Distribution network automation equipment group fault diagnosis method, system, equipment and medium

Abstract

The invention discloses a distribution network automation equipment group fault diagnosis method and system, equipment and a medium, and relates to the technical field of distribution network automation, and the method comprises the steps: obtaining fault data of a distribution network automation equipment group, and carrying out the fault feature extraction; inputting the fault features into the Cauchy distribution probability neural network to obtain an output result with scale parameters, performing parameter optimization by adopting a fractional order war strategy optimization algorithm, and updating the optimized scale parameters to obtain a trained Cauchy distribution probability neural network; and inputting the real-time fault data of the distribution network automation equipment group into the optimal Cauchy distribution probabilistic neural network to obtain various faults. According to the method, data with extreme values or abnormal values are processed through Cauchy distribution, output with scale parameters is better obtained, the convergence speed of the algorithm can be increased through adjustment of fractional calculus, the algorithm can be helped to jump out of a local optimal solution, and therefore the quality of an output result is improved, and the diagnosis effect is improved.

Description

Technical Field

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

[0002] The distribution network is a key link in the power grid system. It directly serves the vast number of electricity users, constitutes a bridge between energy production and consumption, and is an indispensable part of the power grid. With the optimization and upgrading of the energy structure and the continuous deepening of the power system reform, the distribution network is ushering in new development opportunities. It is necessary to pursue the maximization of efficiency and benefits while ensuring safety and quality, and gradually evolve into a safe, reliable, green, environmentally friendly, interactive, economical and efficient smart distribution network. As the cornerstone of smart grid construction, distribution automation plays an irreplaceable role in monitoring the operating status of the distribution network and improving power supply reliability.

[0003] There are a large number of terminal devices in the distribution automation system, and the communication lines involved are intricate. Especially in the joint debugging of a large number of distribution automation terminal devices connected to the master station system, close cooperation between the master station and the site is required. In indoor and outdoor switchgear, there are many types of equipment, and the communication lines of the distribution automation system are particularly complex, which requires the master station and the site to cooperate in real time. According to different distribution scenarios and operation and maintenance requirements, fault detection of equipment groups such as switch cabinets and distribution automation terminals is a key issue that needs to be solved in the distribution detection process. At present, fault detection of distribution network automation equipment groups mainly relies on manual inspections, which consumes a lot of manpower and time, resulting in high human resource costs, unpredictable power outage time, and difficulty in cross-departmental coordination, all of which directly affect the reliability of power supply.

[0004] In the existing technical solutions, the fault signals of the distribution network automation equipment group are generally dominated by current, voltage and other signals. Traditional time domain fault feature extraction methods, such as mean, root mean square, kurtosis and skewness, and frequency domain fault feature extraction methods, such as power spectrum, envelope spectrum analysis and fast Fourier transform, etc. The fault current and voltage data of the distribution network automation equipment group is a kind of nonlinear fault feature data. The traditional time domain fault feature extraction method is difficult to analyze such nonlinear signals. In addition, due to the nesting of hidden layer parameters in the existing fault feature extraction network, its loss function is a strongly nonlinear non-convex function, which is extremely difficult to globally optimize. Generally, only the optimization method that relies on gradient information can be used to solve its local suboptimal solution, which affects the network learning effect and leads to inaccurate fault identification and speed difference when diagnosing equipment faults. Summary of the invention

[0005] The purpose of the present invention is to provide a distribution network automation equipment group fault diagnosis method, system, equipment and medium to solve the problems in the prior art in view of the above-mentioned deficiencies in the prior art.

[0006] The present invention specifically provides the following technical solutions:

[0007] A method for diagnosing faults of a distribution network automation equipment group comprises the following steps:

[0008] Obtain the fault data of the distribution network automation equipment group, and use fractional attention entropy to extract fault features from the fault data;

[0009] The fault feature is input into the Cauchy distribution probability neural network to obtain an output result with a scale parameter, and the attack strategy or the defense strategy in the fractional order war strategy optimization algorithm is used to optimize the scale parameter, and the Cauchy distribution probability neural network is updated by the optimized scale parameter to obtain a trained Cauchy distribution probability neural network;

[0010] The real-time fault data of the distribution network automation equipment group is input into the trained Cauchy distribution probability neural network to obtain the optimal output result, and the optimal output result is output as the fault category to perform fault diagnosis on the distribution network automation equipment group.

