Quad-rotor unmanned aerial vehicle health state assessment method and system for substation inspection

Abstract

The invention provides a substation inspection quad-rotor unmanned aerial vehicle health state assessment method and device, and the method comprises the steps: carrying out the working condition division operation of a process in which a quad-rotor unmanned aerial vehicle executes an inspection flight task, and constructing an unmanned aerial vehicle health state assessment data set; analyzing the characteristics of the operation parameters of the unmanned aerial vehicle under each working condition, respectively performing forward processing to obtain a standardized matrix, a model optimal solution vector and a model worst solution vector, comparing the model optimal solution vector with the model worst solution vector, and updating to obtain an ideal superior and inferior solution vector; calculating the weight of each index by using a combined weighting method, and solving a distance value and a health state index value between the flight data of the to-be-evaluated unmanned aerial vehicle and the ideal superior and inferior solution vector; dividing health state grades, constructing a membership function, and determining the health state of the unmanned aerial vehicle. The method solves the technical problems that the subjective degree of weighting is high and the health assessment precision of the unmanned aerial vehicle is restricted due to the fact that unbalanced characteristics of positive and negative samples are not fully considered and individual indexes are abnormal under multiple working conditions.

Description

Technical Field

[0001] The present invention relates to the field of monitoring and control of substation inspection drones, and in particular to a health status assessment method and system for substation inspection quad-rotor drones. Background Art

[0002] The existing invention patent application document "Health Status Assessment Method for Substation Inspection Quad-rotor UAV" with publication number CN118395070A, the existing method includes: obtaining dynamic parameters and basic parameters of the UAV; calculating the abnormal probability value of the dynamic parameter abnormality according to the normal range value of the dynamic parameter; using the probability membership distribution function to calculate the abnormal probability value to obtain a first membership matrix; determining the parameter abnormality severity level score corresponding to each dynamic parameter, and using the severity membership distribution function to calculate the maximum score value corresponding to each dynamic parameter to obtain a second membership matrix; grey-correlating the first membership matrix and the second membership matrix with each health status level vector respectively to obtain a first health status membership vector; using the degradation membership distribution function to calculate each basic parameter to obtain a second health status membership vector; based on the first health status membership vector and the second health status membership vector, determining the comprehensive health status of the UAV. However, the above existing scheme considers the abnormal probability and severity of dynamic parameters and the degradation of basic parameters to evaluate the health status of drones, but the frequency of drone flight data reporting is high, there is an imbalance between positive and negative flight samples, and the abnormal probability cannot measure the health status of drones well. The determination of its severity and the weight vector of each indicator based on the hierarchical analysis method rely on expert scoring, and subjective factors have a great influence on the evaluation results. At the same time, in the above existing scheme, the abnormal probability and severity of dynamic parameters and the degradation of basic parameters are considered to evaluate the health status of drones, but the frequency of drone data reporting is high, and the number of abnormal samples is relatively small, which cannot reflect the health status of drones well. In addition, the severity relies on expert scoring, and the weight of its parameters adopts the hierarchical analysis method, which has a strong subjective influence.

[0003] The existing invention patent application document "System and method for safety risk assessment of drone take-off and landing based on multimodal data fusion" with publication number CN117829599A, in this existing system, the multimodal data perception module is used to transfer the acquired data to the risk assessment module; the risk assessment module calculates the quantitative value of the risk of drone take-off and landing on slippery road surface; the risk warning module predicts the future risk change trend through the quantitative value of the risk of drone take-off and landing on slippery road surface, and gives a warning. The method includes constructing a risk assessment index system for drone take-off and landing on slippery road surface; establishing a risk assessment index system for drone take-off and landing on slippery road surface; determining the influence weight of each risk factor index; and dividing the risk level. The existing scheme considers the slippery state of the road surface, the performance of the drone, and the ability of the drone pilot as evaluation factors for the risk assessment of drone take-off and landing, adopts hierarchical scoring and quantification for each evaluation index, adopts the combined weighting method of hierarchical analysis method and coefficient of variation method to determine the weight vector of each index, and combines fuzzy comprehensive evaluation to obtain the risk value. At the same time, the aforementioned existing scheme considers the slippery state of the road surface, the performance of the UAV, and the ability of the UAV pilot as evaluation factors for the risk assessment of UAV take-off and landing, adopts a hierarchical scoring quantification for each evaluation indicator, and uses the hierarchical analysis method and the coefficient of variation method to combine the weighting method to determine the weight vector of each indicator, and combines the fuzzy comprehensive evaluation to obtain the risk value. However, for the case where only individual parameters are abnormal, it is impossible to intuitively reflect their impact on the health status.

