New energy station operation and maintenance fault diagnosis method and device and readable storage medium

Abstract

The invention discloses a new energy station operation and maintenance fault diagnosis method and device and a readable storage medium. The method comprises the following steps: clustering each photovoltaic power station in a new energy station, and establishing a prediction model corresponding to each class group subgroup; determining a predicted value based on a prediction model corresponding to the photovoltaic power station to be diagnosed, and comparing the predicted value with the measured value; and when the deviation exceeds a threshold value, inputting the electrical characteristic parameters of each group of photovoltaic modules in the photovoltaic power station to be diagnosed into the fault analysis model so as to obtain the fault type of each group of photovoltaic modules. According to the method, the BP neural network algorithm is improved through the ant colony algorithm, the final reserved population is selected by adopting a fitness ranking method in the ant colony algorithm, the initial population number of the genetic algorithm can be reduced to a certain extent, the optimization time of the genetic algorithm is shortened, and the classification efficiency of fault diagnosis is improved.

Description

Technical Field

[0001] The present application relates to the field of operation and maintenance management technology, and specifically to a new energy station operation and maintenance fault diagnosis method, equipment and computer-readable storage medium. Background Art

[0002] In recent years, the development of new energy represented by wind power and photovoltaic power generation has achieved remarkable results, with the installed capacity and power generation proportion steadily increasing, and the energy structure adjustment and carbon reduction effects gradually emerging. New energy stations are generally built in remote locations, with dispersed geographical locations, wide distribution areas, and complex on-site operating conditions. The existing operation and maintenance mode is mostly on-site maintenance, which requires more operation and maintenance personnel to perform regular inspection and maintenance functions in these places. For new energy power generation companies, establishing an intelligent automatic fault analysis and early warning system is the best solution to achieve safe operation, easy management, and reduce costs.

[0003] In the existing technology, when performing multi-feature fault analysis of equipment, BP neural network, genetic algorithm, sparrow search algorithm and other methods are often used for fault diagnosis. However, genetic algorithm belongs to the random algorithm, which has a certain dependence on the selection of the initial chromosome population size and quality. In some cases, the initial chromosome population of the genetic algorithm has poor traversal, and in the subsequent optimization process, it is difficult for the algorithm to complete the search of the entire interval. When solving problems, the traditional BP neural network sometimes has multiple peaks in function processing due to the large differences between data samples, which makes it easy to fall into the local optimum and unable to jump out. Summary of the invention

[0004] In order to solve the above technical problems, the present application provides a new energy station operation and maintenance fault diagnosis method, equipment and computer-readable storage medium.

[0005] In a first aspect, an embodiment of the present application provides a method for diagnosing faults in operation and maintenance of a new energy station, the method comprising:

[0006] Cluster multiple photovoltaic power stations in the new energy station according to similarity to obtain multiple groupings;

[0007] For each cluster group, the photovoltaic power generation power and corresponding meteorological factor data of each photovoltaic power station included in the cluster group are obtained and used as training data, and a photovoltaic power generation power prediction model corresponding to the cluster group is established based on the training data;

[0008] Determine the target group to which the photovoltaic power station to be diagnosed belongs;

[0009] Input the meteorological factor data corresponding to the time point to be predicted into the target photovoltaic power prediction model corresponding to the target group group;

[0010] Compare the photovoltaic power generation power prediction value output by the target photovoltaic power generation power prediction model with the photovoltaic power generation power actual measurement value of the photovoltaic power station to be diagnosed at the time point to be predicted, and determine the deviation value;

[0011] When the deviation exceeds the threshold, the electrical characteristic parameters of each group of photovoltaic components included in the photovoltaic power station to be diagnosed are input into the fault analysis model, wherein the fault analysis model is obtained based on training the improved BP neural network model, and the BP neural network model is improved by the ant colony algorithm to obtain the improved BP neural network model;

[0012] Based on the fault type classification result matrix output by the fault analysis model, the fault type of each group of photovoltaic modules is determined.

[0013] In a second aspect, an embodiment of the present application provides a new energy station operation and maintenance fault diagnosis device, which includes a processor, a memory, and a new energy station operation and maintenance fault diagnosis program stored in the memory and executable by the processor, wherein when the new energy station operation and maintenance fault diagnosis program is executed by the processor, the steps of the new energy station operation and maintenance fault diagnosis method described in the first aspect are implemented.

