Photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model

Abstract

The invention relates to a photovoltaic array fault diagnosis method, in particular to a photovoltaic array fault diagnosis method based on a VMD-COMTBO-LSTM model, and provides a COMTBO algorithm fusing an eagle optimization algorithm, chaotic mapping and a Cauchy variation strategy, and the COMTBO algorithm is used for optimizing LSTM and building a photovoltaic array fault diagnosis model of COMTBO-LSTM. And in order to better utilize the fault features of the photovoltaic array, a photovoltaic array model is built in an MATLAB / Simulink software environment, and fault data is obtained. VMD (variational mode decomposition) is used to decompose fault data into a plurality of components, and according to the magnitude of energy, the mode component corresponding to the maximum energy is selected as the COMTBO-LSTM input quantity. And finally, through comparison with diagnosis results of other algorithms, the optimization effect of the improved COMTBO is better than that of other algorithms, and the fault diagnosis rate of the photovoltaic array reaches 98%.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and in particular to a photovoltaic array fault diagnosis method based on a VMD-COMTBO-LSTM model. Background Art

[0002] In recent years, with the depletion of fossil fuels such as oil, the energy crisis has become a common problem faced by countries around the world. Photovoltaic power generation has become an ideal alternative energy source. Photovoltaic arrays are composed of several parts, and photovoltaic panels are the most important component. Since photovoltaic panels need to be exposed to the outside for a long time, they are prone to many failures, resulting in reduced power generation efficiency and even fires.

[0003] In the prior art, photovoltaic fault detection methods can be roughly divided into electrical performance monitoring, thermal imaging technology, electronic parameter monitoring, light intensity sensor, remote monitoring and communication technology, data analysis and machine learning. Electrical performance monitoring mainly monitors the electrical performance parameters of the photovoltaic array, such as current, voltage and power, to identify potential faults. Abnormal electrical performance data may indicate battery short circuit, open circuit, contact problem or other electrical faults; thermal imaging technology is mainly infrared thermal imaging technology that can be used to detect hot spots in photovoltaic modules. Hot spots may be caused by local battery failure, connection problems or local shadows. Thermal imaging can quickly locate and diagnose these problems; electronic parameter monitoring mainly monitors electronic parameters in the photovoltaic array, such as reverse leakage current, insulation resistance, etc., which helps to identify electrical faults. Abnormal changes in these parameters may indicate equipment aging, corrosion or other electronic faults; light intensity sensors are mainly light intensity sensors that can be used to monitor lighting conditions to help identify whether system performance degradation is caused by weather conditions. For example, shadows, cloud penetration and pollution may cause performance degradation of local or overall photovoltaic arrays.

[0004] Remote monitoring and communication technology mainly utilizes remote monitoring systems to remotely access the performance data of photovoltaic arrays in real time. When system performance degrades or a failure occurs, maintenance personnel can be notified remotely so that timely measures can be taken; data analysis and machine learning mainly utilize data analysis and machine learning algorithms to process and analyze large amounts of real-time data to identify normal and abnormal patterns. This includes the use of supervised learning, unsupervised learning, and deep learning methods to build models to improve the accuracy of fault diagnosis.

[0005] However, all of the above-mentioned photovoltaic fault detection methods have defects. Electrical performance monitoring may not be able to accurately distinguish different types of electrical faults. Thermal imaging technology may not be sensitive enough under low light conditions, and in some cases, hot spots may not be obvious. Electronic parameter monitoring may be affected by environmental conditions. Light intensity sensors are mainly used to detect light conditions, but cannot directly diagnose problems inside the photovoltaic array. Remote monitoring and communication technologies rely on the reliable communication of remote monitoring systems. In some remote areas or when the network is unstable, data transmission may be delayed or interrupted, affecting timely fault response. Data analysis and machine learning are based on data analysis and machine learning methods that rely on a large amount of high-quality data to train the model. Therefore, in the case of poor data quality or insufficient data, the accuracy of the model may decrease. Although the model can be improved by using algorithms, there will be problems such as slow convergence and falling into local optimality. The current photovoltaic array fault diagnosis method based on long short-term memory neural network (LSTM) has low fault accuracy. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model. The method integrates the Osprey and Mountain Climbing Optimization Algorithm (COMTBO) to optimize LSTM, and the implementation method is simple, low-cost, and has high fault diagnosis accuracy and efficiency. The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model is stable, thereby improving the diagnosis accuracy.