[0004] In summary, the existing technology has technical problems such as high subjectivity in weighting and restricting the accuracy of UAV health assessment due to insufficient consideration of the imbalance characteristics of positive and negative samples and the abnormality of individual indicators under multiple working conditions. Summary of the invention

[0005] The technical problem to be solved by the present invention is: how to solve the technical problem in the prior art that the imbalance characteristics of positive and negative samples and the abnormalities of individual indicators under multiple working conditions are not fully considered, resulting in a high degree of subjectivity in weighting and restricting the accuracy of drone health assessment.

[0006] The present invention adopts the following technical solutions to solve the above technical problems: A method for evaluating the health status of a substation inspection quad-rotor drone comprises:

[0007] S1. During the process of the quad-rotor UAV performing the inspection flight mission, the working condition division operation is performed, the operating parameter data of the quad-rotor UAV under each working condition is collected, and the UAV health status assessment data set is constructed according to the working condition;

[0008] S2. Analyze the characteristic information of the operating parameters of the quadrotor drone under each working condition, perform forward processing respectively, and obtain the standardized matrix under each working condition to obtain the model optimal solution vector and the model worst solution vector of the drone operating parameters;

[0009] S3, calculating the optimal solution vector and the worst solution vector when inputting real-time parameters, comparing them with the optimal solution vector and the worst solution vector of the model, and updating the ideal solution vector;

[0010] S4. Combining the variable weight method and entropy weight method, a combined weighting method is designed. The weight of each indicator is calculated using the weighting method to obtain the distance value d between the flight data of the drone to be evaluated and the ideal good and bad solution vector. + d - , and process to obtain the health status index value h:

[0011] S5. Calculate the real-time operating parameters reported by the quadrotor drone and the distance from the ideal good and bad solution vector, calculate the health index h according to the distance, divide the health status level, construct the membership function of the health index h to the health status, and determine the health status of the drone according to the maximum membership principle.

[0012] The present invention aims at evaluating the health status of a quad-rotor UAV for substation inspection. Taking into account the differences in indicator types under multiple working conditions of UAV flight missions, the parameter characteristics under multiple working conditions are analyzed, and the superior and inferior solutions under each working condition are calculated. Taking into account the imbalanced characteristics of positive and negative samples of flight data, the superior and inferior solutions are updated, and a combined weighting method combining variable weight theory and entropy weight method is used to highlight the influence of individual indicator abnormalities on the evaluation results, thereby weakening the influence of subjective weighting. Finally, the ideal solution distance is used as a measure of the health indicator, and a membership function is established to determine the final health status of the UAV.

[0013] In a more specific technical solution, S1 includes:

[0014] S11, the collected flight data sample set {x1, x2, ..., x m} as input, determine the number of clusters k, C = {C1, C2, ..., C k} is a set of clusters, and sets the maximum number of iterations p;

[0015] S12. From the sample set in the drone health status assessment dataset, determine k samples as the initial centroid μ i , i=1,2,…,k;

[0016] Among them, a data point set is obtained and input from the drone health status assessment data set. From the data point set, a point is randomly selected as the first cluster center μ1, and each point x is calculated. i The distance from the nearest cluster center among the selected cluster centers is selected, and a point whose D(x) is greater than the preset threshold is selected as the new cluster center:

[0017] D(x i )=arg min||x i -μ j|| (1)

[0018] In the formula, μ j is the currently selected centroid vector, arg min represents the variable value when the function reaches the minimum value;

[0019] S13, repeat S12 until k centroids are selected;

[0020] S14. Calculate sample x i With the center of mass μ j The Euclidean distance between the samples x i Mark the cluster label value corresponding to the centroid with the smallest distance, and update to obtain k cluster sets;

[0021] S15. For each sample point in any cluster, use the following logic to recalculate the new centroid:

[0022]

[0023] S16. Repeat S13 and S14 until it is determined that all the centers of mass no longer change, and obtain the final UAV working condition classification result C.

[0024] The present invention divides the flight mission of the quad-rotor UAV into multiple working conditions in detail according to the characteristics of the flight mission, and analyzes the parameter characteristics of the evaluation index under different working conditions according to the characteristics of each working condition, so as to know and evaluate the health status of the UAV in a more refined way.

[0025] The present invention establishes working condition clustering for the original flight data of the quad-rotor drone for the data set under various working conditions. Taking into account that the selection of the initial center point will have a great influence on the iteration efficiency and iteration effect, the original arbitrary selection of the centroid equal to the number of working conditions as the initial centroid is abandoned. A centroid is first selected to calculate its distance with the sample point, and the second centroid is selected based on this, and the selections are made in sequence until the centroid equal to the number of working conditions is selected for subsequent calculations, which greatly improves the iteration efficiency.

