CI(s) t ,a t () represents the confidence interval estimate. σ is the estimated expected reward. R To estimate the standard deviation, κ is the confidence level parameter; Operation selection based on confidence intervals is represented as follows: in, For the optimal action at time step t, Q(s) t Let a) be the value function of action a. Let A be the number of times the state-action pair is accessed, 'a' be the action parameter, 'λ' be the exploration parameter, and 't' be the time step. t Let t be the set of actions at time step t; The formula for generating a robust operation strategy is expressed as follows: Where, π robust (s t ) represents a robust adaptive policy, U(s) t a) is the set of successor states considering uncertainties, s t+1 For the state at time step t+1, R(s) t a) is in state s t The immediate reward gained from performing action a, V(s) t+1 ) represents state s t+1 The value function of , where γ is the discount factor; The formula for constructing the error boundary of new energy power output prediction is expressed as follows: Among them, U(s) t a) represents the set of successor states after considering uncertainties, f(s) t ,a,H RE H is the state transition function. RE Actual contribution to new energy sources For the predicted output of new energy sources, ΔH is the prediction error boundary, and ∈ indicates a membership relationship; The risk assessment model for self-healing operations is expressed as follows: Prob(e|s t ,a t )=F risk (s t ,a t ,ΔH RE ) Among them, Risk(s) t ,a t ) represents state s t Take action a t The risk value, E, is the set of potential risk events, Prob(e|s) t ,a t ) is in state s t Take action a t The conditional probability of risk event e occurring, Impact(e) is the degree of influence of event e, e is the index of the risk event variable, and F risk (s t ,a t ,ΔH RE Let ΔH be the risk function. RE Fluctuations in the output of new energy sources; Decision adjustments based on risk assessment are expressed as follows: Where Q(s) t Let a) be the value function of action a. The optimal action selected after risk adjustment under state s, A t Let be the set of actions at time step t, μ be the risk aversion coefficient, and Risk(s) be the risk aversion coefficient. t a) represents state s t The risk value of taking action 'a'.

7. The method for time-series decision-making of distribution network fault self-healing under new energy fluctuations as described in claim 6, characterized in that: The updated decision parameters include collecting system state data after the self-healing operation, correcting the state transition probability function and reward function weights, iteratively optimizing the characteristic function parameters, incorporating the energy storage system control vector into the self-healing operation sequence, restoring power supply in stages according to the load importance weights, and pre-verifying the feasibility of the operation through Monte Carlo simulation.

8. A system employing the method for time-series decision-making based on distribution network fault self-healing under renewable energy fluctuations as described in any one of claims 1 to 7, characterized in that: It includes a status assessment and prediction module, a self-healing decision optimization module, an adaptive adjustment module, and an execution and update module; The state assessment and prediction module is used to establish a new energy output fluctuation model, quantify the impact of random disturbances, predict load change trends and network state evolution, and generate a composite state vector containing voltage amplitude, phase angle and switching state. The self-healing decision optimization module is used to construct a decision framework that includes network topology, device status and load power supply, design a reward mechanism that integrates load recovery, operating costs and system security, and generate operation sequences through multi-stage optimization. The adaptive adjustment module is used to calculate the confidence interval of the operation sequence, respond to real-time fluctuations in new energy, construct the prediction error boundary, generate a robust operation strategy, quantify the operation risk index, and dynamically adjust the execution threshold. The execution and update module is used to execute the final self-healing operation sequence and monitor the effect, and verify the robustness of the scheme through Monte Carlo simulation, and update the core parameters in a closed loop.

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 method for self-healing timing decision-making of distribution network faults based on new energy fluctuations 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 method for self-healing timing decision-making of distribution network faults based on new energy fluctuations as described in any one of claims 1 to 7.

Images