[0007] The technical solution proposed by the present invention is:

[0008] A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model, the method comprising the following steps:

[0009] A photovoltaic array fault diagnosis model is built based on the LSTM model, and the parameters of the LSTM model are optimized using the Cauchy-Osprey-Mountaineering Team-Based Optimization algorithm that combines Cauchy mutation and Osprey to build a COMTBO-LSTM fault diagnosis model.

[0010] The photovoltaic array fault data under short circuit fault, open circuit fault, ground fault and line fault are used as training data sets, and then the data is decomposed by variational mode decomposition. The decomposed data is used to train the constructed OMTBO-LSTM fault diagnosis model to obtain the trained VMD-COMTBO-LSTM fault diagnosis model.

[0011] Obtain the PV array fault data to be diagnosed and extract characteristic parameters, and perform VMD decomposition on the fault data;

[0012] The decomposed data is input into the trained COMTBO-LSTM fault diagnosis model, and the fault diagnosis result is output.

[0013] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model is a fusion of the mountain climbing optimization algorithm (MTBO) and the osprey optimization algorithm (OOA), and the steps are as follows:

[0014] (1) Using the Circle mapping strategy to initialize the climbers, relative to the randomly distributed climbers;

[0015] (2) The first stage of OOA is used to replace the assisted mountaineering position update formula in the original MTBO;

[0016] (3) Cauchy mutation is used to replace the random member replacement stage.

[0017] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model includes the following steps of training the constructed COMTBO-LSTM fault diagnosis model using a training data set:

[0018] (1) Build a photovoltaic array model in Matlab / Simulink, simulate different faults, and collect photovoltaic array fault data;

[0019] (2) Label the data according to different fault types and establish a fault data set;

[0020] (3) Normalization of fault data;

[0021] (4) Set VMD parameters, perform VMD decomposition on the normalized data, and select the component with the largest energy to fuse with the label;

[0022] (5) Randomly select the test set and training set according to the ratio of test set to training set of 0.8;

[0023] (6) Set the COMTBO-LSTM parameters according to Table 4 and import the training set for training;

[0024] (7) After training, the model performs fault diagnosis on the test set.

[0025] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model, the number of VMD parameter components K.

[0026] The photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model, the initialization COMTBO algorithm parameters include the population number N, the maximum number of iterations Tmax .

[0027] The photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model, wherein the LSTM parameters include the optimal hidden unit 1, the number of L1 and L2, the number of iterations K, the optimal initial learning rate L r .

[0028] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model uses the root mean square error (MSE) between the predicted and actual outputs to calculate the fitness value of the climber according to the following formula:

[0029]

[0030] Where n is the number of test sets, y i is the actual value, is the predicted value.

[0031] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model described above integrates the Cauchy mutation and the Osprey mountain climbing algorithm, and firstly uses the Circle mapping strategy to initialize the mountain climbers:

[0032]

[0033] x i =r i cos(θ i )

[0034] y i =r i sin(θ i )

[0035] In the formula, n is the number of populations, i is the index of an individual, and r is i Represents the random generation radius, a random number uniformly distributed in the interval [0, 1).

[0036] The first stage of OOA is used to replace the assisted climbing position update formula in the original MTBO:

[0037]

[0038] x i,j =lb i,j +r i，j ·(ub j -lb j ), i=1,2,…,N, j=1,2,…,m,

[0039] Using Cauchy mutation to replace the random replacement member stage can expand the search range of MTBO and improve the algorithm's ability to jump out of the local optimal solution; the Cauchy mutation position update formula is as follows:

[0040]

[0041] Where cauchy(0,1) is the standard Cauchy distribution function; is multiplied. One-dimensional Cauchy variogram centered at the origin:

[0042]

[0043] The photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model adopts characteristic parameters including light intensity G, temperature T, short-circuit current I SC , open circuit voltage U oc , Maximum power point P m , Maximum power point voltage U m , Maximum power point current I m, F1, F2, F3, F4, and filling factor FF are used as fault characteristics. The calculation formulas for F1, F2, F3, F4, and filling factor FF are as follows:

[0044]

[0045] Its V oc，STC ,I sc，STC 、V m，STC and I m，STC is the corresponding value of the ideal value condition, and the parameters (α and β) are I sc and V oc Temperature coefficient; V oc,n , I sc,n , V m,n , I m，n At G = 1000w / m 2 , the data record corresponding to the ideal value (STC) at T = 25°C after normalization is calculated according to the following formula:

