Power distribution network fault positioning method and system

Abstract

The method comprises the following steps: acquiring historical parameter data of a power distribution network fault, establishing a first power distribution network fault positioning model according to the historical parameter data of the power distribution network fault, and updating the first power distribution network fault positioning model according to a preset error correction strategy to obtain a second power distribution network fault positioning model; and converting the second power distribution network fault positioning model into a Markov decision process, solving the Markov decision process according to an improved conservative Q learning algorithm, and performing power distribution network fault positioning. The matrix algorithm and the optimization algorithm need accurate topological structure information and fault current data of the power distribution network, and the conservative Q learning algorithm does not need to collect real topological data of the power distribution network. After the intelligent agent learns the strategy, the fault positioning action can be directly carried out through a strategy function in the changing power distribution network environment, and repeated iterative calculation is not needed.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network fault location, and in particular to a distribution network fault location method and system. Background Art

[0002] The widespread use of distributed generation technology has significantly changed the characteristics of distribution networks, making fault location more complicated. However, fast and accurate fault location is the key to ensuring rapid restoration of power supply. Therefore, it is particularly urgent to study and develop fault location methods suitable for large-scale distributed generation systems.

[0003] In most cases, phase-to-phase short-circuit faults have obvious characteristics, and their related information can be accurately collected and analyzed by the feeder terminal unit (FTU), thereby achieving fault location. At present, the location of phase-to-phase short-circuit faults mainly relies on positioning algorithms, which are not only suitable for densely populated areas, but also have become the core method in this research field.

[0004] The core strategy of fault location is to use the fault current information uploaded by the feeder terminal unit (FTU) to the supervisory control and data acquisition (SCADA) system, and use the location algorithm for precise location. These algorithms are mainly divided into two categories: matrix algorithms and optimization algorithms. The matrix algorithm constructs a fault discrimination matrix by combining the topology of the distribution network and the fault current data, and determines the specific section where the fault occurs through matrix operations. The optimization algorithm uses the method of simulating the fault of the distribution network components to convert the fault location problem into an optimization problem by finding a fault state that highly matches the actual fault current information. Given the complexity of the problem, optimization algorithms, such as genetic algorithms and particle swarm optimization algorithms, tend to be used to achieve fast and accurate solutions. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a distribution network fault location method and system, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a distribution network fault location method, comprising:

[0010] Acquire distribution network fault historical parameter data, and establish a first distribution network fault location model according to the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device;

[0011] Update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model;

[0012] The second distribution network fault location model is transformed into a Markov decision process, and the Markov decision process is solved according to an improved conservative Q-learning algorithm to perform distribution network fault location.

[0013] As a preferred solution of the distribution network fault location method described in the present invention, wherein: the acquisition of distribution network fault historical parameter data, and the establishment of a first distribution network fault location model based on the fault historical parameter data, the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device, including:

[0014] According to the distribution network fault historical parameter data, establishing the first distribution network fault location model includes locating the fault according to the historical monitoring data provided by the feeder terminal device, which is specifically expressed as follows:

[0015] FTUs=(s1,s2,...,s k ...,s n )

[0016] Among them, s k represents the fault current direction information detected by the kth feeder terminal device. If the current direction is positive, then s k =1, if the current direction is negative, then s k = -1, if no fault current is detected, then s k =0;

[0017] The fault current direction is that when there is distributed energy in the feeder section, the direction from the system power supply to the distributed energy is regarded as a positive direction, otherwise, the direction is regarded as a negative direction.

[0018] As a preferred solution of the distribution network fault location method described in the present invention, the preset error correction strategy includes:

[0019] Input initial current state S i (L), distribution network structure, fault point p;

[0020] Number all distributed energy resources and feeder terminal devices;

[0021] Based on the Dijkstra algorithm, for any i and j, find the path between the i-th distributed energy source and the j-th feeder terminal device;

[0022] Determine whether there is a path between the i-th distributed energy source and the j-th fault point p;

[0023] If it exists, correct the line status of the feeder where the distributed energy is located;

[0024] Output the correct current state S(L).

