A method for locating faults in wind farm collector lines

Abstract

The present invention discloses a method for locating a fault in a wind farm collector line, comprising the following steps: constructing a reinforced positioning model based on positioning error points and positioning accurate points at the same monitoring time sequence, and using the reinforced positioning model to perform reinforcement learning on the fault positioning error model so that the fault positioning error model converges toward the reinforced positioning model to obtain a fault positioning accurate model. The present invention uses a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line on the basis of positioning technology, optimizes the trade-off between the positioning error and the positioning cost generated by the online monitoring points to achieve a primary reduction in positioning error, and uses the reinforced positioning model to perform reinforcement learning on the fault positioning error model so that the fault positioning error model converges toward the reinforced positioning model to obtain a fault positioning accurate model, so as to achieve a secondary reduction in positioning error to improve the accuracy of fault positioning of the wind farm collector line.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power line fault monitoring, and in particular to a method for locating a wind farm collector line fault. Background Art

[0002] Wind power generation occupies a large area, and the internal collection line wiring is complex. Currently, wind farm collection lines are mostly overhead, cable, T-connected branch line, and mixed line. The wind farm terrain is also complex, so when a wind farm collection line fails, it is often very difficult to find the fault point. Judging from the line outages that have occurred in the group's wind farms in recent years, the distance measurement given by the fault recording device and the protection device cannot accurately reflect the distance to the fault point, resulting in the wind farm being unable to find the fault point or failing to find the fault point for a long time, causing frequent line outages. Moreover, the collection lines of the Nanxuanfeng Wind Farm within the group are all overhead lines, and they are mountain wind farms. The Zhangguang Wind Farm is an overhead cable mixed structure, and the line structure is relatively complex, which makes it difficult to find line faults and takes a long time, affecting emergency repairs and power generation efficiency.

[0003] During the operation of the collector line, in addition to being susceptible to sudden faults such as lightning strikes, floating objects, wind deviation, etc., there are also a large number of "preventable" potential anomalies, such as insulator cracking, contamination, hardware breakage and detachment, vegetation overheight, ice cover, etc. If these hidden discharges are not handled in time, they will cause serious safety risks to the stable operation of the wind farm. Due to the lack of efficient technical means in routine inspections, it is impossible to directly identify and investigate the above hidden discharges before flashover. In the existing technology, the distance measurement given by the fault recording device and the protection device cannot accurately reflect the distance to the fault point, resulting in the wind farm being unable to find the fault point or failing to find the fault point for a long time, causing the line to frequently be in a non-stop state. Summary of the invention

[0004] The purpose of the present invention is to provide a method for locating faults in a wind farm collector line, so as to solve the technical problem that the distance measurement given by the fault recording device and the protection device in the prior art cannot accurately reflect the distance of the fault point, resulting in the situation that the fault point cannot be found in the wind farm or the fault point is not found for a long time, causing the line to frequently be in a non-stop state.

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

[0006] A method for locating a wind farm collector line fault comprises the following steps:

[0007] Step S1, constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location error, and constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location cost, and using a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line, so as to achieve a primary reduction in the positioning error by optimizing the trade-off between the positioning error and the positioning cost generated by the online monitoring points;

[0008] Step S2, obtaining a gradual discharge feature at an online monitoring point, and using the gradual discharge feature to locate the fault using the traveling wave location technology to obtain the line fault position as the location error point, and using the actual position of the line fault as the accurate location point, and constructing a fault location error model based on the gradual discharge feature and the location error point at the same monitoring time sequence;

[0009] Step S3, construct an enhanced positioning model based on the positioning error points and accurate positioning points at the same monitoring time sequence, and use the enhanced positioning model to perform reinforcement learning on the fault location error model so that the fault location error model converges towards the enhanced positioning model to obtain a fault location accuracy model, so as to achieve a secondary reduction in the positioning error and improve the fault location accuracy of the wind farm collector line.

