Self-healing comprehensive evaluation method for intelligent power distribution network

Abstract

The invention discloses a self-healing comprehensive evaluation method for an intelligent power distribution network, which relates to the field of power grids and comprises the following steps: establishing a power distribution network self-healing evaluation comprehensive index system; operating data of the intelligent power distribution network are acquired, and preprocessing and feature extraction are performed on the acquired data by using a machine learning algorithm; calculating a self-healing evaluation index in a fault state; calculating a risk evaluation index in a normal operation state; constructing a dynamic weight adjustment model; according to the method, the multi-dimensional indexes of the fault state and the normal operation state are integrated, machine learning and dynamic weight adjustment are combined, and comprehensive, accurate and dynamic evaluation is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grids, and in particular to a comprehensive evaluation method for self-healing performance of a smart distribution network. Background Art

[0002] Self-healing evaluation indicators are used to measure the self-recovery ability of smart grids after a fault occurs. The main indicators include self-healing credibility, self-healing rate and self-healing speed, which measure the possibility of recovery, the success rate of recovery and the time required for recovery respectively. These indicators have some shortcomings: credibility is difficult to quantify all fault scenarios; self-healing rate requires a large amount of historical data and ignores the recovery differences of different types of faults; self-healing speed is affected by many factors and may ignore new faults that occur during the recovery process. In terms of the evaluation of the self-healing of distribution networks, existing studies often ignore the evaluation of the self-elimination of fault hazards of distribution networks under normal operating conditions, and also lack a comprehensive evaluation framework, lack of standardization, and difficulty in adapting to dynamic changes. Summary of the invention

[0003] In view of the problems existing in the existing power grid, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is to solve the problems that existing studies often ignore the evaluation of the self-elimination of fault hazards of the distribution network under normal operating conditions, lack of a comprehensive evaluation framework, lack of standardization, and difficulty in adapting to dynamic changes.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a method for comprehensively evaluating the self-healing performance of a smart distribution network.

[0007] It includes, establishing a comprehensive indicator system for evaluating self-healing of distribution networks;

[0008] Collect the operation data of the smart distribution network and use machine learning algorithms to pre-process and extract features from the collected data;

[0009] Calculate the self-healing evaluation index under fault conditions;

[0010] Calculate risk assessment indicators under normal operating conditions;

[0011] Build a dynamic weight adjustment model;

[0012] Calculate comprehensive evaluation indicators of self-healing and provide decision support and optimization based on the evaluation results.

[0013] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, wherein: the comprehensive index system for self-healing evaluation of the distribution network includes self-healing evaluation indicators under fault conditions: self-healing rapidity index, self-healing correctness index and self-healing effectiveness index;

[0014] Risk assessment indicators under normal operating conditions: including line overload risk, voltage over-limit risk, load loss risk, distributed energy access risk and demand response failure risk.

[0015] As a preferred solution of the comprehensive evaluation method for the self-healing performance of a smart distribution network described in the present invention, the operation data of the smart distribution network includes distribution network operation parameters: such as line current, node voltage, switch status; distributed energy output data; load demand data; network topology data; meteorological data; user behavior data; and socio-economic data.

[0016] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the machine learning algorithm includes:

[0017] Unsupervised learning algorithms for data cleaning and anomaly detection;

[0018] Principal component analysis (PCA) or autoencoders for feature extraction and dimensionality reduction;

[0019] Support vector machine (SVM) or random forest algorithm for predicting failure probability.