[0014] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, on which a new energy station operation and maintenance fault diagnosis program is stored. When the new energy station operation and maintenance fault diagnosis program is executed by a processor, the steps of the new energy station operation and maintenance fault diagnosis method described in the first aspect are implemented.

[0015] The beneficial effects brought by the technical solution provided in the embodiments of the present application include:

[0016] In the embodiment of the present application, the photovoltaic power stations in the new energy station are clustered, and a prediction model corresponding to each group grouping is established; the prediction value is determined based on the prediction model corresponding to the photovoltaic power station to be diagnosed, and compared with the measured value; when the deviation exceeds the threshold, the electrical characteristic parameters of each group of photovoltaic components in the photovoltaic power station to be diagnosed are input into the fault analysis model to obtain the fault type of each group of photovoltaic components. Through the embodiment of the present application, the BP neural network algorithm is improved by the ant colony algorithm, and the fitness ranking method in the ant colony algorithm is used to select the final retained group, which can reduce the initial population of the genetic algorithm to a certain extent, reduce the time used for the genetic algorithm to find the best, and improve the classification efficiency of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of an embodiment of a method for diagnosing faults in operation and maintenance of a new energy station of the present application;

[0018] Figure 2This is a schematic diagram of the hardware structure of the new energy station operation and maintenance fault diagnosis equipment involved in the embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0021] In a first aspect, an embodiment of the present application provides a method for diagnosing faults in operation and maintenance of a new energy station.

[0022] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of an embodiment of a method for diagnosing faults in operation and maintenance of a new energy station in this application. Figure 1 As shown in the figure, the fault diagnosis method for operation and maintenance of new energy stations includes:

[0023] Step S10, clustering multiple photovoltaic power stations in the new energy station according to similarity to obtain multiple groupings;

[0024] In this embodiment, the characteristic value of each photovoltaic power station is obtained, and then the similarity between two photovoltaic power stations is calculated based on the characteristic values ​​of two photovoltaic power stations, so as to perform clustering, that is, photovoltaic power stations with high similarity are grouped into one group, thereby obtaining multiple group groups.

[0025] Furthermore, in one embodiment, step S10 includes:

[0026] Step S101, grouping each photovoltaic power station in the new energy station into a group;

[0027] Step S102, using a clustering algorithm to calculate the distance between every two cluster groups;

[0028] In this embodiment, it should be noted that the distance referred to here is a cluster distance, for example, a Euclidean distance calculated based on multi-dimensional parameters of two cluster groups.

[0029] When a group includes multiple photovoltaic power stations, the multi-dimensional parameters of the multiple photovoltaic power stations are first averaged to obtain the average value of each dimensional parameter, and the average value of each dimensional parameter is used as the multi-dimensional parameter of the group.

[0030] When a group includes one photovoltaic power station, the multi-dimensional parameters of the photovoltaic power station are used as the multi-dimensional parameters of the group.

[0031] Step S103, calculating the correlation of each grouping based on the output power and meteorological parameters of the photovoltaic power station included in each grouping;

[0032] In this embodiment, for each grouping, assuming that it includes multiple photovoltaic power stations, the correlation between the output power of each photovoltaic power station and the meteorological parameter is calculated respectively, and then the multiple correlations are averaged to obtain the correlation of the grouping. If only one photovoltaic power station is included, the correlation between the output power of the photovoltaic power station and the meteorological parameter is calculated and used as the correlation of the grouping. The meteorological parameters may include irradiance, temperature, humidity, air quality index, etc.

[0033] Step S104, calculating the similarity between each two groupings based on the distance between each two groupings and the correlation between each two groupings;

[0034] In this embodiment, a similarity calculation method is used to calculate the similarity between every two groupings based on the distance between every two groupings and the correlation between every two groupings.

[0035] Step S105 , merging the two cluster groups with the highest similarity into a new cluster group, and returning to the step of calculating the distance between every two cluster groups using a clustering algorithm until the number of cluster groups reaches a preset number.

[0036] In this embodiment, assuming that there are group groups 1 to 3, the similarity results obtained according to step S104 include similarity 1 between group group 1 and group group 2, similarity 2 between group group 1 and group group 3, and similarity 3 between group group 2 and group group 3, where similarity 2 is the maximum value of the three similarities, then group group 1 and group group 3 are merged, and the current number of group groups is 2.

[0037] If the number of the grouped categories reaches the preset number, the clustering ends; if the number of the grouped categories does not reach the preset number, the process returns to step S102.