[0026] In a more specific technical solution, S2 includes:

[0027] S21. Analyze the parameter index types of the selected evaluation parameters under different working conditions, and take the collected parameters x i =(x i1 ,x i2 ,…,x im ) constitutes the health evaluation matrix X n×m =(x1,x2,…,x n ) T ;

[0028] S22, for the evaluation matrix under each working condition, according to the parameter index type, corresponding forward processing is performed to obtain a forward matrix, and the forward matrix is ​​dimensionlessly processed to obtain a standardized evaluation matrix;

[0029] S23. Based on the standardized evaluation matrix, calculate the optimal solution vector and the worst solution vector s of the model of the operating data of the quadcopter under each working condition. + 、s - :

[0030] s + =[max{y 11 ,y 21 ,…,y n1},max{y 12 ,y 22 ,…,y n2},…,max{y 1m ,y 2m ,…,y nm}](7)

[0031] s - =[min{y 11 ,y 21 ,…,y n1},min{y 12 ,y 22 ,…,y n2},…,min{y 1m ,y 2m ,…,y nm}] (8)

[0032] In the formula, x best is the optimal ideal value of the intermediate indicator, a, b are the optimal interval boundaries [a, b] of the interval indicator;

[0033] In a more specific technical solution, S22 also includes:

[0034] The flight data of the quadrotor drone under each working condition is processed forward and dimensionless to obtain the standardized matrix Y n×m ; Among them, the following logic is used to positively process extremely small indicators:

[0035]

[0036] Use the following logic to positively process the intermediate indicators:

[0037]

[0038] Use the following logic to positively process the interval indicators:

[0039]

[0040]

[0041] In a more specific technical solution, in S3, the data to be evaluated is input into the health evaluation matrix X under the corresponding working condition according to the corresponding working condition category. (n+1)×m , perform forward processing and dimensionless processing to obtain a new standardized matrix to obtain the ideal good and bad solution vector s' + 、s' - :

[0042] s *+ =max{s + ,s' +} (9)

[0043] s *- =min{s - ,s' _} (10).

[0044] The present invention calculates health indicators based on ideal superior and inferior solutions, makes full use of historical UAV flight data, and considers that the ratio of positive and negative samples in UAV flight data differs greatly. The method of updating the ideal solution vector is adopted to improve the ideal solution vector in real time, making the evaluation result more reliable.

[0045] In a more specific technical solution, in S4, the distance value d is processed using the following logic: + d - :

[0046]

[0047]

[0048] Among them, ω j , j = 1, 2, ..., m + 1, is the jth element of the weight vector ω of the evaluation index,

[0049]

[0050] In a more specific technical solution, S4 also includes:

[0051] S41, establish a pairwise evaluation matrix A of evaluation indicators, and obtain a constant weight vector w1 by the hierarchical analysis method;

[0052] S42, constructing a state variable weight vector G, processing to obtain a variable weight matrix W, taking the last line of the variable weight matrix as the variable weight vector w2 of the input parameter; wherein, the state variable weight vector is constructed using the following logic:

[0053] G(Y i )=[G1(Yi ),G2(Y i ),…,G n (Y i ),G n+1 (Y i )]as follows:

[0054]

[0055] In the formula, i = 1, 2, ..., n + 1; j = 1, 2, ..., m; α> 0 is the penalty level factor; 0 < β ≤ 1; the variable weight vector G is obtained (n+1)×m ;

[0056] The following logic is used to calculate the variable weight matrix W, and the variable weight vector w2 of the last row of the variable weight matrix is ​​used as the real-time input data as the current state weight;

[0057]

[0058] S43, standardize the matrix Y n×m Perform forward and standardization processing to obtain the probability matrix P;

[0059]

[0060] S44. Calculate the information entropy e of each indicator based on the probability matrix P j And calculate the entropy weight method weight w3;

[0061] S45, obtaining a final weight vector w according to the combined weighting method, wherein the weight vector w is obtained by using the following logical processing according to the combined weighting method;

[0062] w=λw2+(1-λ)w3 (19)

[0063] In the formula, λ is the weight under variable weight empowerment.

[0064] Aiming at the situation that the constant weight cannot well reflect the abnormality of individual indicators, the present invention adopts the combination weighting method of hierarchical analysis method combined with variable weight theory and entropy weight method and the superior and inferior solution distance method to make the evaluation result more accurate and reliable.

[0065] Specifically, the present invention reflects the abnormal situation of individual parameters through variable weight vectors in response to the abnormal situation of individual parameters, and combines the subjective factors of expert scoring based on the hierarchical analysis method and the objective factors of the entropy weight method, so as to avoid the excessive influence of the variable weight method on the health status when the parameters with smaller impact on the health status are abnormal, and reflects the abnormal situation of various parameters in real time with dynamic weights, so that the evaluation results are more accurate.

[0066] In a more specific technical solution, in S44, the information entropy e of each indicator is calculated using the following logic:j ;

[0067]

[0068] Using the following logic, calculate the entropy weight method weight w3;

[0069]

[0070] In a more specific technical solution, in S5, a membership function that conforms to the Gaussian distribution is used as the health status membership function of the health indicator. According to the maximum membership principle, the health status of the drone is determined using the following logic.