[0020] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the method for calculating the self-healing evaluation index under a fault state is:

[0021] The self-healing rapidity index is expressed as:

[0022] T st =T sc +T fl

[0023] Where T st The self-healing strategy formulation time refers to the time required from detecting a fault to determining appropriate self-healing measures, including the self-healing signal collection time and the self-healing fault location time. sc is the self-healing signal collection time, T fl Locate the time for self-healing faults;

[0024] The self-healing remote control time is expressed as:

[0025]

[0026] Where T cssis the single-step remote control time, N is the number of remote control switches, and the self-healing remote control time is T c ;

[0027] The full self-healing action time is expressed as:

[0028] T sh =T st +T c

[0029] Where T sh is the full self-healing action time, T st Set time for self-healing strategy, and the self-healing remote control time is T c ;

[0030] Self-healing correctness indicators:

[0031] The accuracy of the self-healing strategy is expressed as

[0032]

[0033] Among them C st represents the accuracy of the self-healing strategy, K c is the number of self-healing actions that conform to the rule, K a The total number of self-healing actions;

[0034] The success rate of self-healing remote control is expressed as

[0035]

[0036] Where Kr is the number of successful self-healing actions of remote control operation, C rc is the success rate of self-healing remote control, K a is the total number of self-healing actions K;

[0037] Self-healing effectiveness indicators:

[0038] The self-healing recovery efficiency is expressed as:

[0039]

[0040] Where K fd is the number of successful self-healing remote control, E fd The self-healing recovery efficiency, K a It is the total number of self-healing actions.

[0041] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the method for calculating the risk evaluation index under normal operating conditions is:

[0042] The line overload risk of the distribution network is defined as

[0043]

[0044] In the formula, Risk L is the line overload risk index, P r (E i ) is the line disconnection probability, sev(P li ) is the overload severity;

[0045] The voltage over-limit risk is expressed as:

[0046]

[0047] Where sev(Vi) is the voltage over-limit severity, P r (V i ) is the voltage over-limit rate;

[0048] The risk of lost load is defined as

[0049] R Li =P Li 2 I Li

[0050] Where P Li is the probability of losing load, I Li is the severity of the lost load;

[0051] Distributed energy access risk DER and demand response failure risk DR The calculation of is similar, taking into account probability and severity.

[0052] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the dynamic weight adjustment model is based on a reinforcement learning algorithm, and continuously optimizes the weight allocation strategy by interacting with the distribution network digital twin system. The specific steps include: initializing the weight coefficient;

[0053] Calculate the comprehensive evaluation index of self-healing based on the current weight;

[0054] Simulate the operation of the distribution network in the digital twin system to obtain reward signals;

[0055] Update the weight coefficients according to the reward signal;

[0056] Repeat the above steps until convergence or the preset number of iterations is reached.

[0057] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention,

[0058] Wherein: the method for calculating the comprehensive evaluation index of self-healing is:

[0059]

[0060] Among them, Tsh0 is the minimum self-healing full action time, w1 to w n is the dynamic weight coefficient of each indicator, I1 to I n are the standardized index values.

[0061] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the method for decision support and optimization based on the evaluation results includes: constructing a multi-dimensional visualization interface to display the comprehensive self-healing evaluation indicators and their composition in real time;

[0062] Use knowledge graphs and expert systems to generate improvement suggestions and optimization measures;

[0063] Input the evaluation results and improvement suggestions into the distribution network self-healing control system to achieve closed-loop optimization;

[0064] A time series model of evaluation indicators is established, and the long short-term memory network (LSTM) is used to predict the changing trend of self-healing performance in the future.

[0065] As a preferred solution of the comprehensive evaluation method for self-healing of a smart distribution network described in the present invention, the following steps are also included: constructing a distributed evaluation system based on blockchain technology to achieve data sharing and collaborative evaluation among multiple distribution network areas;

[0066] By using federated learning technology, model collaborative optimization among multiple distribution network operators can be achieved while protecting data privacy.