[0046]

[0047] The photovoltaic array fault diagnosis method based on the VMD-COMTBO-LSTM model comprises:

[0048] Construct an LSTM model to diagnose photovoltaic array faults, use the COMTBO algorithm to optimize the LSTM model parameters, perform VMD decomposition on photovoltaic fault data as training data, and construct a VMD-COMTBO-LSTM fault diagnosis model;

[0049] Data acquisition: Acquisition of photovoltaic fault data under short circuit fault, open circuit fault, line-to-line fault, aging and shading fault to form the initial data set;

[0050] Data processing: VMD data decomposition is performed on the initial data set, and the modal component with the maximum energy is selected and fused with the label as the data set. The test set and training set are randomly selected in a ratio of 4:1;

[0051] Set the COMTBO-LSTM parameters and import the training set for training. After the training is completed, the model performs fault diagnosis on the test set and outputs the fault diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the implementation process of the photovoltaic array fault diagnosis method of the VMD-COMTBO-LSTM model;

[0053] Figure 2 At G = 1000w / m 2 ,The normal, short circuit, open circuit, aging, and line faults are simulated at T = 25 ℃, and the UI image and UP image are obtained;

[0054] Figure 3 It is the basic structure of LSTM neural network;

[0055] Figure 4 It is a 4*8 photovoltaic array simulation diagram constructed for specific applications;

[0056] Figure 5 is the result matrix of the photovoltaic array fault diagnosis method of VMD-COMTBO-LSTM model;

[0057] Figure 6 This is a comparison chart of fault diagnosis accuracy.

[0058] Specific real-time method

[0059] The present invention is further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the protection scope of the present invention is not limited thereby.

[0060] LSTM (Long Short-Term Memory) is a supervised learning model for analyzing sequence data, and is often used for modeling and predicting time series data. It is a special recurrent neural network (RNN) that can effectively capture long-term dependencies and has achieved good performance in many fields. The MTBO algorithm is inspired by the climbing of a mountaineering team, and compares the optimization problem solving process to the climbing of a mountaineering team. In MTBO, each individual represents a climber, the position of the individual represents the location of the climber, and the fitness of the individual represents the height of the climber's position (that is, the objective function value of the problem).

[0061] The COMTBO algorithm combines the MTBO and FES algorithms, and realizes diverse search by introducing the individual position update strategy in MTBO and the osprey predation strategy in FES. Specifically, in each iteration, COMTBO selects the leader individual according to the fitness of the individual, and uses the predation strategy of FES to update the position of the individual. In addition, COMTBO also introduces the Cauchy mutation strategy of the osprey algorithm to increase the diversity of the search.

[0062] The present invention considers the characteristics of multiple types of faults in photovoltaic arrays, adopts an LSTM model to build a fault diagnosis model, introduces an algorithm to optimize the parameters of the LSTM model, optimizes the LSTM by integrating Cauchy mutation and Osprey's mountain climbing optimization algorithm, builds a COMTBO-LSTM model, performs VMD decomposition on fault data, inputs the model into the COMTBO-LSTM model, and finally builds a VMD-COMTBO-LSTM photovoltaic array fault diagnosis model, which can effectively identify the normal, short circuit, open circuit, aging, and line fault states of the photovoltaic array, and improve the accuracy and efficiency of fault diagnosis.

[0063] like Figure 1 As shown, the steps of the photovoltaic array fault diagnosis method of the VMD-CCOMTBO-LSTM model of the present invention include:

[0064] Step 1: Build a photovoltaic array model in Matlab / Simulink, simulate different faults, and collect photovoltaic array fault data;

[0065] Step 2: Label the data according to different fault types and establish a fault data set;

[0066] Step 3: Normalize the fault data;

[0067] Step 4: Set VMD parameters, perform VMD decomposition on the normalized data, and select the component with the largest energy to fuse with the label;

[0068] Step 5: Randomly extract the test set and training set according to the ratio of test set to training set of 0.8;

[0069] Step 6: Set the COMTBO-LSTM parameters according to Table 4 and import the training set for training;

[0070] Step 7: After the training is completed, the model performs fault diagnosis on the test set and outputs the results.