[0071]

[0072] Where c is the mean of the curve and σ is the standard deviation.

[0073] The present invention takes into account the imbalance of positive and negative samples of drones, starts from the data reported by the drones, mines the ideal good and bad solutions of the drone data themselves, and characterizes the health status of the drone by the distance from the ideal good and bad solutions. In the process, the combined weighting of AHP-variable weight and entropy weight is considered, and the dynamic weight vector reflects the abnormal situation of the parameters in real time. The evaluation result combines subjective and objective factors to make the result more accurate.

[0074] In a more specific technical solution, a substation inspection quadcopter drone health status assessment system includes:

[0075] The data set construction module is used to divide the working conditions of the quad-rotor UAV during the inspection flight mission, collect the operating parameter data of the quad-rotor UAV under each working condition, and construct the UAV health status assessment data set according to the working condition;

[0076] The standardized matrix construction module is used to analyze the characteristic information of the operating parameters of the quadcopter under each working condition, and perform forward processing respectively to obtain the standardized matrix under each working condition, so as to obtain the model optimal solution vector and the model worst solution vector of the operating parameters of the drone. The standardized matrix construction module is connected with the data set construction module;

[0077] The good and bad solution vector updating module is used to calculate the optimal solution vector and the worst solution vector when the real-time parameters are input, compare them with the model optimal solution vector and the model worst solution vector, and update the ideal good and bad solution vector. The good and bad solution vector updating module is connected with the standardized matrix construction module;

[0078] The combined weighting module is used to combine the variable weight method and the entropy weight method to design a combined weighting method, and use the weighting method to calculate the weight of each indicator to obtain the distance value d between the flight data of the drone to be evaluated and the ideal good and bad solution vector. + d- , the health status index value h is obtained by processing, and the combined weighting module is connected with the good and bad solution vector update module:

[0079] The drone health status determination module is used to calculate the real-time operating parameters reported by the quadrotor drone and the distance from the ideal good and bad solution vector, calculate the health index h based on the distance, divide the health status level, and construct the membership function of the health index h to the health status. According to the principle of maximum membership, the drone health status is determined. The drone health status determination module is connected to the combined weighting module.

[0080] Compared with the prior art, the present invention has the following advantages:

[0081] The present invention aims at evaluating the health status of a quad-rotor UAV for substation inspection. Taking into account the differences in indicator types under multiple working conditions of UAV flight missions, the parameter characteristics under multiple working conditions are analyzed, and the superior and inferior solutions under each working condition are calculated. Taking into account the imbalanced characteristics of positive and negative samples of flight data, the superior and inferior solutions are updated, and a combined weighting method combining variable weight theory and entropy weight method is used to highlight the influence of individual indicator abnormalities on the evaluation results, thereby weakening the influence of subjective weighting. Finally, the ideal solution distance is used as a measure of the health indicator, and a membership function is established to determine the final health status of the UAV.

[0082] The present invention divides the flight mission of the quad-rotor UAV into multiple working conditions in detail according to the characteristics of the flight mission, and analyzes the parameter characteristics of the evaluation index under different working conditions according to the characteristics of each working condition, so as to know and evaluate the health status of the UAV in a more refined way.

[0083] The present invention establishes working condition clustering for the original flight data of the quad-rotor drone for the data set under various working conditions. Taking into account that the selection of the initial center point will have a great influence on the iteration efficiency and iteration effect, the original arbitrary selection of the centroid equal to the number of working conditions as the initial centroid is abandoned. A centroid is first selected to calculate its distance with the sample point, and the second centroid is selected based on this, and the selections are made in sequence until the centroid equal to the number of working conditions is selected for subsequent calculations, which greatly improves the iteration efficiency.

[0084] The present invention calculates health indicators based on ideal superior and inferior solutions, makes full use of historical UAV flight data, and considers that the ratio of positive and negative samples in UAV flight data differs greatly. The method of updating the ideal solution vector is adopted to improve the ideal solution vector in real time, making the evaluation result more reliable.

[0085] Aiming at the situation that the constant weight cannot well reflect the abnormality of individual indicators, the present invention adopts the combination weighting method of hierarchical analysis method combined with variable weight theory and entropy weight method and the superior and inferior solution distance method to make the evaluation result more accurate and reliable.

[0086] The present invention focuses on mining the UAV flight data itself, divides the UAV flight mission into multiple working conditions, calculates and analyzes the ideal solution vector under each working condition, overcomes the problem of imbalanced positive and negative samples of the UAV, and combines the variable weight method to highlight the impact of abnormal parameters on the health status, wherein the subjective weight of the hierarchical analysis method and the objective weight of the entropy weight method are considered, avoiding the excessive impact of the variable weight method on the health status when the parameters with little impact on the health status are abnormal, thereby ensuring the accuracy and rationality of the evaluation results.