[0011] Preferably, the extracting fault features from fault data using fractional attention entropy includes:

[0012] Each point in the fault data time series is regarded as a state, the time series is represented by a peak point sequence, and the interval between two consecutive peak points is obtained;

[0013] Get the Shannon entropy of the frequency interval between adjacent peak points, where the fractional Shannon entropy FrH of the jth strategy j The specific expression of (X) is:

[0014]

[0015] Where: p(x i ) is the position parameter x when the i-th strategy i The probability of occurrence at ; b is the number of interval point types, where Γ(·) is the gamma function, ψ(·) is the digamma function, and α is the fractional order;

[0016] The mean of the fractional Shannon entropy obtained by various different strategies is defined as the fractional attention entropy. The specific expression is:

[0017]

[0018] Fault feature extraction is performed on fault data via fractional attention entropy.

[0019] Preferably, the fault feature is input into a Cauchy distribution probability neural network for training to obtain an output result with a scale parameter, including:

[0020] Get the probability function f(·) of the Cauchy distribution. The specific expression is:

[0021]

[0022] Among them, x is the location parameter variable, x0 is the location parameter that defines the peak position of the distribution; γ is the scale parameter of the half width at half the maximum value;

[0023] The fault features are input into the probabilistic neural network, and the training samples of the fault features are processed and output using the Cauchy distribution probability density function represented by the Euclidean distance to obtain the output results with scale parameters, and the output results are added by category, and the addition results are judged to obtain the output with the maximum probability value;

[0024] Among them, the Cauchy distribution probability density function Y represented by the Euclidean distance ik The specific expression is:

[0025]

[0026] Add the output by class g i The specific expression of (X) is:

[0027]

[0028] Get the output with the maximum probability value The specific expression is:

[0029]

[0030] Where X is the random variable of the location parameter, X ik is a random variable with the location parameter that defines the location of the peak of the distribution, n i is the n categories for the i-th strategy, and k is the k-th category.

[0031] Preferably, after obtaining the output with the maximum probability value, the method further comprises:

[0032] The root mean square error formula is selected to obtain the error, the Cauchy distribution probability neural network is optimized according to the error, and the output result with scale parameters is obtained again through the optimized Cauchy distribution probability neural network; the specific expression of error acquisition is:

[0033]

[0034] Among them, y j is the true value corresponding to the j-th group of test data, is the predicted value of the jth group of test data, E is the objective function, that is, the error, and m is the amount of test data.

[0035] Preferably, the method of optimizing the scale parameter by using the attack strategy or the defense strategy in the fractional-order war strategy optimization algorithm includes:

[0036] In the attack strategy, the position of each soldier is updated according to the position of the king and the commander. The new position of each soldier is the latest result of the scale parameter. The king is the soldier with the highest attack power, and the commander is the soldier with the second highest attack power. The specific expression of the new position of the soldier is:

[0037] X i (t+1)-X i (t) = 2 × ρ × (CK) + rand × (W i ×KX i (t));

[0038] Among them, X i (t+1) is the soldier's new position, X i is the position before the soldier, C is the commander's position, K is the king's position, W is the i is the weight, t is the variable, rand represents the random number, and ρ is the random constant;

[0039] The Grünwald-Letnikov type fractional derivative is used to update the position of each soldier. The specific expression is:

[0040]

[0041] The position of each search agent is obtained by using the position of the king, the commander, and the rank of each soldier to perform parameter optimization of the scale parameter. The specific expression is:

[0042] X i (t+1)=(X i (t+1))×(F n ≥F p )+(X i (t))×(F n <F p );

[0043] Among them, F n is the new position, F p For the previous location, the search agent is the soldier;

[0044] In the defense strategy, the search agent position is updated based on the positions of the king, the army leader, and a random soldier to obtain the parameter optimization result of the scale parameter. The specific expression is:

[0045]

[0046] Among them, the rank and weight updates in the defense strategy remain unchanged, X rand (t) is a random position; for each parameter optimization process, the weak soldier is relocated to the median of the entire army on the battlefield. The specific expression is:

[0047] X w (t+1)=-(1-randn)×(X w (t)-median(X))+K;

[0048] Among them, randn is a function that generates random numbers or matrices with standard normal distribution, median is the median function, weak soldiers represent soldiers whose fitness is lower than the threshold, and X w (t+1) Reposition the weak soldier.