[0067] The beneficial effects of the present invention are as follows: the comprehensive evaluation method for the self-healing of the smart distribution network achieves a comprehensive, accurate and dynamic evaluation by integrating multi-dimensional indicators of fault status and normal operating status, combined with machine learning and dynamic weight adjustment. It not only provides real-time visualization and prediction capabilities, but also uses blockchain and federated learning technology to achieve multi-region collaborative evaluation. The method significantly enhances risk management, improves the degree of automation and decision-making support capabilities, and takes into account new factors such as distributed energy. This innovative method greatly improves the reliability, efficiency and adaptability of the distribution network, brings significant economic and social benefits, and provides strong technical support for the continuous optimization and sustainable development of smart grids. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0069] Figure 1 It is a flow chart of a comprehensive evaluation method for self-healing of smart distribution network;

[0070] Figure 2 It is the voltage distribution probability diagram. DETAILED DESCRIPTION

[0071] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0072] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0073] Secondly, the term "one 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 term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0074] Example 1

[0075] Reference Figure 1 This embodiment describes in detail the specific implementation process of a method for comprehensive evaluation of the self-healing performance of a smart distribution network. The method is applied to a smart distribution network system in a certain city, which includes 100 10kV distribution lines, serves about 500,000 users, and is connected to a large number of distributed photovoltaic power generation equipment and electric vehicle charging facilities.

[0076] S1: Establish a comprehensive indicator system for self-healing evaluation of distribution network First, establish a comprehensive indicator system that includes self-healing evaluation indicators under fault conditions and risk evaluation indicators under normal operating conditions:

[0077] Self-healing evaluation index under fault state: Self-healing rapidity index: T st ,T c ,T sh

[0078] Self-healing correctness index: C st ,C rc

[0079] Self-healing effectiveness index: E fd ,E xf ,E ur

[0080] Risk assessment indicators under normal operating conditions: Line overload risk: Risk L

[0081] Voltage over-limit risk: Risk V

[0082] Risk of lost load: R Li

[0083] Distributed Energy Access Risk: RiskDER

[0084] Risk of Demand Response Failure: RiskDR

[0085] S2: Collecting smart distribution network operation data

[0086] Deploy a network of smart sensors to collect the following data:

[0087] Distribution network operation parameters: line current, node voltage, switch status and other data are collected every 5 minutes.

[0088] Distributed energy output data: photovoltaic power output power is recorded every 15 minutes.

[0089] Load demand data: The load conditions of each distribution transformer are recorded every 15 minutes.

[0090] Network topology data: Network topology information is updated once a day.

[0091] Weather Data: Get local weather forecast information every hour.

[0092] User behavior data: Analyze smart meter data and generate user electricity usage behavior reports every day.

[0093] Socio-economic data: Regional economic development indicators are updated monthly.

[0094] S3: Data preprocessing and feature extraction using machine learning algorithms

[0095] Data cleaning: Use the isolation forest algorithm to detect and process abnormal data points.

[0096] Feature extraction: The principal component analysis (PCA) method is applied to extract the main features from the raw data.

[0097] Fault prediction: Use the random forest algorithm to train a fault prediction model based on historical data.

[0098] S4: Calculate the self-healing evaluation index under fault conditions

[0099] Taking a certain fault as an example, the calculation process is as follows:

[0100] Self-healing speed index: Tsc (Self-healing signal collection time) = 2 seconds

[0101] T fl (Self-healing fault location time) = 3 seconds

[0102] T st =T sc +T fl = 5 seconds

[0103] N (number of remote switches) = 3

[0104] T css (Single-step remote control time) = 1 second

[0105] T c =T css *N = 3 seconds

[0106] T sh =T st +T c =8 seconds

[0107] Self-healing correctness index: K a (Total number of self-healing actions) = 100

[0108] K c( The number of self-healing actions that meet the rules) = 98

[0109]

[0110] K r (Number of successful self-healing actions by remote control) = 97

[0111]

[0112] Self-healing effectiveness index: K fd (Number of feeders successfully restored by self-healing remote control) = 3

[0113] K a (Total number of feeders affected) = 3

[0114]

[0115] N xf (Number of distribution transformers successfully restored by self-healing remote control) = 15