[0087] The present invention solves the technical problem in the prior art that the imbalance characteristics of positive and negative samples and the abnormality of individual indicators under multiple working conditions are not fully considered, resulting in a high degree of subjectivity in weighting and restricting the accuracy of drone health assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 This is a schematic diagram of the basic steps of a health status assessment method for a substation inspection quad-rotor drone according to Embodiment 1 of the present invention;

[0089] Figure 2 This is a schematic diagram of specific steps for establishing a data set of various working conditions of a drone according to Embodiment 1 of the present invention;

[0090] Figure 3 This is a schematic diagram of specific steps for establishing a standardized evaluation matrix for health assessment of drones under multiple working conditions according to Example 1 of the present invention;

[0091] Figure 4 This is a schematic diagram of the specific steps of calculating the weight vector of the health assessment parameters of the UAV under multiple working conditions according to the second embodiment of the present invention. DETAILED DESCRIPTION

[0092] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are 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.

[0093] Example 1

[0094] like Figure 1 As shown, the present invention provides a method for evaluating the health status of a substation inspection quad-rotor drone, comprising the following basic steps:

[0095] S1. The process of the quadrotor drone performing the inspection flight mission is divided into four working conditions: lifting, moving forward, turning, and hovering. The operating parameter data of the drone under each working condition is collected to construct a training set.

[0096] In this embodiment, the UAV flight reporting data is collected, and according to the divided working condition categories, a UAV health status assessment data set under each working condition is established;

[0097] like Figure 2 As shown, step S1 of establishing the data set of each working condition of the UAV also includes the following specific steps:

[0098] S11, the collected flight data sample set {x1, x2, ..., x m} as input, determine the number of clusters k, C = {C1, C2, ..., C k} is a set of clusters, and sets the maximum number of iterations p;

[0099] S12, determine k samples from the sample set as the initial centroid μ i , i=1,2,…,k;

[0100] In this embodiment, a point is randomly selected from the input data point set as the first cluster center μ1, and the value of each point x is calculated. i The distance from the nearest cluster center among the selected cluster centers, and then select a point with a larger D(x) as the new cluster center:

[0101] D(x i )=arg min||x i -μ j || (1)

[0102] Among them, μ j is the currently selected centroid vector, arg min represents the variable value when the function reaches the minimum value;

[0103] S13, repeat the above step S12 until k centroids are selected;

[0104] S14. Calculate sample x i With the center of mass μ j The Euclidean distance between i Mark the cluster label value corresponding to the centroid with the smallest distance, and update the k cluster sets;

[0105] S15. Recalculate the new centroid for each sample point in any cluster:

[0106]

[0107] S16. Repeat the above steps S13 and S14 until it is determined that all the centers of mass no longer change, and obtain the final UAV working condition classification result C.

[0108] S2. According to the characteristics of the UAV operating parameters under each working condition, forward processing is performed respectively to obtain the standardized matrix under each working condition, and the optimal and worst solution vectors of the UAV operating parameters are obtained;

[0109] like Figure 3 As shown, in this embodiment, step S2 of establishing a standardized evaluation matrix for health assessment of drones under multiple working conditions also includes the following specific steps:

[0110] S21. Analyze the parameter index types of the selected evaluation parameters under different working conditions. The parameter index types are divided into very small indexes, intermediate indexes, very large indexes and interval indexes;

[0111] In this embodiment, the types of health assessment indicators of drones under various working conditions are analyzed, and the collected parameters x i =(x i1 ,x i2 ,…,x im ) constitutes the evaluation matrix X n×m =(x1,x2,…,x n ) T ;

[0112] S22, performing corresponding forward processing on the data under each working condition according to the parameter type, and performing dimensionless processing on the forward matrix to obtain a standardized evaluation matrix;

[0113] In this embodiment, the flight data of the UAV under the above working conditions are forward processed, and then dimensionless processed to obtain the standardized matrix Y n×m ;

[0114] In this embodiment, the following formula is used to perform positive transformation of the extremely small index:

[0115]

[0116] Use the following formula to make the intermediate indicators positive:

[0117]

[0118] Use the following formula to make the interval indicator positive:

[0119]

[0120]

[0121] S23. Based on the obtained standardized evaluation matrix, the advantages and disadvantages solution vectors of the UAV operation data under each working condition are calculated.