[0010] As a preferred solution of the present invention, the step of constructing a fitting relationship between the number of online monitoring points of wind farm collector lines and the fault location error includes:

[0011] A plurality of values ​​are set for the number of online monitoring points of the wind farm collector line, and online monitoring points are set on the wind farm collector line according to the set value of each number of online monitoring points, and the fault location error corresponding to each number of online monitoring points is obtained by using the traveling wave positioning technology for fault location, wherein the fault location error is the position distance between the line fault position obtained by the traveling wave positioning technology for fault location and the actual position of the line fault;

[0012] The number of online monitoring points is used as an input item of the BP neural network, the fault location error is used as an output item of the BP neural network, and the BP neural network is used to perform model training on the input item and the output item of the BP neural network to obtain a positioning error calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location error;

[0013] The positioning error calculation formula is:

[0014] Pw=BP(n);

[0015] Where Pw is the fault location error, n is the number of online monitoring points, and BP is the neural network.

[0016] As a preferred solution of the present invention, the step of constructing a fitting relationship between the number of online monitoring points of wind farm collector lines and the fault location cost includes:

[0017] The online monitoring points are set on the wind power collection line according to the set value of the number of each online monitoring point, and the economic cost of setting the online monitoring points is calculated as the fault location cost according to the set value of the number of online monitoring points;

[0018] The number of online monitoring points is used as the second input item of the BP neural network, the fault location cost is used as the second output item of the BP neural network, and the BP neural network is used to perform model training on the second input item of the BP neural network and the second output item of the BP neural network to obtain a positioning cost calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location cost;

[0019] The positioning cost calculation formula is:

[0020] Rw=BP(n);

[0021] Where Rw is the fault location cost, n is the number of online monitoring points, and BP is the neural network.

[0022] As a preferred solution of the present invention, the method of setting the number of online monitoring points of the wind farm collector line by using a multi-objective optimization algorithm includes:

[0023] Taking the positioning error as the optimization target, a positioning error optimization function for the number of online monitoring points is constructed. The positioning error optimization function is:

[0024] F1 = min(Pw);

[0025] Taking the positioning cost as the optimization target, a positioning cost optimization function for the number of online monitoring points is constructed. The positioning cost optimization function is:

[0026] F2 = min(Rw);

[0027] Solving the positioning error optimization function and the positioning cost optimization function to obtain the number of online monitoring points that bidirectionally optimizes the fault positioning cost and the fault positioning error as the optimal number of online monitoring points;

[0028] The optimal online monitoring points are set on the wind power collection line according to the optimal number of online monitoring points to achieve the lowest fault location cost and the lowest fault location error.

[0029] As a preferred solution of the present invention, the gradual discharge characteristics are discharge characteristics of gradual faults, and the gradual faults include: composite insulator degradation, conductor hardware floating discharge, vegetation discharge, insulator pollution flashover, and insulator ice flashover.

[0030] As a preferred embodiment of the present invention, the fault location error model constructed based on the gradual discharge characteristics and location error points at the same monitoring time sequence includes:

[0031] Taking the gradual discharge characteristics and the location error points at the same monitoring time sequence as the input item and the output item of the CNN neural network respectively, and using the CNN neural network to perform network training on the input item and the output item of the CNN neural network to obtain the fault location error model;

[0032] The model expression of the fault location error model is:

[0033] P = CNN(S);

[0034] In the formula, P is the location error point, S is the gradual discharge characteristic, and CNN is the CNN neural network.

[0035] As a preferred embodiment of the present invention, the enhanced location model constructed based on the location error points and location accurate points at the same monitoring time sequence includes:

[0036] Taking the location accurate points and the location error points at the same monitoring time sequence as the input item and the output item of the reinforcement learning model respectively, and using the reinforcement learning model to perform network training on the input item and the output item of the reinforcement learning model to obtain the enhanced location model;

[0037] The model expression of the enhanced location model is:

[0038] P = Rmodel(Pd);

[0039] In the formula, Pd is the location accurate point, P is the location error point, and Rmodel is the reinforcement learning model.