[0025] As a preferred solution of the distribution network fault location method described in the present invention, the preset error correction strategy also includes:

[0026] The corrected objective function is as follows:

[0027]

[0028] Among them, σ represents the penalty coefficient set to 0.5-1.0 to avoid misjudgment, N FTU Indicates the number of feeder terminal devices; N L Indicates the number of lines, Indicates the state of the line. When the kth line fails, L k =1, otherwise, L k =0; K j represents the switch corresponding to the jth distributed energy; Indicates the fault current state corresponding to the fault line state L, FTU i Represents the fault current state detected by the i-th feeder terminal device.

[0029] As a preferred solution of the distribution network fault location method described in the present invention, wherein: the conversion of the second distribution network fault location model into a Markov decision process also includes: converting the distribution network fault location model into a Markov decision process by defining an intelligent agent, an environment, a state, an action and a reward, wherein the intelligent agent is a distribution network monitor or a fault location program, and the environment is the distribution network itself.

[0030] As a preferred solution of the distribution network fault location method described in the present invention, wherein: the method of solving the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault includes: initializing the Q network, updating the entropy regularization coefficient, function Q and strategy in the Q network, completing the conservative Q learning algorithm design, and solving the Markov decision process according to the designed conservative Q learning algorithm.

[0031] As a preferred solution of the distribution network fault location method described in the present invention, wherein: solving the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault also includes: the conservative Q learning algorithm includes designing a conservative Q learning model, the input of the model is the historical data of distribution network fault location, and the output is the distribution network fault location strategy function.

[0032] In a second aspect, the present invention provides a distribution network fault location system, comprising:

[0033] A first model building module is used to obtain distribution network fault history parameter data, and establish a first distribution network fault location model according to the fault history parameter data, wherein the fault history parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device;

[0034] A second model building module is used to update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model;

[0035] The positioning module is used to convert the second distribution network fault positioning model into a Markov decision process, and solve the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault.

[0036] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a distribution network fault location method and system, obtains distribution network fault historical parameter data, and establishes a first distribution network fault location model based on the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power-related information when the fault occurs and historical monitoring data provided by the feeder terminal device; updates the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model; converts the second distribution network fault location model into a Markov decision process, and solves the Markov decision process according to an improved conservative Q learning algorithm to locate the distribution network fault. Matrix algorithms and optimization algorithms require accurate topological structure information and fault current data of the distribution network, while the conservative Q learning algorithm does not need to collect real distribution network topological data. After the intelligent agent learns the strategy, it can directly perform fault location actions through the strategy function in the face of a changing distribution network environment without repeated iterative calculations. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing 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 labor. Among them:

[0040] Figure 1 A method flow chart of a distribution network fault location method and system provided by an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of misjudgment of a common fault location model of a distribution network fault location method and system provided by an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of the interaction between an intelligent agent and an environment in a reinforcement learning of a distribution network fault location method and system provided by an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of a Markov decision process of a distribution network fault location method and system provided by an embodiment of the present invention;

[0044] Figure 5 A schematic diagram showing the difference between online reinforcement learning (left) and offline reinforcement learning (right) of a distribution network fault location method and system provided by an embodiment of the present invention;

[0045] Figure 6 A schematic diagram of an intelligent agent training process of a distribution network fault location method and system provided by an embodiment of the present invention;

[0046] Figure 7 A schematic diagram of an IEEE-33 node distribution network of a distribution network fault location method and system provided by an embodiment of the present invention;

[0047] Figure 8 An internal structural diagram of a computer device of a distribution network fault locating method and system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, 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 drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0049] Example 1

[0050] Reference Figure 1-Figure 8 , which is the first embodiment of the present invention, and provides a distribution network fault location method and system, including:

[0051] In the related art, (1) the shortcomings of matrix algorithms are:

[0052] High data requirements: Matrix algorithms rely on precise topological information of the distribution network and require accurate fault current data. If the data is not collected accurately or is noisy, the accuracy of fault location may be affected.

[0053] The calculation is more complicated: In a large-scale distribution network, the size of the matrix may be very large, resulting in a complex calculation process and requiring a long processing time.

[0054] Poor adaptability to dynamic changes: In the case of dynamic changes in the distribution network, such as load fluctuations or the randomness of distributed generation, the matrix algorithm may need to be updated frequently to adapt to these changes.