[0049] Preferably, each parameter optimization process further includes:

[0050] Random soldiers replace weak soldiers. The specific expression is:

[0051] X w (t+1)=Lb+rand×(Ub-Lb);

[0052] Among them, Lb is the lower limit of the exploration space, Ub is the upper limit of the exploration space, and rand is a random soldier.

[0053] The present invention provides a distribution network automation equipment group fault diagnosis system, comprising:

[0054] The acquisition module is used to obtain the fault data of the distribution network automation equipment group and extract the fault features of the fault data using fractional attention entropy;

[0055] A neural network optimization module is used to input the fault characteristics into a Cauchy distribution probability neural network to obtain an output result with a scale parameter, and use an attack strategy or a defense strategy in a fractional-order war strategy optimization algorithm to optimize the scale parameter, and update the Cauchy distribution probability neural network through the optimized scale parameter to obtain a trained Cauchy distribution probability neural network;

[0056] The diagnosis module is used to input the real-time fault data of the distribution network automation equipment group into the trained Cauchy distribution probability neural network to obtain the optimal output result, and output the optimal output result as the fault category to perform fault diagnosis on the distribution network automation equipment group.

[0057] The present invention provides a computer device, comprising a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of the above-mentioned distribution network automation equipment group fault diagnosis method.

[0058] The present invention provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for diagnosing faults of a distribution network automation equipment group are implemented.

[0059] Compared with the prior art, the present invention has the following significant advantages:

[0060] The present invention uses fractional attention entropy to extract fault features from fault data without decomposing or transforming the data. It directly measures the complexity of the data and can extract more comprehensive fault features from the fault data. The Cauchy distribution probability neural network is trained through the fault features. The data with extreme values ​​or abnormal values ​​are processed through the Cauchy distribution to better obtain the output with scale parameters. Since the loss function in the traditional network is a non-smooth and non-convex loss function, parameter optimization is extremely difficult. The present invention uses a derivative-free optimized fractional war strategy optimization algorithm to optimize the parameters of the Cauchy distribution probability neural network. The convergence speed of the algorithm can be accelerated through the adjustment of fractional calculus. At the same time, global and local searches can be performed to help the algorithm jump out of the local optimal solution, thereby improving the quality of the output results, thereby improving the accuracy and diagnosis speed of the equipment fault diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is an overall flow chart of a method for diagnosing faults in a distribution network automation equipment group according to the present invention;

[0062] Figure 2 This is the CPNN network structure diagram in the present invention. DETAILED DESCRIPTION

[0063] The following is a clear and complete description of the technical solutions of the embodiments of the present invention in conjunction with the drawings in 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 should fall within the scope of protection of the present invention.

[0064] Through artificial intelligence technology, the distribution network automation equipment group can more accurately perceive and analyze the working status and health of the distribution equipment, thereby achieving more efficient and accurate data collection and analysis, timely discovering potential faults and abnormal conditions, and improving the accuracy of fault diagnosis and the efficiency of detection and operation and maintenance. Using a data-driven approach combined with artificial intelligence technology, the data of the distribution network automation equipment group is analyzed and processed. By establishing AI intelligent diagnosis and analysis algorithms and quantitative models, real-time monitoring, fault diagnosis and prediction of equipment operating status can be achieved. This can improve the operation and maintenance personnel's understanding of the equipment status so that they can quickly discover and solve problems, and also provide a basis for intelligent operation and maintenance. Table 1 shows common fault categories and their causes.

[0065] Table 1 Common fault categories and their causes

[0066]

[0067] The main idea of ​​the present invention is to use fractional-order attention entropy to extract comprehensive fault features from the fault data of distribution network automation equipment group, and then input the obtained fault features into the Cauchy distribution probability neural network for fault multi-classification processing, in which the algorithm optimizer adopts the fractional-order war strategy optimization algorithm. Figure 1 shown.

[0068] like Figure 1 As shown, the present invention provides a distribution network automation equipment group fault diagnosis method to be described below, which specifically includes the following steps:

[0069] Step S1: Obtain the fault data of the distribution network automation equipment group, use fractional attention entropy to extract fault features from the fault data, and directly measure the complexity of the fault data.