[0116] N a (Number of distribution transformers affected by the fault) = 16

[0117]

[0118] N ur (Number of users who successfully restored power through self-healing remote control) = 9800

[0119] N au (Number of users affected by the failure) = 10,000

[0120]

[0121] S5: Calculate risk assessment indicators under normal operating conditions

[0122] Taking a certain line as an example, the calculation process is as follows:

[0123] Line overload risk RiskL: Assume that the line is close to overload three times within 24 hours

[0124] Calculate Pr(Ei) and sev(Pli) for each state

[0125] For example, for the first approach to overload:

[0126] Voltage over-limit risk V : Assume that within 24 hours, the line voltage is below 0.93 per unit for 5% of the time

[0127] P r (V i )=0.05

[0128] sev(V i )=(0.9-0.93)^2 / 0.9≈0.001

[0129]

[0130] Risk of load loss RLi: Assume that 100 failure scenarios are simulated, 10 of which result in load loss

[0131] P Li =10 / 100=0.1

[0132] The average load loss caused by each fault is 2MW, and the distribution network capacity is 100MW.

[0133] I Li =(1 / 100)*Σ(2 / 100)=0.0002

[0134] R Li =P Li *I Li =0.00002

[0135] S6: Building a dynamic weight adjustment model

[0136] Use deep reinforcement learning algorithms such as DDPG to build dynamic weight adjustment models:

[0137] Define the state space: including the current indicator values ​​and system operation status.

[0138] Define the action space: the weight adjustment of each indicator.

[0139] Define the reward function: based on the improvement degree of the adjusted comprehensive evaluation index of self-healing.

[0140] Training process: Initialization weight: w1 = w2 = ... = wn = 1 / n

[0141] Simulating distribution network operation in a digital twin system

[0142] Select actions based on current state (adjust weights)

[0143] Calculate new comprehensive self-healing evaluation indicators and rewards

[0144] Update neural network parameters

[0145] Repeat the above steps until convergence

[0146] S7: Calculate the comprehensive evaluation index of self-healing

[0147] Assume that the weights after dynamic adjustment are: w1 = 0.15, w2 = 0.15, w3 = 0.1, w4 = 0.1, w5 = 0.1, w6 = 0.15, w7 = 0.15, w8 = 0.1

[0148]

[0149] S8: Decision support and optimization based on evaluation results

[0150] Build a multi-dimensional visualization interface: Use the D3.js library to create an interactive dashboard that displays Ecei values ​​and various sub-indicators in real time.

[0151] Design a red, yellow and green three-color early warning system, when Ecei < 60 is red, 60 ≤ Ecei < 80 is yellow, and Ecei ≥ 80 is green.

[0152] Generate improvement suggestions: Analyze the reasons for low Ecei based on knowledge graph and expert system.

[0153] In this case, the system will recommend optimizing the self-healing control algorithm to shorten T sh , while paying attention to line load conditions and considering increasing line capacity or load shifting.

[0154] Input to the self-healing control system: The evaluation results and improvement suggestions are transmitted to the self-healing control system through a standardized interface.

[0155] The self-healing control system automatically adjusts control parameters based on the recommendations, such as thresholds to shorten fault location time.

[0156] Predict future trends: Use the LSTM network to predict the change trend in the next 7 days based on the Ecei data of the past 30 days.

[0157] If the forecast shows that Ecei has a downward trend, the system will issue an early warning and recommend preventive maintenance.

[0158] Example 2

[0159] Self-healing strategy development time

[0160] The self-healing strategy formulation time refers to the time required from detecting a fault to determining the appropriate self-healing measures. This time includes the self-healing signal collection time and the self-healing fault location time. It can be expressed as follows:

[0161] T st =T sc +T fl

[0162] Where T st Represents the self-healing strategy formulation time, T sc Represents the self-healing signal collection time, T fl Indicates the self-healing fault location time;

[0163] Self-healing remote control time

[0164] Self-healing remote control time T c It refers to the time required to send control commands to the remote terminal (such as RTU or DTU, etc.) on site and complete all related operations after the self-healing strategy is formulated. Depending on whether self-healing is automatically executed, it can be divided into two modes:

[0165] Self-healing remote control time in semi-automatic mode: In this mode, although the self-healing strategy is automatically generated, the actual operation still requires manual confirmation.