[0055] (2) Shortcomings of the optimization algorithm:

[0056] High consumption of computing resources: Optimization algorithms, especially heuristic algorithms such as genetic algorithms and particle swarm optimization algorithms, may require a lot of computing resources, especially in large-scale networks.

[0057] Model assumptions: Optimization algorithms are usually based on some simplified model assumptions, which may not fully conform to the actual distribution network conditions, thus affecting the accuracy of fault location.

[0058] Real-time issues: Although heuristic optimization algorithms can usually find good solutions, they may converge slowly, especially when the parameters are poorly set. In fault location scenarios that require fast response, optimization algorithms may not be able to guarantee real-time performance, especially when the network scale is large or the fault types are complex.

[0059] The present application provides a method for effectively solving the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the distribution network fault location method;

[0060] Figure 1 A method flow chart of a distribution network fault location method and system is shown, including:

[0061] S101, obtaining distribution network fault historical parameter data, and establishing a first distribution network fault location model according to the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by a feeder terminal device;

[0062] It should be noted that the fault location problem of the distribution network is the premise for ensuring the stability and restoration of power supply. Distributed energy is connected to the distribution network in large quantities, and the distribution network becomes a multi-power system. The power generation characteristics of distributed energy are random, intermittent, susceptible to environmental influences, and small reserves. They have a direct impact on the current, loss, voltage distribution and short-circuit level of the entire line, and may cause power flow reversal problems. The characteristics of short-circuit faults in distribution networks are obvious, and feeder terminal units (FTUs) can be used to collect fault information. Fault location information collected based on FTUs is considered to be the mainstream method in this field.

[0063] In an embodiment of the present application, distribution network fault history parameter data is obtained, and a first distribution network fault location model is established according to the fault history parameter data. The fault history parameter data includes at least current, voltage, and power related information when the fault occurs, and the historical monitoring data provided by the feeder terminal device includes:

[0064] According to the distribution network fault historical parameter data, establishing the first distribution network fault location model includes locating the fault according to the historical monitoring data provided by the feeder terminal device, which is specifically expressed as follows:

[0065] FTUs=(s1,s2,...,s k ...,s n )

[0066] Among them, s k represents the fault current direction information detected by the kth feeder terminal device. If the current direction is positive, then s k =1, if the current direction is negative, then sk = -1, if no fault current is detected, then s k =0;

[0067] It should be noted that the direction of the fault current is that when there is distributed energy in the feeder section, the direction from the system power supply to the distributed energy is considered to be the positive direction, otherwise, the direction is considered to be the negative direction.

[0068] It should be noted that the commonly used fault location model is as follows:

[0069]

[0070] Among them, N FTU Indicates the number of FTUs; N L Indicates the number of lines, Indicates the state of the line. If the kth line fails, L k =1, otherwise, L k =0; K j represents the switch corresponding to the jth distributed energy source; SET down Represents the downstream line set of line j; SET up represents the upstream line set of line j; Indicates the fault current state corresponding to the fault line state L.

[0071] It should be noted that the main idea of ​​the commonly used fault location model is to find a fault state whose corresponding fault current state is most similar to the fault state collected by the FTU. However, due to the access of distributed energy, the commonly used model may face the challenge of misjudgment, such as Figure 2 As shown in Figure 2. According to the model constraints, no fault current flows, which is obviously inconsistent with the actual situation.

[0072] S102, updating the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model;

[0073] Among them, the preset error correction strategies include:

[0074] Input initial current state S i (L), distribution network structure, fault point p;

[0075] Number all distributed energy resources and feeder terminal devices;

[0076] Based on the Dijkstra algorithm, for any i and j, find the path between the i-th distributed energy source and the j-th feeder terminal device;

[0077] Determine whether there is a path between the i-th distributed energy source and the j-th fault point p;

[0078] If it exists, correct the line status of the feeder where the distributed energy is located;

[0079] Output the correct current state S(L).

[0080] Furthermore, the preset error correction strategy also includes:

[0081] The corrected objective function is as follows:

[0082]

[0083] Among them, σ represents the penalty coefficient set to 0.5-1.0 to avoid misjudgment, N FTU Indicates the number of feeder terminal devices; N L Indicates the number of lines, Indicates the state of the line. When the kth line fails, L k =1, otherwise, L k =0; K j represents the switch corresponding to the jth distributed energy; Indicates the fault current state corresponding to the fault line state L, FTU i Represents the fault current state detected by the i-th feeder terminal device.