[0070] The present invention uses fractional attention entropy to extract fault features from fault data, without decomposing or transforming the data, and directly measures the complexity of the data. This overcomes the shortcomings of traditional time and frequency domain fault feature extraction methods that often cannot effectively extract hidden fault features. In addition, this parameter-free measure can accurately estimate in a computationally efficient manner, and further considers fractional order information, and can measure the dynamic changes of fractional order time series, so that more comprehensive fault features can be extracted from fault data.

[0071] Compared with traditional entropy, attention entropy has the advantages of strong robustness to time series length and no need to set hyperparameters. The calculation of time series attention entropy can be summarized as follows:

[0072] (1) If each point in the time series is considered as a system, the change of its state can be regarded as the adjustment of the system to the environment. The peak point can effectively characterize the change of the upper and lower bounds of the local state, so the local peak point is defined as the key point.

[0073] (2) Given a finite sequence X, first define the critical sample Ω. Then, calculate the subsequence u of any given interval X. i 、u k and u j , satisfying u i and u j matches in the example Ω, but for any i<k<j, u k There is no match in Ω.

[0074] Define a point x if one of the following conditions is met i is the peak point, including the local maximum and local minimum:

[0075] x i-1 <x i And x i >x i+1 (x i is defined as a local maximum).

[0076] x i <x i-1 And x i <x i+1 (x i is defined as a local minimum).

[0077] If each point in the fault data time series is regarded as a state of the system, then the change of state can be regarded as the adjustment of the system to the environment. Complex systems are expected to have complex state change processes when adapting to the environment. Peak points represent the local upper and lower limits of state changes, which makes them potential key examples. Then, the time series is represented by the peak point sequence, and the interval between two consecutive peak points is calculated.

[0078] According to four different strategies, the interval from local maximum to local maximum (Max-Max), the interval from local minimum to local minimum (Min-Min), the interval from local maximum to local minimum (Max-Min), and the interval from local minimum to local maximum (Min-Max) are set as key points, and the number of interval points between adjacent key points is calculated, and the frequency of all intervals is calculated.

[0079] (3) Obtain the Shannon entropy of the interval frequency between adjacent peak points. The following is the fractional Shannon entropy formula for the jth strategy:

[0080]

[0081] Where: p(x i ) is the position parameter x when the i-th strategy iThe probability of occurrence at the point; b is the number of interval points, where Γ(·) is the gamma function, ψ(·) is the digamma function, and α is the fractional order.

[0082] (4) The mean of the fractional Shannon entropy calculated by the four different strategies is defined as the fractional attention entropy FrATE(X), that is:

[0083]

[0084] Step S2: Input the fault characteristics into the Cauchy distribution probability neural network to obtain an output result with a scale parameter, and use the attack strategy or defense strategy in the fractional-order war strategy optimization algorithm to optimize the scale parameter, and update the Cauchy distribution probability neural network with the optimized scale parameter to obtain a trained Cauchy distribution probability neural network.

[0085] The fault features are input into the Cauchy distribution probability neural network for training, and the output results with scale parameters are obtained, including:

[0086] The Cauchy distribution is also called the Cauchy-Lorentz distribution. It is a continuous probability distribution. The probability function f(·) of the Cauchy distribution is obtained as follows:

[0087]

[0088] The present invention uses L1 norm to characterize its density function:

[0089]

[0090] Among them, x is the location parameter variable, x0 is the location parameter that defines the peak position of the distribution; γ is the scale parameter of the half width at half the maximum value.

[0091] The fault features are input into the probabilistic neural network, and the Cauchy distribution probability density function represented by the Euclidean distance is used to process and output the training samples of the fault features to obtain the output results with scale parameters, and the output results are added by category, and the addition results are judged to obtain the output with the maximum probability value. Specifically:

[0092] CPNN has the characteristics of simple training, fast convergence, and no local minimum problem, which is suitable for real-time processing and nonlinear approximation. CPNN contains four network layers: input layer, sample layer, summation layer, and competition layer.

[0093] Input layer: The number of neurons is the dimension of the fault feature vector. In the input layer, the network will calculate the distance between the input vector and all training sample vectors. Assume X = (x1, x2, ..., x n ) is the input fault feature data, and C is the number of categories.