[0166] Self-healing remote control time in fully automatic mode: In this mode, the self-healing strategy and its execution are fully automated without human intervention.

[0167] For each self-healing action, different lines and fault areas may require different self-healing strategies, so the number of switches that need to be remotely controlled will also be different. In order to facilitate the comparison of the self-healing control time of different numbers of switches, a "single-step remote control time T css ”, which is the average time required to remotely control a switch.

[0168] Where N is the number of remote control switches during the self-healing action.

[0169] Self-healing full action time

[0170] The self-healing full action time refers to the total time required from the detection of a fault until all self-healing measures are fully executed. It is as follows:

[0171] T sh =T st +T c

[0172] Self-healing correctness index

[0173] Self-healing strategy accuracy

[0174] C st represents the accuracy of the self-healing strategy, K c represents the number of self-healing actions that conform to the rules, K a Indicates the total number of self-healing actions;

[0175] Self-healing remote control success rate

[0176] C rc represents the success rate of self-healing remote control, K r Indicates the number of successful self-healing actions of remote control operation, K a Represents the total number of self-healing actions K;

[0177] Self-healing effectiveness index

[0178] Self-healing recovery efficiency

[0179] E fd Indicates the self-healing recovery efficiency, K fd Indicates the number of successful self-healing remote control (within a certain period of time), K a Indicates the total number of self-healing actions;

[0180] Self-healing restoration efficiency at the distribution transformer (medium voltage user) level

[0181]

[0182] E xf Represents the self-healing recovery efficiency at the distribution transformer level, ∑N xf Indicates the number of distribution transformers that have been successfully restored to power by self-healing remote control (in a certain period of time), ∑N a Indicates the number of distribution transformers affected by the fault (in a certain period);

[0183] Self-healing power restoration efficiency at the user (low-voltage user) level

[0184]

[0185] E ur Represents the self-healing power restoration efficiency at the user level, ∑N ur Indicates the number of users who have successfully restored power through self-healing remote control, ∑N ua Indicates the number of users affected by the failure;

[0186] Risk assessment indicators under normal operating conditions

[0187] Line overload risk

[0188] Line overload risks are common in distribution networks, and the main reasons include increased load, aging equipment, unreasonable design, faults or abnormal conditions, lack of effective monitoring and management, unreasonable scheduling arrangements, load transfer after unplanned power outages, and insufficient distribution automation. These factors may cause the power grid to be unable to effectively carry the load under certain conditions, thus causing overload problems.

[0189] The probability of power line disconnection within a period of time is

[0190]

[0191] Where P r (E i ) represents the probability of line disconnection in time period t, E i represents the i-th accident, λ represents E i The number of occurrences in time period t

[0192] The severity of line overload is as follows:

[0193]

[0194] In the formula, sev(P li ) indicates that due to accident E i The resulting overload severity of line l, P li Indicates accident E i Active power of line l when it occurs, P lmax Indicates the maximum transmission power set for line 1.

[0195] The line overload risk of the entire distribution network is defined as

[0196]

[0197] In the formula, Risk L It is the line overload risk indicator.

[0198] Voltage over-limit risk

[0199] Voltage over-limit risk refers to the possibility that the node voltage in the distribution network exceeds the allowable operating range, which may be caused by load fluctuations, insufficient reactive power compensation, line losses, equipment failures, changes in network structure, extreme weather conditions, and access to distributed energy sources. Voltage over-limit not only affects the quality of electricity consumption for users, but may also damage electrical equipment and pose a threat to the safe and stable operation of the power grid.