[0071] The selection of input features of the VMD-COMTBO-LSTM photovoltaic fault diagnosis model determines whether the model can quickly and accurately classify the fault type. Before extracting the feature parameters, this embodiment also includes generating the photovoltaic array volt-ampere characteristic curve under different fault types according to the fault data when the photovoltaic array has different fault types, and determining the feature parameters for fault diagnosis according to the photovoltaic array volt-ampere characteristic curve under each fault type. By analyzing the output characteristics of the photovoltaic array in normal and fault states, the fault mechanism characteristics of the photovoltaic array can be accurately determined, so that the various types of faults of the photovoltaic array can be effectively classified using the feature parameters.

[0072] In the selection of specific fault characteristics, when G = 1000w / m 2 ,T=25℃, the normal, short circuit, open circuit, aging, and line fault are simulated, and the UI image and UP image are obtained as follows Figure 2 It can be seen that when a short circuit occurs in the photovoltaic array, the short-circuit current (I sc ) does not change much, but the open circuit voltage (U oc ), significantly smaller, the maximum power point (P m ) also decreases significantly, and the maximum power point current (I m ); when an open circuit occurs, U oc Not much change, I sc and P m becomes smaller; when a shading failure occurs, U oc and I sc The change is not big, but P m When aging failure occurs, U oc and I sc The change is not big, but P m The reduction is large.

[0073] However, the PV array works in a complex and changing environment, which causes the IV curve to be very different from the ideal situation. Therefore, a normalization method is needed to eliminate the influence of different temperatures and irradiances. These four important parameters are extracted from the IV curve of the PV array and divided by their ideal values ​​under standard test conditions (STC). The corresponding data after normalization is recorded as V oc,n ,I sc,n ,Vm,n ,Im ,n .

[0074]

[0075]

[0076] Where G is the irradiance of the photovoltaic panel and T is the temperature of the photovoltaic module. The ideal value of light and temperature is G STC =1000w / m 2 , T STC =25℃. Its V oc，STC ,I sc，STC 、V m，STC and I m，STC is the corresponding value of the ideal value condition, and the parameters (α and β) are I sc and V oc The temperature coefficient of

[0077] Although these four parameters are different under different fault conditions, the differences are still not obvious. The present invention also introduces the following five fault parameters F1, F2, F3, F4, FF as fault characteristics, FF is the filling factor; the formula is as follows:

[0078]

[0079] In summary, the present invention selects G, T, I SC , U oc , P m , U m , F1, F2, F3, F4, and FF are used as fault characteristics.

[0080] In specific applications, by establishing a 4*6 photovoltaic array model such as Figure 4 As shown, the simulation of normal, short circuit, open circuit, aging, and line fault is based on G, T, and I SC , U oc , P m , U m , F1, F2, F3, F4, and FF are used as fault features to obtain founder data as the initial sample data.

[0081] The parameters of the selected photovoltaic modules are shown in Table 1

[0082] Table 1: PV module parameters

[0083]

[0084]

[0085] In the present invention, VMD decomposition decomposes a given vibration signal f into a series of modal components (IMFs) with sparse characteristics, seeks to minimize the sum of the bandwidths of each mode, and constrains the sum of all modes to be equal to the original signal. Each component has its own center frequency and finite bandwidth.

[0086]

[0087] Where k is the number of VMD components, k = 1, 2, ..., K; u k is the kth modal component; H is u k Estimate broadband minimum; ω k for u k The corresponding center frequency.

[0088] By introducing the Lagrange multiplier λ and the penalty factor α, the constrained problem is transformed into an unconstrained problem, and the augmented Lagrange equation is obtained. Then the alternating direction multiplier method is used to solve the equation and find λ, u k and ω k The optimal solution of is as follows:

[0089]

[0090] Update according to the above formula With ω k , according to the given ε (precision), the iteration stops when the following formula is satisfied

[0091]

[0092] Initialize λ before iterative solution 1 , and set the maximum number of iterations.

[0093] Since the COMTBO algorithm is introduced for parameter optimization in the process of LSTM algorithm training fault diagnosis model, the position of each climber is first initialized to consist of the corresponding LSTM parameters. In the process of training the training data set with LSTM, the COMTBO algorithm is used to continuously update, calculate and reorder the fitness value and optimal position of the climber until the optimal solution of the LSTM parameters is found. The advantages of the COMTBO algorithm can be fully utilized to quickly find the optimal solution of the LSTM parameters. It not only has fast convergence speed, high search efficiency and few operating parameters, but also can effectively improve the accuracy of multi-classification fault diagnosis of photovoltaic arrays.