[0094] Sample layer: The number of neurons is the number of training samples. The activation function of the sample layer is the Cauchy distribution probability density function represented by the Euclidean distance. The Cauchy distribution probability density function Y ik The specific expression is:

[0095]

[0096] Where X ik is the kth training sample of the i-th category, n i is the total number of training samples of the i-th class, γ is the scale parameter, and Y ik It is the output of the k-th fault feature data of the i-th category after the activation function, that is, the output of the sample layer.

[0097] Sum layer: The number of neurons is the number of categories. The output of the sample layer is summed by category g i (X), which is equivalent to C adders, and the specific expression is:

[0098]

[0099] Among them, g i (·) is the adder of the ith type.

[0100] Competition layer: The number of neurons is 1. The judgment result is output by the competition layer. There is only one 1 in the output result, and the rest are 0. The output result with the largest probability value is 1. The output with the largest probability value is obtained. The specific expression is:

[0101]

[0102] Where X is the random variable of the location parameter, X ik is a random variable with the location parameter that defines the location of the peak of the distribution, n i is the n categories for the i-th strategy, and k is the k-th category.

[0103] In this method, the scale parameter γ is an important parameter that affects the results of the algorithm. It is necessary to construct an objective optimization function and perform parameter optimization for this parameter.

[0104] The root mean square error formula is selected for error calculation, and the Cauchy distribution probability neural network is optimized according to the error. The output result with scale parameters is obtained again through the optimized Cauchy distribution probability neural network; let the true value corresponding to the jth group of test data be y j , the predicted value is Then the objective function is the error E, and the specific expression for error acquisition is:

[0105]

[0106] Among them, y j is the true value corresponding to the j-th group of test data, is the predicted value of the jth group of test data, E is the objective function, that is, the error, and m is the amount of test data.

[0107] In order to facilitate the subsequent swarm intelligence method to perform parameter optimization, the objective function E is changed into a fitness function. The objective function E is the minimum problem, and the adaptive function is f(γ).

[0108]

[0109] Its essence is equivalent to a strongly non-smooth and non-convex optimization problem.

[0110] Taking the 3-class classification problem with 2 training samples in each class as an example, the network structure Figure 2 .

[0111] In the parameter optimization of the scale parameters using the attack strategy or defense strategy in the fractional order war strategy optimization algorithm, in order to ensure that their army can reach the predetermined goal and achieve the goal, the emperor and the commander of each unit will coordinate their troops in a specific mode. War strategies are formulated based on mission objectives, threats, difficulties and prospects. War strategy is a continuous dynamic process in which the armed forces simply coordinate and fight against the enemy. As the war progresses, this strategy can adapt to changing conditions. The position of the king and commander has a continuous impact on the position of the soldiers in the army. The flags on the chariots of the king and the army commander represent their positions, which are visible to all soldiers. Soldiers in the team are trained to follow the strategy according to the sound of drums or other instruments. When a military commander dies, the strategy changes, and each other commander must learn how to rebuild and continue to build war strategies. The goal of the king is to conquer the enemy king or leader, while the main goal of the army soldiers is to attack the other side and improve their ranks.

[0112] 1. Attack strategy:

[0113] Two war strategies are simulated. In the first case, the position of each soldier is updated according to the positions of the king and the commander, and each soldier's new position is taken as the latest result of the scale parameter. The king assumes a favorable position to launch a large-scale attack on the opponent. Therefore, the soldier with the greatest attack power or fitness is regarded as the king, and the commander is the soldier with the second highest attack power after the king. All soldiers have the same rank and weight at the beginning of the war. If a soldier successfully executes the strategy, his rank will be promoted. However, as the war progresses, the ranks and weights of all soldiers will be updated according to the success of the strategy. As the war approaches the end, the positions of the king, the army commander, and the soldiers remain very close as they approach the goal. The specific expression of the new position of the soldier is:

[0114] X i (t+1)=X i (t)+2×ρ×(CK)+ rand×(W i ×KX i (t)) (10);

[0115] Can be rewritten as:

[0116] X i (t+1)-X i (t)=2×ρ×(CK)+ rand×(W i ×KX i (t)) (11);