[0040] As a preferred embodiment of the present invention, the use of the enhanced location model to perform reinforcement learning on the fault location error model to make the fault location error model converge towards the enhanced location model to obtain the fault location precision model includes:

[0041] Storing the location error points calculated by the fault location error model at each monitoring time sequence in the result matrix;

[0042] Using the reinforcement learning model to perform sampling training on the location error points and the location accurate points at each monitoring time sequence to realize iterative update of the reinforcement learning model, and outputting the location error points at each monitoring time sequence during each iterative update;

[0043] After each iteration of the reinforcement learning model, the fault location error model receives the output of the reinforcement learning model and performs update training at the corresponding monitoring time sequence to obtain the location error point at the corresponding monitoring time sequence, and constructs a loss function by combining the location error point with the location accuracy point.

[0044] The fault location error model is iteratively updated by minimizing the loss function, so as to achieve the convergence of the fault location error model to reinforcement learning to obtain the accurate fault location model.

[0045] As a preferred solution of the present invention, the loss function is a binary norm function of the position error between the positioning error point and the positioning accurate point.

[0046] As a preferred solution of the present invention, the gradual discharge characteristics at each monitoring time sequence are normalized.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention uses a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line on the basis of positioning technology, optimizes the trade-off between the positioning error and the positioning cost generated by the online monitoring points to achieve a primary reduction in the positioning error, and uses a reinforced positioning model to perform reinforcement learning on the fault location error model so that the fault location error model converges toward the reinforced positioning model to obtain a precise fault location model, thereby achieving a secondary reduction in the positioning error to improve the fault location accuracy of the wind farm collector line. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the implementation of the present invention or the technical solution in the prior art, the following briefly introduces the drawings required for the implementation or the prior art description. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0050] Figure 1 A flow chart of a method for locating a wind farm collector line fault provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] like Figure 1As shown, the present invention provides a method for locating a wind farm collector line fault, comprising the following steps:

[0053] Step S1, constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location error, and constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location cost, and using a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line, so as to achieve a primary reduction in the positioning error by optimizing the trade-off between the positioning error and the positioning cost generated by the online monitoring points;

[0054] Construct a fitting relationship between the number of online monitoring points of wind farm collector lines and the fault location error, including:

[0055] A plurality of values ​​are set for the number of online monitoring points of the wind farm collector line, and online monitoring points are set on the wind farm collector line according to the set value of each number of online monitoring points, and the fault location error corresponding to each number of online monitoring points is obtained by using the traveling wave positioning technology for fault location, and the fault location error is the position distance between the line fault position obtained by the traveling wave positioning technology for fault location and the actual position of the line fault;

[0056] The number of online monitoring points is used as the input item of the BP neural network, and the fault location error is used as the output item of the BP neural network. The BP neural network is used to train the input item and the output item of the BP neural network to obtain a positioning error calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location error.

[0057] The positioning error calculation formula is:

[0058] Pw=BP(n);

[0059] Where Pw is the fault location error, n is the number of online monitoring points, and BP is the neural network.

[0060] Construct a fitting relationship between the number of online monitoring points of wind farm collector lines and the fault location cost, including:

[0061] The online monitoring points are set on the wind power collection line according to the set value of the number of each online monitoring point, and the economic cost of setting the online monitoring points is calculated as the fault location cost according to the set value of the number of online monitoring points;

[0062] The number of online monitoring points is taken as the second input item of the BP neural network, and the fault location cost is taken as the second output item of the BP neural network. The BP neural network is used to train the second input item and the second output item of the BP neural network to obtain a location cost calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location cost.

[0063] The formula for calculating positioning cost is:

[0064] Rw=BP(n);

[0065] Where Rw is the fault location cost, n is the number of online monitoring points, and BP is the neural network.

[0066] The number of online monitoring points of wind farm collector lines is set using a multi-objective optimization algorithm, including:

[0067] Taking the positioning error as the optimization target, the positioning error optimization function of the number of online monitoring points is constructed. The positioning error optimization function is:

[0068] F1 = min(Pw);

[0069] Taking the positioning cost as the optimization target, the positioning cost optimization function of the number of online monitoring points is constructed. The positioning cost optimization function is:

[0070] F2 = min(Rw);

[0071] The positioning error optimization function and the positioning cost optimization function are solved to obtain the number of online monitoring points that optimizes the fault positioning cost and the fault positioning error in both directions as the optimal number of online monitoring points;

[0072] The optimal online monitoring points are set on the wind power collection line according to the optimal number of online monitoring points to achieve the lowest fault location cost and the lowest fault location error.