[0094] In the present invention, two LSTM layers are used to construct the LSTM model. The output of each LSTM layer is passed to the next LSTM layer, and finally output to the fully connected layer and the softmax layer.

[0095] The output of the LSTM layer is calculated based on the input of the current time step and the hidden state of the previous time step. The calculation method is as follows:

[0096] Input gate: i t =σ(W xi x t +W hi h t-1 +b i )

[0097] Forget gate: f t =σ(W xf x t +W hf h t-1 +b f )

[0098] Update candidate value:

[0099] Cell state update:

[0100] Output gate: o t =σ(W xo x t +W ho h t-1 +b o )

[0101] Hidden state update: h t =o t ⊙tanh(C t )

[0102] In these formulas, x t is the input of the current time step, h t-1 is the hidden state W of the previous time step xi , W hi , W xf , W hf , W xc , W hc , W xo , W ho is the weight matrix, b i 、b f 、b c 、b o is the bias vector, σ is the sigmoid activation function, ⊙ represents element-wise multiplication, and tanh is the hyperbolic tangent activation function.

[0103] In order to verify the reliability of the VMD-COMTBO-LSTM fault diagnosis model of the present invention, a photovoltaic array simulation model was built on the Matlab / Smulink simulation platform. According to the five fault types mentioned above, simulation data under 11 fault labels were obtained. The temperature was selected from [10, 45] °C with a step size of 1 °C. The light intensity was selected from [100, 1000] W / m 2 , at 20W / m 2 is the step length; a total of 36×46×11=18216 sets of fault data are obtained, and the ratio of the test set to the training set is 0.8. The simulation labels are as follows in Table 2:

[0104] Table 2: Fault selection

[0105]

[0106] In order to verify that the VMD-COMTBO-LSTM fault diagnosis model of the present invention can more accurately diagnose the faults of photovoltaic arrays, the VMD-LSTM model, mountain climbing optimization algorithm (MTBO), osprey optimization algorithm (OOA), sparrow search algorithm (SSA) and gray wolf optimization algorithm (GWO) are optimized by comparative analysis. The same training set and test set are used for experiments. The fault diagnosis results of VMD-LSTM, VMD-MTBO-LSTM, VMD-OOA-LSTM, VMD-SSA-LSTM, VMD-GWO-LSTM models and VMD-COMTBO-LSTM model of the present invention are as follows: Figure 6 Shown

[0107] from Figure 6 It can be seen that the accuracy of VMD-LSTM, VMD-SSA-LSTM, and VMD-MTBO-LSTM are all below 90%, while the remaining three models are all above 90%. VMD-COMTBO-LSTM has the best effect, with an accuracy of about 98%, which verifies the feasibility and accuracy of photovoltaic array fault diagnosis based on the VMD-COMTBO-LSTM model of the present invention.

[0108] The present invention integrates MTBO and OOA to achieve parameter optimization for LSTM. Compared with SSA, GWO and MTBO and OOA before integration, the optimization effect is better than the above algorithms, showing better local wit escape ability and parameter optimization ability;

[0109] The present invention is a photovoltaic array fault diagnosis model based on VMD-COMTBO-LSTM, which can accurately identify photovoltaic array short circuit, open circuit, shading, aging and line faults, avoid irreversible losses to photovoltaic panels, and improve the power generation efficiency of photovoltaic arrays.

[0110] The simulation proves that the VMD-COMTBO-LSTM model has high feasibility and application prospects, and its fault diagnosis rate reaches about 98%, which is higher in accuracy than the other four groups of models.

Claims

1. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model, characterized in that: The method comprises the following steps: A photovoltaic array fault diagnosis model is built based on the LSTM model, and the parameters of the LSTM model are optimized using the Cauchy-Osprey-Mountaineering Team-Based Optimization algorithm that combines Cauchy mutation and Osprey to build a COMTBO-LSTM fault diagnosis model. The photovoltaic array fault data under short circuit fault, open circuit fault, ground fault and line fault are used as training data sets, and then the data is decomposed by variational mode decomposition. The decomposed data is used to train the constructed OMTBO-LSTM fault diagnosis model to obtain the trained VMD-COMTBO-LSTM fault diagnosis model. Obtain the PV array fault data to be diagnosed and extract characteristic parameters, and perform VMD decomposition on the fault data; The decomposed data is input into the trained COMTBO-LSTM fault diagnosis model, and the fault diagnosis result is output.