[0117] Among them, X i (t+1) is the soldier's new position, X i is the position before the soldier, C is the commander's position, K is the king's position, W is the i is the weight, t is the variable, rand represents the random number, and ρ is the random constant. Figure 1 The colored circle around the pawn in the middle represents the W based on the king's position. i × i (t) is the trajectory point. If W i >1, then W i × i (t) is beyond the king's position, so the pawn's updated position is beyond the commander's position. i <1, then W i × i The position of (t) is between the king's position and the soldier's current position. Compared with the previous case, the soldier's updated position is closer. If W i tends to zero, then the soldier’s updated position is very close to the commander’s position, which represents the final stage of the war.

[0118] In order to increase the historical dependence of the algorithm, the invention will use the discrete form of the Grünwald-Letnikov type fractional order derivative to generalize the algorithm, and use the Grünwald-Letnikov type fractional order derivative to process the update method of each soldier's position, specifically:

[0119] The Grünwald-Letnikov fractional derivative of X(θ) is defined as:

[0120]

[0121] where α∈(0,1],Γ(·) is a gamma function.

[0122] While integer derivatives imply only a finite sequence, fractional derivatives require an infinite number of terms; integer derivatives are "local" operators, whereas fractional derivatives implicitly contain a "memory" of all past events.

[0123] The approximate expression of formula (5) is:

[0124]

[0125] Where T is the sampling period and r is the truncation order.

[0126] The X on the left side of formula (4) i (t+1)-X i (t) is the discrete form of the differential when the order α = 1 (assuming T = 1), so:

[0127] D α [X i (t+1)]=2×ρ×(CK)+rand×(W i ×KX i (t)) (14);

[0128] If the characteristics of fractional calculus are considered, the order of velocity differential can be extended to real number α∈[0,1], which can obtain smoother changes and longer memory effects. Considering the first r=4 term of the differential given by expression (13), expression (14) can be rewritten as:

[0129]

[0130] When r≥4, the degree of approximation does not change much, so it is discarded. Therefore, we have:

[0131]

[0132] Among them, α is the fractional order, X k-1 (t) The position of the k-1th soldier.

[0133] 2. Update of military ranks and weights:

[0134] The position update of each search agent depends on the interaction of the king, the commander's position, and the rank of each soldier. The rank of each soldier depends on his success history on the battlefield, which will subsequently affect the weight factor W i The rank of each soldier reflects how close the soldier (search agent) is to the target (fitness value). The weight W i will change exponentially with the factor τ. If the new position F n The attack power (fitness) is less than the previous position F p of attack power, the soldier will take up his previous position.

[0135] The position of each search agent is obtained by using the position of the king, the commander, and the rank of each soldier to perform parameter optimization of the scale parameter. The specific expression is:

[0136] X i (t+1)=(X i (t+1))×(F n ≥F p )+(X i (t))×(F n <F p )(17);

[0137] Among them, F n is the new position, F p For the previous position.

[0138] If the soldier successfully updates his position, the soldier's rank R i Will be promoted:

[0139] R i =(R i +1)×(F n ≥F p )+(R i )×(F n <F p ) (18);

[0140] Based on military rank, the new weights are calculated as:

[0141]

[0142] 3. Defense strategy:

[0143] The second strategy updates the position based on the position of the king, army leader, and a random soldier. The rank and weight updates remain the same.

[0144] That is, based on the positions of the king, the army leader and a random soldier, the search agent position is updated to obtain the parameter optimization result of the scale parameter. The specific expression is:

[0145]

[0146] Among them, X rand (t) is a random position.

[0147] This war strategy explores more of the search space than the previous one because it involves a random soldier placement. i For larger values ​​of W, the soldiers take larger steps and update their positions. i The smaller the value of , the smaller the steps the soldier takes when updating its position.

[0148] 4. Replacement (repositioning) of weak soldiers:

[0149] For each iteration, identify the weak soldier with the worst fitness. One of the simplest ways is to replace the weak soldier with a random soldier, as follows:

[0150] X w (t+1)=Lb+rand×(Ub-Lb) (21);

[0151] Among them, Lb is the lower limit of the exploration space, Ub is the upper limit of the exploration space, and rand is a random soldier.