[0122] In this embodiment, the ideal good and bad solutions of the UAV operating parameters are calculated by the standardized matrix under each working condition.+ 、s - :

[0123] s + =[max{y 11 ,y 21 ,…,y n1},max{y 12 ,y 22 ,…,y n2},…,max{y 1m ,y 2m ,…,y nm}](7)

[0124] s - =[min{y 11 ,y 21 ,…,y n1},min{y 12 ,y 22 ,…,y n2},…,min{y 1m ,y 2m ,…,y nm}] (8)

[0125] Among them, x best is the optimal ideal value of the intermediate indicator, a, b are the optimal interval boundaries [a, b] of the interval indicator;

[0126] S3, calculating the good and bad solution vector when inputting real-time parameters, comparing it with the good and bad solution vector in the model, and updating the good and bad solution vector;

[0127] In this embodiment, the data to be evaluated is input into the health evaluation matrix X under the corresponding working condition according to its working condition category. (n+1)×m After forward processing and dimensionless processing, a new standardized matrix is ​​obtained and its good and bad solution vector is calculated. After comparing with the good and bad solution constructed by the model, the good and bad solution vector s' is updated. + 、s' - :

[0128] s *+ =max{s + ,s' +} (9)

[0129] s *- =min{s - ,s' _} (10)

[0130] S4. Introduce the combined weighting method of variable weight method and entropy weight method to calculate the weight of each indicator, and combine the combined weighting method to calculate the distance value d between the flight data of the drone to be evaluated and the updated ideal good and bad solution vector + d- , and then get the value of the health status index h:

[0131]

[0132]

[0133] Among them, ω j , j = 1, 2, ..., m + 1, is the jth element of the weight vector ω of the evaluation index,

[0134]

[0135] like Figure 4 As shown, in this embodiment, step S4 of calculating the weight vector of the health assessment parameter of the drone under multiple working conditions also includes the following specific steps:

[0136] S41, establish a pairwise evaluation matrix A of evaluation indicators, and obtain a constant weight vector w1 by the hierarchical analysis method;

[0137] S42, construct a state variable weight vector G, obtain a variable weight matrix W, and take the last line of the variable weight matrix as the variable weight vector w2 of the input parameter;

[0138] In this embodiment, a state variable weight vector is constructed, an exponential state variable weight vector is selected, and a state variable weight vector G(Y i )=[G1(Y i ),G2(Y i ),…,G n (Y i ),G n+1 (Y i )]as follows:

[0139]

[0140] In the formula, i = 1, 2, ..., n + 1; j = 1, 2, ..., m; α> 0 is the penalty level factor; 0 < β ≤ 1; the variable weight vector G is obtained (n+1)×m , and then calculate the variable weight matrix W according to formula (16), and use the last row of the variable weight matrix as the current state weight w2 of the real-time input data;

[0141]

[0142] S43, calculating the probability matrix P according to the matrix Y after the forward and normalization processing;

[0143]

[0144] S44. Calculate the information entropy e of each indicator j And calculate the entropy weight method weight w3;

[0145] In this embodiment, the information entropy e of each indicator is calculated using the following formula: j ;

[0146]

[0147] Use the following formula to calculate the entropy weight vector w3;

[0148]

[0149] S45. Obtain the final weight vector w according to the combined weighting method.

[0150] In this embodiment, according to the combined weighting method, the final weight vector w is obtained using the following formula:

[0151] w=λw2+(1-λ)w3 (19)

[0152] Among them, λ is the weight under variable weight empowerment.

[0153] S5. Then calculate the distance between the real-time operating parameters reported by the drone and the updated good and bad solution vector, calculate the health index h according to the obtained distance, divide the health status level and construct the membership function of the obtained health index h to the health status, and determine the health status of the drone according to the maximum membership principle.

[0154] In this embodiment, the health status is divided into levels, including but not limited to: healthy, sub-healthy, degraded, deteriorated and faulty, and a membership function of the obtained health indicator h to the health status is constructed. The membership function that conforms to the Gaussian distribution is used as the health status membership function of the health indicator, and the health status of the drone is determined according to the maximum membership principle.

[0155]

[0156] Where c is the mean of the curve and σ is the standard deviation, which controls the width of the curve.

[0157] In this embodiment, a membership function conforming to Gaussian distribution is used as the health status membership function of the health indicator, and the health status of the drone is determined according to the maximum membership principle. Formulas (21) to (25) are the health status membership functions of the health indicator h for the five status levels of health, sub-health, degradation, deterioration, and failure:

[0158]

[0159]

[0160]

[0161]

[0162]

[0163] In summary, the present invention evaluates the health status of a quadcopter UAV for substation inspection. It takes into account the differences in indicator types under multiple working conditions of UAV flight missions, analyzes parameter characteristics under multiple working conditions, calculates the superior and inferior solutions under each working condition, and takes into account the imbalance of positive and negative samples in flight data. It adopts a method of updating the superior and inferior solutions, and uses a combined weighting method that combines variable weight theory with the entropy weight method to highlight the impact of individual indicator abnormalities on the evaluation results, weakening the impact of subjective weighting. Finally, the ideal solution distance is used as a measure of the health indicator, and a membership function is established to determine the final health status of the UAV.