[0084] S103, converting the second distribution network fault location model into a Markov decision process, and solving the Markov decision process according to an improved conservative Q-learning algorithm to locate the distribution network fault.

[0085] Converting the second distribution network fault location model into a Markov decision process also includes: converting the distribution network fault location model into a Markov decision process by defining an agent, an environment, a state, an action and a reward, wherein the agent is a distribution network monitor or a fault location program, and the environment is the distribution network itself.

[0086] The Markov decision process is solved according to the improved conservative Q-learning algorithm, and the distribution network fault location is performed, including: initializing the Q network, updating the entropy regularization coefficient, function Q and strategy in the Q network, completing the conservative Q-learning algorithm design, and solving the Markov decision process according to the designed conservative Q-learning algorithm.

[0087] Solving the Markov decision process according to the improved conservative Q-learning algorithm, distribution network fault location also includes: the conservative Q-learning algorithm includes designing a conservative Q-learning model, the input of the model is the historical data of distribution network fault location, and the output is the distribution network fault location strategy function.

[0088] It should be noted that the subject of reinforcement learning is an agent, which is responsible for taking actions and making decisions. The object that interacts with the agent is called the environment, which can be abstractly understood as the rules or mechanisms in the interaction process.

[0089] The mathematical foundation and modeling tool of reinforcement learning is Markov decision process (MDP). An MDP consists of a state space Action Space It consists of state transfer function P(s'|s,a), reward function r(s,a), discount factor γ∈[0,1), etc.

[0090] State refers to the summary of the current environment. State space refers to the set of all possible states, written as cursive letters. The state space can be discrete or continuous. The state space can be a finite set or an infinite countable set.

[0091] An action is a decision made by an agent based on its current state. Actions can be chosen deterministically or randomly. The action space is the set of all possible actions, written as cursive letters.

[0092] Reward refers to a value returned by the environment to the agent after the agent performs an action. Rewards are often defined by oneself, and the quality of reward definition greatly affects the results of reinforcement learning. Reward is a function of the current state s and the current action a, denoted as r(s,a).

[0093] The return is the sum of all rewards from the current moment to the end of this round, so the return is also called cumulative future reward. Let the return at time t be the random variable U t . Define the return as:

[0094] U t =R t +R t+1 +R t+2 +R t+3 +…+R n

[0095] Among them, R t is the reward value at time t, R t+1 is the reward value at time t+1, and so on.

[0096] The goal of reinforcement learning is to find a strategy that maximizes the expected return. This strategy is called the optimal policy.

[0097] In MDP, a discounted return is often used to discount future rewards. The definition of discounted return is:

[0098] U t =R t +γ·R t+1 +γ 2 ·R t+2 +γ 3 ·R t+3 +…

[0099] Among them, γ∈[0,1) refers to the discount rate, which discounts future rewards.

[0100] State transitions may be random, and reinforcement learning usually assumes that state transitions are random, and the randomness comes from the environment. Use the state transition probability function to describe state transitions. The value of this formula represents the probability that when the agent performs action a in the current state s, the state of the environment becomes s′.

[0101] Usually, the state transition has the Markov property, which can be expressed as This formula indicates that the state S at the next moment t+1 Depends only on the current state S t and action A t , and does not depend on past states and actions.

[0102] In the embodiment of the present application, it is assumed that the distribution network has n nodes, N L Branch road, N FTU FTUs.

[0103] When a branch fails, the fault location information vector is:

[0104] FTUs=(s1,s2,...,s k ...,s n )

[0105] Among them, s k Indicates the fault current direction information detected by the kth FTU. If the current direction is positive, then s k =1, if the current direction is negative, then s k = -1, if no fault current is detected, then s k =0.

[0106] The status of the line is:

[0107]

[0108] By defining the agent, environment, state, action and reward, the distribution network fault location problem can be transformed into a Markov decision process. The agent is the distribution network monitor or fault location program. The environment is the distribution network itself.