2. According to claim 1, a photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model is characterized in that: The steps for integrating Mountain Climbing Optimization (MTBO) and Osprey Optimization Algorithm (OOA) are as follows: (1) Using the Circle mapping strategy to initialize the climbers, relative to the randomly distributed climbers; (2) The first stage of OOA is used to replace the assisted mountaineering position update formula in the original MTBO; (3) Cauchy mutation is used to replace the random member replacement stage.

3. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 1, characterized in that: The steps for training the constructed COMTBO-LSTM fault diagnosis model using the training dataset include: (1) Build a photovoltaic array model in Matlab / Simulink, simulate different faults, and collect photovoltaic array fault data; (2) Label the data according to different fault types and establish a fault data set; (3) Normalization of fault data; (4) Set VMD parameters, perform VMD decomposition on the normalized data, and select the component with the largest energy to fuse with the label; (5) Randomly select the test set and training set according to the ratio of test set to training set of 0.8; (6) Set the COMTBO-LSTM parameters according to Table 4 and import the training set for training; (7) After training, the model performs fault diagnosis on the test set.

4. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 2, characterized in that: The number K of VMD parameter components.

5. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 2, characterized in that: The initialization parameters of the COMTBO algorithm include the population number N, the maximum number of iterations T max .

6. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 2, characterized in that: The LSTM parameters include the optimal hidden unit 1, 2 number L1 and L1, the number of iterations K, the optimal initial learning rate L r .

7. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 2, characterized in that: The fitness value of the climber is calculated using the root mean square error (MSE) between the predicted and actual outputs as follows: Where n is the number of test sets, y i is the actual value, is the predicted value.

8. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to claim 2, characterized in that: Combining the Cauchy mutation and Osprey's mountain climbing algorithm, we first use the Circle mapping strategy to initialize the climbers: x i =r i cos(θ i ) y i =r i sin(θ i ) In the formula, n is the number of populations, i is the index of an individual, and r is i Indicates the random generation radius, a random number uniformly distributed in the interval [0,1); The first stage of OOA is used to replace the assisted climbing position update formula in the original MTBO: x i,j =lb i,j +r i,j ·(ub j -lb j ),i=1,2,…,N,j=1,2,…,m, Using Cauchy mutation to replace the random replacement member stage can expand the search range of MTBO and improve the algorithm's ability to jump out of the local optimal solution; the Cauchy mutation position update formula is as follows: Where cauchy(0,1) is the standard Cauchy distribution function; is multiplied; one-dimensional Cauchy variogram centered at the origin:

9. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to any one of claims 1 to 8, characterized in that: The characteristic parameters used include light intensity G, temperature T, short-circuit current I SC , open circuit voltage U oc , Maximum power point P m , Maximum power point voltage U m , Maximum power point current I m , F1, F2, F3, F4, and filling factor FF are used as fault characteristics; the calculation formula for F1, F2, F3, F4, and filling factor FF is as follows: Its V oc,STC ,I sc,STC 、V m,STC and I m,STC is the corresponding value of the ideal value condition, and the parameters (α and β) are I sc and V oc Temperature coefficient; V oc,n ,I sc,n ,V m,n ,I m,n At G = 1000w / m 2 , the data records corresponding to the normalized ideal value (STC) at T = 25°C are calculated according to the following formula:

10. A photovoltaic array fault diagnosis method based on VMD-COMTBO-LSTM model according to any one of claims 1 to 8, characterized in that: The method comprises: Construct an LSTM model to diagnose photovoltaic array faults, use the COMTBO algorithm to optimize the LSTM model parameters, perform VMD decomposition on photovoltaic fault data as training data, and construct a VMD-COMTBO-LSTM fault diagnosis model; Data acquisition: Acquisition of photovoltaic fault data under short circuit fault, open circuit fault, line-to-line fault, aging and shading fault to form the initial data set; Data processing: VMD data decomposition is performed on the initial data set, and the modal component with the maximum energy is selected and fused with the label as the data set. The test set and training set are randomly selected in a ratio of 4:1; Set the COMTBO-LSTM parameters and import the training set for training. After the training is completed, the model performs fault diagnosis on the test set and outputs the fault diagnosis results.

Images