[0152] The second method is to reposition the weak soldiers near the median of the entire army on the battlefield for each parameter optimization process, as shown below:

[0153] X w (t+1)=-(1-randn)×(X w (t)-median(X))+K (22);

[0154] Among them, randn is a function that generates random numbers or matrices with standard normal distribution, median is the median function, weak soldiers represent soldiers whose fitness is lower than the threshold, and X w (t+1) Reposition the weak soldier.

[0155] Step S3: input the real-time fault data of the distribution network automation equipment group into the trained Cauchy distribution probability neural network to obtain the optimal output result, and output the optimal output result as the fault category to perform fault diagnosis on the distribution network automation equipment group.

[0156] Based on the above method, the present invention provides a distribution network automation equipment group fault diagnosis system, including: an acquisition module, a neural network optimization module and a diagnosis module.

[0157] Among them, the acquisition module is used to obtain the fault data of the distribution network automation equipment group, and use fractional-order attention entropy to extract fault features from the fault data; the neural network optimization module is used to input the fault features into the Cauchy distribution probability neural network to obtain an output result with scale parameters, and use the attack strategy or defense strategy in the fractional-order war strategy optimization algorithm to optimize the scale parameters, and update the Cauchy distribution probability neural network through the optimized scale parameters to obtain a trained Cauchy distribution probability neural network; the diagnosis module is used to input the real-time fault data of the distribution network automation equipment group into the trained Cauchy distribution probability neural network to obtain the optimal output result, and output the optimal output result as the fault category to perform fault diagnosis on the distribution network automation equipment group.

[0158] The present invention also provides a computer device, including a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a distribution network automation equipment group fault diagnosis method.

[0159] In accordance with the disclosed embodiments, a computing device may communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth communications, etc.), or with any device (e.g., routers, modems, etc.) that enables a computing device to communicate with one or more other computing devices.

[0160] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of a distribution network automation equipment group fault diagnosis method are implemented.

[0161] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, such as but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0162] The above content is a further detailed description of the present invention in combination with a specific preferred embodiment. For technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for diagnosing faults in a distribution network automation equipment group, characterized in that: include: Obtain the fault data of the distribution network automation equipment group, and use fractional attention entropy to extract fault features from the fault data; The fault feature is input into the Cauchy distribution probability neural network to obtain an output result with a scale parameter, and the attack strategy or the defense strategy in the fractional order war strategy optimization algorithm is used to optimize the scale parameter, and the Cauchy distribution probability neural network is updated by the optimized scale parameter to obtain a trained Cauchy distribution probability neural network; The real-time fault data of the distribution network automation equipment group is input into the trained Cauchy distribution probability neural network to obtain the optimal output result, and the optimal output result is output as the fault category to perform fault diagnosis on the distribution network automation equipment group.

2. A distribution network automation equipment group fault diagnosis method as claimed in claim 1, characterized in that: The method of extracting fault features from fault data using fractional attention entropy includes: Each point in the fault data time series is regarded as a state, the time series is represented by a peak point sequence, and the interval between two consecutive peak points is obtained; Get the Shannon entropy of the frequency interval between adjacent peak points, where the fractional Shannon entropy FrH of the jth strategy j The specific expression of (X) is: Where: p(x i ) is the position parameter x when the i-th strategy i The probability of occurrence at ; b is the number of interval point types, where Γ(·) is the gamma function, ψ(·) is the digamma function, and α is the fractional order; The mean of the fractional Shannon entropy obtained by various different strategies is defined as the fractional attention entropy FrATE(X), and the specific expression is: Fault feature extraction is performed on fault data via fractional attention entropy.

3. A distribution network automation equipment group fault diagnosis method as claimed in claim 1, characterized in that: The fault feature is input into the Cauchy distribution probability neural network to obtain an output result with a scale parameter, including: Get the probability function f(·) of the Cauchy distribution. The specific expression is: Among them, x is the location parameter variable, x0 is the location parameter that defines the peak position of the distribution; γ is the scale parameter of the half width at half the maximum value; The fault features are input into the probabilistic neural network, and the training samples of the fault features are processed and output using the Cauchy distribution probability density function represented by the Euclidean distance to obtain the output results with scale parameters, and the output results are added by category, and the addition results are judged to obtain the output with the maximum probability value; Among them, the Cauchy distribution probability density function Y represented by the Euclidean distance ik The specific expression is: Add the output by class g i The specific expression of (X) is: Get the output with the maximum probability value The specific expression is: Where X is the random variable of the location parameter, X ik is a random variable with the location parameter that defines the location of the peak of the distribution, n i is the n categories for the i-th strategy, and k is the k-th category.