[0109] The state S at time t t It is expressed as:

[0110]

[0111] Wherein, FTUs is the fault location information collected by FTUs, L is the state of the line, and S(L) is the fault current state corresponding to the fault line state L.

[0112] The action is to determine which branch has a fault. Action A at time t t It is expressed as:

[0113] A t =k,k∈[1,2,...,N L ]

[0114] The reward function is transformed by the judgment index, and the reward R at time t t It is expressed as:

[0115]

[0116] It should be noted that Q-learning is the most classic and traditional reinforcement learning algorithm. Its idea can be simply understood as obtaining a "prophet" who can predict the future through training. This prophet must be able to foresee the possible results of taking all actions and choose actions based on the foreseen results. In reinforcement learning, the prophet is the action-value function, also known as the Q function, as shown in the following formula.

[0117]

[0118] Among them, the action value function Q π (s t ,a t ) is the value of the current state S t Next, we select action A based on strategy π t Return U t expected value.

[0119] The action-value function is generally represented by a neural network. At the beginning, the parameters of the network are random, which means that the prophet knows nothing. However, as the intelligent agent interacts with the environment and accumulates enough experience, it continuously optimizes the action-value function and eventually obtains the optimal action-value function.

[0120]

[0121] Among them, the optimal action value function The value of is the action value function Qπ(s t ,a t )’s maximum value.

[0122] In this process, the "wisdom" of the intelligent agent is not ideal at the beginning, but through training, it can eventually reach the level of a "prophet".

[0123] Ultimately, the agent selects an action based on the optimal action-value function, and its strategy is to select an action with the largest output result in the action-value function.

[0124] It should be noted that traditional reinforcement learning algorithms, including Q learning, can be collectively referred to as online reinforcement learning. During the training process, the agent can continuously interact with the environment to obtain new feedback data. During the training process, the data is stored in the experience replay pool and sampled when needed.

[0125] However, in the distribution network operation scenario, allowing intelligent agents that have not yet learned to interact directly with the distribution network is also likely to cause danger or huge losses.

[0126] Offline reinforcement learning can solve this problem. The goal of offline reinforcement learning is to obtain a better strategy through reinforcement learning algorithm only from the collected and determined data set without the agent interacting with the environment, such as Figure 5 As shown in Figure 2. Offline reinforcement learning is also called batch-constrained reinforcement learning. Batch constraint means “historical data constraint”.

[0127] However, there is a common problem of extrapolation error in offline reinforcement learning. Extrapolation error refers to the error caused by the mismatch between the distribution of state-action combinations that the current strategy may access and the state-action combinations sampled from the dataset. The learning agent will overestimate the action value function, resulting in extrapolation error.

[0128] Why is the online reinforcement learning algorithm not affected by the extrapolation error? Because for online reinforcement learning, the agent has the opportunity to sample new data in time by interacting with the environment, thereby correcting these errors. However, in offline reinforcement learning, the agent cannot interact with the environment and cannot make corrections.

[0129] Therefore, in general, offline reinforcement learning algorithms should try to limit the size of the extrapolation error as much as possible to obtain a better strategy.

[0130] In the embodiments of the present application, experiments have shown that extrapolation errors mainly lead to overestimation of functions at points far from the data set, and even often cause values ​​to diverge upward. Therefore, if some method can be used to keep the function at points that deviate from the data set in the algorithm at a very low value, it may be possible to eliminate some of the effects of extrapolation errors. This is the basic idea of ​​the conservative Q-learning algorithm. CQL (conservative Q-learning algorithm) introduces some additional restrictions on the ordinary Bellman equation to achieve this goal.

[0131] In the traditional Q-learning algorithm, the Q function is updated by the following formula.

[0132]

[0133] in, is the Bellman operator of the actual calculation strategy. It is a collection of state-action combinations. and are the Q functions before and after iteration respectively.