[0200] Assume that the voltage distribution probability of node i in time period t is as follows: Figure 2 As shown, Figure 2 is the voltage distribution probability. Using the voltage distribution probability, the voltage over-limit rate of node i in time period t can be calculated as:

[0201]

[0202] Where P r (V i ) represents the voltage crossing rate, f(V t,i ) represents the voltage probability density function of the ith node in period t, 1.1 and 0.9 are the upper and lower limits of the voltage per unit value fluctuation allowed, respectively; V down,t,i,max and V up,t,i,max They are the maximum lower limit voltage and the maximum upper limit voltage of the i-th node in period t, which are used to characterize the severity of voltage crossing.

[0203] The severity of voltage over-limit is expressed by the following formula:

[0204]

[0205] Where, sev(V i ) is the voltage over-limit severity of node i.

[0206] The voltage over-limit risk of the entire distribution network is defined as

[0207]

[0208] In the formula, Risk V represents the voltage over-limit risk; Nb represents the total number of weak nodes in the power grid;

[0209] Risk of lost load

[0210] In the future distribution network, due to the large-scale access of distributed generation and electric vehicles, the number of components in the distribution network is increasing and the operation mode is becoming more and more complex. The risk of lost load refers to the possibility that power cannot be supplied to users due to faults or other abnormal conditions in the distribution network.

[0211] The probability of loss of load in the distribution network is

[0212] Where i represents the number of the expected fault, Nf represents the total number of expected faults, L(i) represents the load loss caused by expected fault i, which is 1, otherwise it is 0, P r (L i ) represents the probability of losing load;

[0213] The severity of the load loss can be measured by

[0214]

[0215] In the formula, P(L i ) represents the load loss caused by fault i, P S represents the capacity of the distribution network, I Li Indicates the severity of the lost load;

[0216] The loss load risk of the distribution network is defined as

[0217] R Li =P Li 2 I Li

[0218] In the formula, R Li Indicates the risk of lost load;

[0219] Comprehensive evaluation of self-healing

[0220] In order to intuitively compare the comprehensive self-healing effects in different regions and periods, various raw data are converted into a directly comparable standardized format through a certain scaling system, combining the distribution network risk assessment indicators under normal operating conditions and the self-healing evaluation indicators under fault conditions mentioned above.

[0221] The formulas for the self-healing characteristic index values ​​are as follows:

[0222]

[0223] Self-healing evaluation weight:

[0224] F=[f1,f2,f3,f4,f5,f6,f7,f8,f9]

[0225] The comprehensive evaluation index E of self-healing using a percentage scoring scale cei

[0226]

[0227] Where, T sh0 Indicates the minimum self-healing full action time.

[0228] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A comprehensive evaluation method for self-healing of a smart distribution network, characterized by: Establish a comprehensive indicator system for self-healing evaluation of distribution networks; Collect the operation data of the smart distribution network and use machine learning algorithms to pre-process and extract features from the collected data; Calculate the self-healing evaluation index under fault conditions; Calculate risk assessment indicators under normal operating conditions; Build a dynamic weight adjustment model; Calculate comprehensive evaluation indicators of self-healing and provide decision support and optimization based on the evaluation results.

2. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 1, characterized in that: The comprehensive index system for self-healing evaluation of the distribution network includes self-healing evaluation indicators under fault conditions: self-healing rapidity index, self-healing correctness index and self-healing effectiveness index; Risk assessment indicators under normal operating conditions: including line overload risk, voltage over-limit risk, load loss risk, distributed energy access risk and demand response failure risk.

3. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 2, characterized in that: The smart distribution network operation data includes distribution network operation parameters: such as line current, node voltage, switch status; distributed energy output data; Load demand data; Network topology data; Meteorological data; user behavior data; socio-economic data.

4. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 3, characterized in that: The machine learning algorithm includes: Unsupervised learning algorithms for data cleaning and anomaly detection; Principal component analysis (PCA) or autoencoders for feature extraction and dimensionality reduction; Support vector machine (SVM) or random forest algorithm for predicting failure probability.

5. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 4, characterized in that: The method for calculating the self-healing evaluation index under the fault state is: The self-healing rapidity index is expressed as: T st =T sc +T fl Where T st The self-healing strategy formulation time refers to the time required from detecting a fault to determining appropriate self-healing measures, including the self-healing signal collection time and the self-healing fault location time. sc is the self-healing signal collection time, T fl Locate the time for self-healing faults; The self-healing remote control time is expressed as: Where T css is the single-step remote control time, N is the number of remote control switches, and the self-healing remote control time is T c ; The full self-healing action time is expressed as: T sh =T st +T c Where T sh is the full self-healing action time, T st Set time for self-healing strategy, and the self-healing remote control time is T c ; Self-healing correctness indicators: The accuracy of the self-healing strategy is expressed as Among them C st represents the accuracy of the self-healing strategy, K c is the number of self-healing actions that conform to the rule, K a The total number of self-healing actions; The success rate of self-healing remote control is expressed as Where Kr is the number of successful self-healing actions of remote control operation, C rc is the success rate of self-healing remote control, K a is the total number of self-healing actions K; Self-healing effectiveness indicators: The self-healing recovery efficiency is expressed as: Where K fd is the number of successful self-healing remote control, E fd The self-healing recovery efficiency, K a It is the total number of self-healing actions.

6. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 5, characterized in that: The method for calculating the risk assessment index under normal operating conditions is: The line overload risk of the distribution network is defined as In the formula, Risk L is the line overload risk index, P r (E i ) is the line disconnection probability, sev(P li ) is the overload severity; The voltage over-limit risk is expressed as: Where sev(Vi) is the voltage over-limit severity, P r (V i ) is the voltage over-limit rate; The risk of lost load is defined as R Li =P Li 2 I Li Where P Li is the probability of losing load, I Li is the severity of the lost load; Distributed energy access risk DER and demand response failure risk DR The calculation of is similar, taking into account probability and severity.

7. The method for comprehensive evaluation of self-healing performance of a smart distribution network according to claim 6, characterized in that: The dynamic weight adjustment model is based on a reinforcement learning algorithm and continuously optimizes the weight allocation strategy by interacting with the distribution network digital twin system. The specific steps include: initializing the weight coefficient; Calculate the comprehensive evaluation index of self-healing based on the current weight; Simulate the operation of the distribution network in the digital twin system to obtain reward signals; Update the weight coefficients according to the reward signal; Repeat the above steps until convergence or the preset number of iterations is reached.

8. A comprehensive evaluation method for self-healing of a smart distribution network, characterized in that: The method for calculating the comprehensive evaluation index of self-healing is: Among them, Tsh0 is the minimum self-healing full action time, w1 to w n is the dynamic weight coefficient of each indicator, I1 to I n are the standardized index values.

9. A comprehensive evaluation method for self-healing of a smart distribution network, characterized in that: The method for decision support and optimization based on the evaluation results includes: constructing a multi-dimensional visualization interface to display the comprehensive self-healing evaluation indicators and their composition in real time; Use knowledge graphs and expert systems to generate improvement suggestions and optimization measures; Input the evaluation results and improvement suggestions into the distribution network self-healing control system to achieve closed-loop optimization; A time series model of evaluation indicators is established, and the long short-term memory network (LSTM) is used to predict the changing trend of self-healing performance in the future.

10. A comprehensive evaluation method for self-healing of a smart distribution network, characterized in that: The following steps are also included: Build a distributed evaluation system based on blockchain technology to achieve data sharing and collaborative evaluation among multiple distribution network areas; By using federated learning technology, model collaborative optimization among multiple distribution network operators can be achieved while protecting data privacy.

Images