[0164] The present invention divides the flight mission of the quad-rotor UAV into multiple working conditions in detail according to the characteristics of the flight mission, and analyzes the parameter characteristics of the evaluation index under different working conditions according to the characteristics of each working condition, so as to know and evaluate the health status of the UAV in a more refined way.

[0165] The present invention establishes working condition clustering for the original flight data of the quad-rotor drone for the data set under various working conditions. Taking into account that the selection of the initial center point will have a great influence on the iteration efficiency and iteration effect, the original arbitrary selection of the centroid equal to the number of working conditions as the initial centroid is abandoned. A centroid is first selected to calculate its distance with the sample point, and the second centroid is selected based on this, and the selections are made in sequence until the centroid equal to the number of working conditions is selected for subsequent calculations, which greatly improves the iteration efficiency.

[0166] The present invention calculates health indicators based on ideal superior and inferior solutions, makes full use of historical UAV flight data, and considers that the ratio of positive and negative samples in UAV flight data differs greatly. The method of updating the ideal solution vector is adopted to improve the ideal solution vector in real time, making the evaluation result more reliable.

[0167] Aiming at the situation that the constant weight cannot well reflect the abnormality of individual indicators, the present invention adopts the combination weighting method of hierarchical analysis method combined with variable weight theory and entropy weight method and the superior and inferior solution distance method to make the evaluation result more accurate and reliable.

[0168] The present invention solves the technical problem in the prior art that the imbalance characteristics of positive and negative samples and the abnormality of individual indicators under multiple working conditions are not fully considered, resulting in a high degree of subjectivity in weighting and restricting the accuracy of drone health assessment.

[0169] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the health status of a substation inspection quadrotor drone, characterized in that: The method comprises: S1. During the process of the quad-rotor UAV performing the inspection flight mission, the operating condition division operation is performed, the operating parameter data of the quad-rotor UAV under each operating condition is collected, and a UAV health status assessment data set is constructed for the operating condition; S2. Analyze the characteristic information of the operating parameters of the quad-rotor drone under each of the working conditions, perform forward processing respectively, and obtain the standardized matrix under each of the working conditions to obtain the model optimal solution vector and the model worst solution vector of the drone operating parameters; S3, calculating the optimal solution vector and the worst solution vector when the real-time parameters are input, comparing them with the model optimal solution vector and the model worst solution vector, and updating to obtain an ideal good and bad solution vector; S4. Combining the variable weight method and the entropy weight method, a combined weighting method is designed, and the weight of each indicator is calculated using the weighting method to obtain the distance value d between the flight data of the drone to be evaluated and the ideal good and bad solution vector. + d - , and process to obtain the health status index value h: S5. Calculate the distance between the real-time operating parameters reported by the quadrotor drone and the ideal good-bad solution vector, calculate the health index h according to the distance, divide the health status level, construct the membership function of the health index h to the health status, and determine the health status of the drone according to the maximum membership principle.

2. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: The S1 includes: S11, the collected flight data sample set {x1, x2, ..., x m } as input, determine the number of clusters k, C = {C1, C2, ..., C k } is a set of clusters, and sets the maximum number of iterations p; S12: From the sample set in the drone health status assessment data set, determine k samples as the initial centroid μ i , i=1,2,…,k; Among them, a data point set is obtained and input from the drone health status assessment data set, and a point is randomly selected from the data point set as the first cluster center μ1, and each point x is calculated i The distance from the nearest cluster center among the selected cluster centers is selected, and a point whose D(x) is greater than the preset threshold is selected as the new cluster center: D(x i )=arg min||x i -μ j || (1) In the formula, μ j is the currently selected centroid vector, arg min represents the variable value when the function reaches the minimum value; S13, repeatedly executing S12 until k centroids are selected; S14. Calculate sample x i With the center of mass μ j The Euclidean distance between the samples x i Mark the cluster label value corresponding to the centroid with the smallest distance, and update to obtain k cluster sets; S15. For each sample point in any cluster, use the following logic to recalculate the new centroid: S16. Repeat S13 and S14 until it is determined that all the centers of mass no longer change, and obtain a final UAV working condition classification result C.

3. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: The S2 includes: S21, analyzing the parameter index types of the selected evaluation parameters under different working conditions, taking the collected parameters x i =(x i1 ,x i2 ,…,x im ) constitutes the health evaluation matrix X n×m =(x1,x2,…,x n ) T ; S22, for the evaluation matrix under each of the working conditions, according to the parameter index type, performing the corresponding forward processing to obtain a forward matrix, and performing dimensionless processing on the forward matrix to obtain a standardized evaluation matrix; S23, based on the standardized evaluation matrix, calculating the optimal solution vector and the worst solution vector s of the model of the operating data of the quadrotor drone under each of the working conditions + 、s - : s + =[max{and 11 ,and 21 ,…,and n1 },max{and 12 ,and 22 ,…,and n2 },…,max{and 1m ,and 2m ,…,and nm }](7) s - =[min{and 11 ,and 21 ,…,and n1 },min{and 12 ,and 22 ,…,and n2 },…,min{and 1m ,and 2m ,…,and nm }] (8) In the formula, x best is the optimal ideal value of the intermediate indicator, and a, b are the optimal interval boundaries [a, b] of the interval indicator.

4. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 3, characterized in that: The S22 further includes: The flight data of the quad-rotor drone under each of the working conditions is subjected to the forward processing and the dimensionless processing to obtain the standardized matrix Y n×m ; Wherein, the following logic is used to perform the positive processing on the extremely small indicators: The intermediate indicators are positively processed using the following logic: The following logic is used to perform the positive processing on the interval-type indicators:

5. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: In S3, the data to be evaluated is input into the health evaluation matrix X under the corresponding working condition according to the corresponding working condition category. (n+1)×m , perform forward processing and dimensionless processing to obtain a new standardized matrix to obtain the ideal good and bad solution vector s' + 、s' - : s *+ =max{s + ,s' + } (9) s *- =min{s - ,s' _ } (10).

6. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: In S4, the distance value d is processed using the following logic: + d - : Among them, ω j , j = 1, 2, ..., m + 1, is the jth element of the weight vector ω of the evaluation index, 7. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: The S4 further comprises: S41, establish a pairwise evaluation matrix A of evaluation indicators, and obtain a constant weight vector w1 by the hierarchical analysis method; S42, constructing a state variable weight vector G, processing to obtain a variable weight matrix W, taking the last line of the variable weight matrix as the variable weight vector w2 of the input parameter; wherein the state variable weight vector is constructed using the following logic: G(Y i ) = [G1(Y i ), G2(Y i ), …, G n (Y i ), G n+1 (Y i )] is as follows: In the formula, i = 1, 2, ..., n + 1; j = 1, 2, ..., m; α> 0 is the penalty level factor; 0 < β ≤ 1; the variable weight vector G is obtained (n+1)×m ; The variable weight matrix W is calculated using the following logic, and the variable weight vector w2 of the last row of the variable weight matrix is ​​used as the real-time input data as the current state weight; S43, the standardized matrix Y n×m Perform forward and standardization processing to obtain the probability matrix P; S44. Calculate the information entropy e of each indicator according to the probability matrix P. j And calculate the entropy weight method weight w3; S45, obtaining a final weight vector w according to a combined weighting method, wherein the weight vector w is obtained by using the following logical processing according to the combined weighting method; w=λw2+(1-λ)w3 (19) In the formula, λ is the weight under variable weight empowerment.

8. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 7, characterized in that: In S44, the information entropy e of each indicator is calculated using the following logic: j ; The entropy weight method weight w3 is calculated using the following logic; 9. A method for evaluating the health status of a substation inspection quad-rotor drone according to claim 1, characterized in that: In S5, the membership function conforming to the Gaussian distribution is used as the health status membership function of the health indicator, and the health status of the drone is determined according to the maximum membership principle and using the following logic. Where c is the mean of the curve and σ is the standard deviation.

10. A health status assessment system for substation inspection quadcopter drone, characterized in that: The system comprises: A data set construction module is used to perform working condition division operations on the process of the quad-rotor UAV performing the inspection flight mission, collect the operating parameter data of the quad-rotor UAV under each working condition, and construct a UAV health status assessment data set for the working condition; A standardized matrix construction module is used to analyze the characteristic information of the operating parameters of the quad-rotor drone under each of the working conditions, perform forward processing respectively, and obtain the standardized matrix under each of the working conditions to obtain the model optimal solution vector and the model worst solution vector of the drone operating parameters. The standardized matrix construction module is connected to the data set construction module; A good and bad solution vector updating module, used for calculating the optimal solution vector and the worst solution vector when inputting real-time parameters, comparing them with the model optimal solution vector and the model worst solution vector, and updating to obtain an ideal good and bad solution vector, wherein the good and bad solution vector updating module is connected to the standardized matrix building module; The combined weighting module is used to combine the variable weight method and the entropy weight method to design a combined weighting method, and use the weighting method to calculate the weight of each indicator to obtain the distance value d between the flight data of the drone to be evaluated and the ideal good and bad solution vector. + ,d - , processing to obtain the health status index value h, the combined weighting module is connected with the superior and inferior solution vector updating module: The drone health status determination module is used to calculate the real-time operating parameters reported by the quad-rotor drone and the distance from the ideal good and bad solution vector, calculate the health index h according to the distance, divide the health status level, construct the membership function of the health index h to the health status, and determine the drone health status according to the maximum membership principle. The drone health status determination module is connected to the combined weighting module.

Images