[0134] In the embodiment of the present application, in order to prevent the overestimation of Q value in each state (especially in the state not in the data set), it is necessary to penalize the high value in some states. Considering the general case, it is hoped that the expected value of Q in a certain distribution π(s,a) is minimized. In the above formula, The calculation of s, a, s′, and a′ is required, but only a′ is generated and may not be in the dataset. Therefore, the action obtained by the state s in the dataset according to the strategy π is penalized:

[0135]

[0136] Where β is the balancing factor. It can be proved that the function Q given by the iterative convergence of the above formula has a value smaller than the true value at any (s, a). However, if the conditions are relaxed and only the expected value on is smaller than the true value, the constraints on the above formula can be slightly relaxed. A natural idea is that for data points that conform to the behavioral strategy used to generate the data set, the valuation of these points can be considered to be more accurate, and there is no need to restrict the value to be small at these points. As compensation for the first term, the above formula is changed to:

[0137]

[0138] Among them, π b The behavior strategy is approximated by the existing data in the dataset. It can be proved that when π b =π, the function Q obtained by iterative convergence of the above formula is not smaller than the true value at every point, but its expectation is smaller than the true value, that is,

[0139]

[0140] In the embodiment of the present application, according to the derivation process of Q learning, the iterative equation can be simplified as:

[0141]

[0142] In the embodiment of the present application, the CQL algorithm process is as follows:

[0143] Initialize Q network Q θ 、Target Q Network Q θ′ and strategy π φ , Entropy regularization coefficient α

[0144] for training times t=1→T do

[0145] Update the entropy regularization coefficient:

[0146]

[0147] Update function Q:

[0148]

[0149] Update strategy:

[0150]

[0151] end for

[0152] In the embodiment of the present application, the input of the model is the historical data of distribution network fault location, which needs to be expressed in the form of a Markov decision process, as shown in the following formula:

[0153] (s,a,r,s′)

[0154] Among them, s is the current state, that is, the distribution network fault information when a fault occurs, a is the fault location implemented in the current state, that is, judging which node has a fault, r is the reward, that is, the feedback information after the action is taken, and s′ is the next state, that is, the distribution network fault information after the fault location action is performed.

[0155] The output of the model is the distribution network fault location strategy function π φ (a|s), the input of this function is state s, and the output is the probability distribution of action a. Generally speaking, the policy function trained by reinforcement learning outputs actions that are infinitely close to the deterministic policy and will not select worse actions.

[0156] In summary, the present invention proposes a distribution network fault location method, which obtains distribution network fault historical parameter data, and establishes a first distribution network fault location model based on the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device; the first distribution network fault location model is updated according to a preset error correction strategy to obtain a second distribution network fault location model; the second distribution network fault location model is converted into a Markov decision process, and the Markov decision process is solved according to an improved conservative Q learning algorithm to locate the distribution network fault. Matrix algorithms and optimization algorithms require accurate topological structure information and fault current data of the distribution network, and conservative Q learning algorithms do not need to collect real distribution network topological data. After the intelligent agent learns the strategy, it can directly perform fault location actions through the strategy function in the face of a changing distribution network environment without repeated iterative calculations.

[0157] This embodiment also provides a distribution network fault location system, including:

[0158] A first model building module is used to obtain distribution network fault historical parameter data and establish a first distribution network fault location model according to the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device;

[0159] A second model building module is used to update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model;

[0160] The positioning module is used to transform the second distribution network fault location model into a Markov decision process, and solve the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault.

[0161] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0162] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 8As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a distribution network fault location method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0163] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0164] Acquire distribution network fault historical parameter data, and establish a first distribution network fault location model according to the fault historical parameter data, the fault historical parameter data at least including current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device;

[0165] Update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model;

[0166] The second distribution network fault location model is transformed into a Markov decision process, and the Markov decision process is solved according to the improved conservative Q-learning algorithm to locate the distribution network fault.

[0167] Example 2

[0168] Reference Figure 2-Figure 7 , which is an embodiment of the present invention, provides a distribution network fault location method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0169] The intelligent agent is trained through the historical distribution network fault location data. The trained intelligent agent can have the fault location capability. The training process is as follows: Figure 6 Shown

[0170] The present invention is based on the IEEE-33 node distribution network system to verify the effectiveness of the proposed method. Figure 7 shown.

[0171] Node 21 is connected to a distributed power source with a base power of 100MVA. Figure 7 As shown, when a fault occurs in the connection line between node 11 and node 12, after inputting the fault information, the intelligent agent can locate the fault. Since it is offline training and online application, the response time can be ignored.