[0073] In fact, the accuracy of traveling wave positioning technology will be affected by terrain sag. The influence of terrain and sag may cause the error of the actual conductor length and the straight-line distance between the poles and towers to reach more than 5%. For a 60-kilometer line, 5% means a positioning error of 3 kilometers. If it is reduced to a monitoring section of 10 kilometers, the positioning error will be reduced to 1 / 6 kilometers. Therefore, the more online monitoring points there are and the shorter the monitoring section is, the smaller the fault location error will be. However, the more online monitoring points there are, the higher the equipment labor cost will be, resulting in a higher fault location cost. In this embodiment, the fault location cost and the fault location error are used as optimization functions. The lower the fault location error, the more online monitoring points are expected to be. The lower the fault location cost, the fewer online monitoring points are expected to be. When the positioning error is minimized and the positioning cost is minimized for target optimization, the number of online monitoring points obtained can be weighed between the two contradictory optimization goals, and a number of online monitoring points is determined so that a lower positioning cost can be achieved while achieving a lower positioning error.

[0074] The number of online monitoring points is set to reduce the positioning error, but there is still a positioning error caused by the influence of terrain sag. This embodiment uses a reinforcement model to perform reinforcement learning on the fault location model to improve the positioning accuracy of the fault location model. On the basis of setting the number of online monitoring points to reduce the positioning error, the positioning error is further reduced. In the model, reinforcement learning is used to overcome the influence of terrain sag on positioning from a mathematical model. The specific steps are as follows:

[0075] Step S2, obtaining a gradual discharge feature at an online monitoring point, and using the gradual discharge feature to locate the fault using the traveling wave location technology to obtain the line fault position as the location error point, and using the actual position of the line fault as the accurate location point, and constructing a fault location error model based on the gradual discharge feature and the location error point at the same monitoring time sequence;

[0076] The gradual discharge characteristics are the discharge characteristics of gradual faults. The discharge characteristics include electrical characteristics such as current, voltage, and power. Gradual faults include: composite insulator deterioration, floating discharge of conductor hardware, vegetation discharge, insulator pollution flashover, and insulator ice flashover. For gradual faults that have not occurred, they should be discovered and eliminated as early as possible to prevent them from happening, thereby improving the real-time monitoring and intelligent operation and maintenance level of wind farm collection lines, significantly shortening power outage time, reducing line tripping rate, and greatly reducing production impact and economic losses caused by delayed fault processing, which has huge economic benefits. Monitoring hidden dangers will help power supply section operation and maintenance personnel to inspect the hidden danger locations, deal with equipment damage and hidden danger locations caused by hidden dangers in advance, and avoid major accidents such as line breaks that may be caused by long-term operation of equipment with problems.

[0077] A fault location error model is constructed based on the gradual discharge characteristics and location error points at the same monitoring time sequence, including:

[0078] The gradual discharge characteristics and positioning error points at the same monitoring time sequence are used as the input items and output items of the CNN neural network respectively, and the CNN neural network is used to train the input items and output items of the CNN neural network to obtain the fault location error model.

[0079] The model expression of the fault location error model is:

[0080] P = CNN(S);

[0081] Where P is the positioning error point, S is the gradual discharge feature, and CNN is the CNN neural network.

[0082] Step S3, construct an enhanced positioning model based on the positioning error points and accurate positioning points at the same monitoring time sequence, and use the enhanced positioning model to perform reinforcement learning on the fault location error model so that the fault location error model converges towards the enhanced positioning model to obtain a fault location accuracy model, so as to achieve a secondary reduction in the positioning error and improve the fault location accuracy of the wind farm collector line.

[0083] An enhanced positioning model is constructed based on the positioning error points and accurate positioning points at the same monitoring time sequence, including:

[0084] The accurate positioning point and the positioning error point at the same monitoring time sequence are used as the input item and the output item of the reinforcement learning model respectively, and the reinforcement learning model is used to perform network training on the input item and the output item of the reinforcement learning model to obtain the reinforcement positioning model;

[0085] The model expression of the enhanced positioning model is:

[0086] P = Rmodel(Pd);

[0087] In the formula, Pd is the accurate positioning point, P is the positioning error point, and Rmodel is the reinforcement learning model.