4. A distribution network automation equipment group fault diagnosis method as claimed in claim 3, characterized in that: After obtaining the output with the maximum probability value, the method further includes: The root mean square error formula is selected to obtain the error, the Cauchy distribution probability neural network is optimized according to the error, and the output result with scale parameters is obtained again through the optimized Cauchy distribution probability neural network; the specific expression of error acquisition is: Among them, y j is the true value corresponding to the j-th group of test data, is the predicted value of the jth group of test data, E is the objective function, that is, the error, and m is the amount of test data.

5. A distribution network automation equipment group fault diagnosis method as claimed in claim 1, characterized in that: The method of optimizing the scale parameters by using the attack strategy or the defense strategy in the fractional-order war strategy optimization algorithm includes: In the attack strategy, the position of each soldier is updated according to the position of the king and the commander. The new position of each soldier is the latest result of the scale parameter. The king is the soldier with the highest attack power, and the commander is the soldier with the second highest attack power. The specific expression of the new position of the soldier is: X i (t+1)-X i (t)=2×ρ×(C-K)+rand×(W i ×K-X i (t)) Among them, X i (t+1) is the soldier's new position, X i is the position before the soldier, C is the commander's position, K is the king's position, W is the i is the weight, t is the variable, rand represents the random number, and ρ is the random constant; The Grünwald-Letnikov type fractional derivative is used to update the position of each soldier. The specific expression is: Among them, α is the fractional order, X k-1 (t) The position of the k-1th soldier; The position of the king, the commander, and the rank of each soldier are used to obtain the position of each search agent and perform parameter optimization of the scale parameter. The specific expression is: X i (t+1)=(X i (t+1))×(F n ≥F p )+(X i (t))×(F n <F p ); Among them, F n is the new position, F p For the previous location, the search agent is the soldier; In the defense strategy, the search agent position is updated based on the positions of the king, the army leader, and a random soldier to obtain the parameter optimization result of the scale parameter. The specific expression is: Among them, the rank and weight updates in the defense strategy remain unchanged, X rand (t) is a random position; for each parameter optimization process, the weak soldier is relocated to the median of the entire army on the battlefield. The specific expression is: X w (t+1)=-(1-randn)×(X w (t)-median(X))+K; Among them, randn is a function that generates random numbers or matrices with standard normal distribution, median is the median function, weak soldiers represent soldiers whose fitness is lower than the threshold, and X w (t+1) Reposition the weak soldier.

6. A distribution network automation equipment group fault diagnosis method as claimed in claim 5, characterized in that: For each parameter optimization process, it also includes: Random soldiers replace weak soldiers. The specific expression is: X w (t+1)=Lb+rand×(Ub-Lb); Among them, Lb is the lower limit of the exploration space, Ub is the upper limit of the exploration space, and rand is a random soldier.

7. A distribution network automation equipment group fault diagnosis system, characterized in that: include: The acquisition module is used to obtain the fault data of the distribution network automation equipment group and extract the fault features of the fault data using fractional attention entropy; A neural network optimization module is used to input the fault characteristics into a Cauchy distribution probability neural network to obtain an output result with a scale parameter, and use an attack strategy or a defense strategy in a fractional-order war strategy optimization algorithm to optimize the scale parameter, and update the Cauchy distribution probability neural network through the optimized scale parameter to obtain a trained Cauchy distribution probability neural network; The diagnosis module is used to input the real-time fault data of the distribution network automation equipment group into the trained Cauchy distribution probability neural network to obtain the optimal output result, and output the optimal output result as the fault category to perform fault diagnosis on the distribution network automation equipment group.

8. A computer device, characterized in that: It comprises a memory and a processor, wherein a program is stored in the memory, and when the program is executed by the processor, the processor executes the steps of a distribution network automation equipment group fault diagnosis method as claimed in any one of claims 1 to 6.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a distribution network automation equipment group fault diagnosis method described in any one of claims 1 to 6 are implemented.

Images