[0172] The comparison of the effects of the method of the present invention and the traditional method is shown in the following table:

[0173] Table 1 Algorithm performance comparison

[0174] method Accuracy Training time The response time of the algorithm after a failure occurs CQL Algorithm 98.6％ 124.07s 0.67s Traditional Algorithms 92.3％ 0 128.45

[0175] According to the comparison results, the CQL algorithm proposed in the present invention can improve the response time of the algorithm.

[0176] 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.

[0177] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0178] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0179] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0180] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0181] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0182] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A distribution network fault location method, It is characterized in that include: Acquire distribution network fault historical parameter data, and establish a first distribution network fault location model according to the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device; Update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model; The second distribution network fault location model is transformed into a Markov decision process, and the Markov decision process is solved according to an improved conservative Q-learning algorithm to perform distribution network fault location.

2. The distribution network fault location method according to claim 1, It is characterized in that The acquiring of distribution network fault historical parameter data and establishing a first distribution network fault location model according to the fault historical parameter data, wherein the fault historical parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device, includes: According to the distribution network fault historical parameter data, establishing the first distribution network fault location model includes locating the fault according to the historical monitoring data provided by the feeder terminal device, which is specifically expressed as follows: FTUs=(s 1 ,s 2 ,...,s k ...,s n ) Among them, s k represents the fault current direction information detected by the kth feeder terminal device. If the current direction is positive, then s k =1, if the current direction is negative, then s k = -1, if no fault current is detected, then s k =0; The fault current direction is that when there is distributed energy in the feeder section, the direction from the system power supply to the distributed energy is regarded as a positive direction, otherwise, the direction is regarded as a negative direction.

3. The distribution network fault location method according to claim 2, It is characterized in that The preset error correction strategy includes: Input initial current state S i (L), distribution network structure, fault point p; Number all distributed energy resources and feeder terminal devices; Based on the Dijkstra algorithm, for any i and j, find the path between the i-th distributed energy source and the j-th feeder terminal device; Determine whether there is a path between the i-th distributed energy source and the j-th fault point p; If it exists, correct the line status of the feeder where the distributed energy is located; Output the correct current state S(L).

4. The distribution network fault location method according to claim 3, It is characterized in that The preset error correction strategy also includes: The corrected objective function is as follows: Among them, σ represents the penalty coefficient set to 0.5-1.0 to avoid misjudgment, N FTU Indicates the number of feeder terminal devices; N L Indicates the number of lines, Indicates the state of the line. When the kth line fails, L k =1, otherwise, L k =0; K j represents the switch corresponding to the jth distributed energy; Indicates the fault current state corresponding to the fault line state L, FTU i Represents the fault current state detected by the i-th feeder terminal device.

5. The distribution network fault location method according to claim 4, It is characterized in that The converting of the second distribution network fault location model into a Markov decision process also includes: converting the distribution network fault location model into a Markov decision process by defining an agent, an environment, a state, an action and a reward, wherein the agent is a distribution network monitor or a fault location program, and the environment is the distribution network itself.

6. The method for locating a distribution network fault according to claim 5, It is characterized in that The method of solving the Markov decision process according to the improved conservative Q learning algorithm and locating the distribution network fault includes: initializing the Q network, updating the entropy regularization coefficient, function Q and strategy in the Q network, completing the design of the conservative Q learning algorithm, and solving the Markov decision process according to the designed conservative Q learning algorithm.

7. The method for locating a distribution network fault according to claim 6, It is characterized in that The method of solving the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault also includes: the conservative Q learning algorithm includes designing a conservative Q learning model, the input of the model is the historical data of distribution network fault location, and the output is the distribution network fault location strategy function.

8. A distribution network fault location system, It is characterized in that include: A first model building module is used to obtain distribution network fault history parameter data, and establish a first distribution network fault location model according to the fault history parameter data, wherein the fault history parameter data at least includes current, voltage, power related information when the fault occurs and historical monitoring data provided by the feeder terminal device; A second model building module is used to update the first distribution network fault location model according to a preset error correction strategy to obtain a second distribution network fault location model; The positioning module is used to convert the second distribution network fault positioning model into a Markov decision process, and solve the Markov decision process according to the improved conservative Q learning algorithm to locate the distribution network fault.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images