[0088] The fault location error model is reinforced by using the reinforcement location model to make the fault location error model converge to the reinforcement location model to obtain an accurate fault location model, including:

[0089] The location error points calculated by the fault location error model at each monitoring time sequence are stored in the result matrix;

[0090] The reinforcement learning model is used to sample and train the positioning error points and accurate positioning points at each monitoring time sequence to achieve iterative update of the reinforcement learning model, and the positioning error points at each monitoring time sequence are output in each iterative update;

[0091] After each iteration of the reinforcement learning model, the fault location error model receives the output of the reinforcement learning model and performs update training at the corresponding monitoring time sequence to obtain the location error point at the corresponding monitoring time sequence, and constructs a loss function by combining the location error point with the location accuracy point.

[0092] The fault location error model is iteratively updated by minimizing the loss function to achieve the convergence of the fault location error model to reinforcement learning to obtain an accurate fault location model.

[0093] Reinforcement learning is used to continuously iterate and update the fault location error model to converge, so that the output result of the fault location error model is closer and closer to the accurate positioning point, thereby achieving the convergence and true position of the fault position according to the gradual discharge characteristics, and improving the positioning accuracy.

[0094] The loss function is the binary norm function of the position error between the positioning error point and the positioning accurate point.

[0095] The gradual discharge characteristics at each monitoring time sequence are normalized.

[0096] The present invention uses a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line on the basis of positioning technology, optimizes the trade-off between the positioning error and the positioning cost generated by the online monitoring points to achieve a primary reduction in the positioning error, and uses a reinforced positioning model to perform reinforcement learning on the fault location error model so that the fault location error model converges toward the reinforced positioning model to obtain a precise fault location model, thereby achieving a secondary reduction in the positioning error to improve the fault location accuracy of the wind farm collector line.

[0097] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for locating faults in wind farm collector lines. It is characterized in that The following steps are involved: Step S1, constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location error, and constructing a fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location cost, and using a multi-objective optimization algorithm to set the number of online monitoring points of the wind farm collector line, so as to achieve a primary reduction in the positioning error by optimizing the trade-off between the positioning error and the positioning cost generated by the online monitoring points; Step S2, obtaining a gradual discharge feature at an online monitoring point, and using the gradual discharge feature to locate the fault using the traveling wave location technology to obtain the line fault position as the location error point, and using the actual position of the line fault as the accurate location point, and constructing a fault location error model based on the gradual discharge feature and the location error point at the same monitoring time sequence; Step S3, construct an enhanced positioning model based on the positioning error points and accurate positioning points at the same monitoring time sequence, and use the enhanced positioning model to perform reinforcement learning on the fault location error model so that the fault location error model converges towards the enhanced positioning model to obtain a fault location accuracy model, so as to achieve a secondary reduction in the positioning error and improve the fault location accuracy of the wind farm collector line.

2. A method for locating a wind farm collector line fault according to claim 1, Features: The constructing of the fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location error includes: A plurality of values ​​are set for the number of online monitoring points of the wind farm collector line, and online monitoring points are set on the wind farm collector line according to the set value of each number of online monitoring points, and the fault location error corresponding to each number of online monitoring points is obtained by using the traveling wave positioning technology for fault location, wherein the fault location error is the position distance between the line fault position obtained by the traveling wave positioning technology for fault location and the actual position of the line fault; The number of online monitoring points is used as an input item of the BP neural network, the fault location error is used as an output item of the BP neural network, and the BP neural network is used to perform model training on the input item and the output item of the BP neural network to obtain a positioning error calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location error; The positioning error calculation formula is: Pw=BP(n); Where Pw is the fault location error, n is the number of online monitoring points, and BP is the neural network.

3. A method for locating a wind farm collector line fault according to claim 2, Features: The construction of the fitting relationship between the number of online monitoring points of the wind farm collector line and the fault location cost includes: The online monitoring points are set on the wind power collection line according to the set value of the number of each online monitoring point, and the economic cost of setting the online monitoring points is calculated as the fault location cost according to the set value of the number of online monitoring points; The number of online monitoring points is used as the second input item of the BP neural network, the fault location cost is used as the second output item of the BP neural network, and the BP neural network is used to perform model training on the second input item of the BP neural network and the second output item of the BP neural network to obtain a positioning cost calculation formula that characterizes the fitting relationship between the number of online monitoring points and the fault location cost; The positioning cost calculation formula is: Rw=BP(n); Where Rw is the fault location cost, n is the number of online monitoring points, and BP is the neural network.

4. A method for locating a wind farm collector line fault according to claim 3, Features: The method of setting the number of online monitoring points of the wind farm collector line by using a multi-objective optimization algorithm includes: Taking the positioning error as the optimization target, a positioning error optimization function for the number of online monitoring points is constructed. The positioning error optimization function is: F1 = min(Pw); Taking the positioning cost as the optimization target, a positioning cost optimization function for the number of online monitoring points is constructed. The positioning cost optimization function is: F2 = min(Rw); Solving the positioning error optimization function and the positioning cost optimization function to obtain the number of online monitoring points that bidirectionally optimizes the fault positioning cost and the fault positioning error as the optimal number of online monitoring points; The optimal online monitoring points are set on the wind power collection line according to the optimal number of online monitoring points to achieve the lowest fault location cost and the lowest fault location error.

5. A method for locating a wind farm collector line fault according to claim 4, Features: The gradual discharge characteristics are discharge characteristics of gradual faults, and the gradual faults include: composite insulator degradation, conductor hardware floating discharge, vegetation discharge, insulator pollution flashover, and insulator ice flashover.

6. A method for locating a wind farm collector line fault according to claim 5, Features: The fault location error model is constructed based on the gradual discharge characteristics and the location error points at the same monitoring time sequence, including: The gradual discharge characteristics at the same monitoring time sequence and the positioning error point are used as the input item of the CNN neural network and the output item of the CNN neural network respectively, and the CNN neural network is used to perform network training on the input item of the CNN neural network and the output item of the CNN neural network to obtain the fault location error model; The model expression of the fault location error model is: P = CNN(S); Where P is the positioning error point, S is the gradual discharge feature, and CNN is the CNN neural network.

7. A method for locating a wind farm collector line fault according to claim 6, It is characterized in that The enhanced positioning model is constructed based on the positioning error points and the positioning accurate points at the same monitoring time sequence, including: The accurate positioning point and the positioning error point at the same monitoring time sequence are used as the input item and the output item of the reinforcement learning model respectively, and the reinforcement learning model is used to perform network training on the input item and the output item of the reinforcement learning model to obtain the reinforcement positioning model; The model expression of the enhanced positioning model is: P = Rmodel(Pd); In the formula, Pd is the accurate positioning point, P is the positioning error point, and Rmodel is the reinforcement learning model.

8. A method for locating a wind farm collector line fault according to claim 7, It is characterized in that The method of using the enhanced positioning model to perform reinforcement learning on the fault positioning error model so that the fault positioning error model converges toward the enhanced positioning model to obtain an accurate fault positioning model includes: The location error points calculated by the fault location error model at each monitoring time sequence are stored in the result matrix; The reinforcement learning model is used to sample and train the positioning error points and the positioning accurate points at each monitoring time sequence to achieve iterative update of the reinforcement learning model, and the positioning error points at each monitoring time sequence are output during each iterative update; After each iteration of the reinforcement learning model, the fault location error model receives the output of the reinforcement learning model and performs update training at the corresponding monitoring time sequence to obtain the location error point at the corresponding monitoring time sequence, and constructs a loss function by combining the location error point with the location accuracy point. The fault location error model is iteratively updated by minimizing the loss function, so as to achieve the convergence of the fault location error model to reinforcement learning to obtain the accurate fault location model.

9. A method for locating a wind farm collector line fault according to claim 8, It is characterized in that The loss function is a binary norm function of the position error between the positioning error point and the positioning accurate point.

10. A method for locating a wind farm collector line fault according to claim 9, It is characterized in that The gradual discharge characteristics at each monitoring time sequence are normalized.

Images