[0038] Step S20, for each grouping, obtaining the photovoltaic power generation power and corresponding meteorological factor data of each photovoltaic power station included in the grouping and using them as training data, and establishing a photovoltaic power generation power prediction model corresponding to the grouping based on the training data;

[0039] In this embodiment, because each cluster grouping represents a type of photovoltaic power stations with high similarity, the power generation power of these photovoltaic power stations and the corresponding meteorological factor data can be used as training data, and the meteorological factor data can be used as input and the corresponding power generation power as a label to train the photovoltaic power generation power prediction model corresponding to the cluster grouping.

[0040] Furthermore, in one embodiment, establishing a photovoltaic power prediction model corresponding to the grouping based on the training data includes:

[0041] The training data set is input into the photovoltaic power generation prediction model for calculation, the performance of the photovoltaic power generation prediction model is evaluated, and the photovoltaic power generation prediction model with the optimal parameters is determined.

[0042] In this embodiment, inputting the training data set into the photovoltaic power generation prediction model specifically refers to inputting the meteorological factor data in the training data set into the photovoltaic power generation prediction model.

[0043] The performance evaluation of the photovoltaic power prediction model can be carried out based on the output of the photovoltaic power prediction model and the corresponding labels. The model performance can be evaluated by means of root mean square error, mean absolute error, accuracy, deviation percentage, etc., so as to determine the photovoltaic power prediction model with the optimal parameters.

[0044] Further, in one embodiment, inputting the training data set into the photovoltaic power prediction model for calculation includes:

[0045] Inputting the training data set into the first power prediction model to obtain an initial prediction value;

[0046] Subtracting the initial predicted value from the actual power data to obtain a correction data set;

[0047] The correction data set is input into the photovoltaic power error correction model to obtain the correction prediction value;

[0048] The initial prediction value is superimposed with the corrected prediction value to obtain the corrected photovoltaic power generation power.

[0049] Furthermore, in one embodiment, the photovoltaic power error correction model adopts the following method:

[0050] Initialize the algorithm parameters, including the number of ant colonies, number of iterations, initial positions and initial speeds of ants, and spatial dimensions;

[0051] Construct the network topology and obtain the parameters that need to be optimized;

[0052] The mean absolute error (MAE) is used as the fitness function, and the fitness value is calculated according to the initial position of the ant, so as to obtain the initial global optimal and individual optimal values ​​of the ant;

[0053] Use the obtained fitness value to update the ant speed and position, calculate the fitness value again based on the updated ant position, and find the individual optimal position of the ant and the global optimal position;

[0054] When the number of iterations reaches the maximum value, the iteration process ends, and the found hyperparameters are brought into the prediction model to predict the output results; if not, return to the previous step to continue the iteration.

[0055] The above scheme of this embodiment predicts the prediction error and adds the output of the first power prediction model to the result of the correction prediction model to obtain the final prediction result. The result is superior to the prediction results of the traditional SVM model and the LSTM model, so that the above photovoltaic power prediction based on error correction in this embodiment can effectively improve the prediction accuracy of photovoltaic power.

[0056] It should be noted that, since photovoltaic power generation is greatly affected by external factors, its time series will show certain volatility. Directly taking the original power data for prediction based on the SVM regression model will produce large errors. Therefore, before using the SVM regression model for prediction, the photovoltaic power generation signal needs to be processed to solve the problem of large prediction errors of the SVM single model.

[0057] Based on the above considerations, this embodiment uses a modal decomposition algorithm to first decompose the fluctuating photovoltaic power generation original power signal into multiple periodic components to solve the problem of local characteristic oscillation.

[0058] Further, in one embodiment, inputting the training data set into the first power prediction model to obtain the initial prediction value includes:

[0059] Step S201, decomposing the historical photovoltaic power data using a decomposition algorithm to obtain a plurality of intrinsic mode components IMF and residual components RES;

[0060] Step S202, using the training sets of each intrinsic mode component IMF and residual component RES to establish respective SVM regression models;

[0061] Step S203, using the sparrow search algorithm to optimize the penalty parameter c and the kernel function parameter g of each SVM regression model, and predicting the test set while obtaining the prediction model of each intrinsic mode component IMF and residual component RES;

[0062] Step S204 , superimposing the photovoltaic power values ​​predicted by the prediction models of the intrinsic mode components IMF and the residual component RES to obtain an initial prediction value.

[0063] Through the above scheme, this embodiment decomposes the historical photovoltaic power data into intrinsic mode components and residual components of different frequencies, inputs each IMF component into the prediction model respectively, and finally reconstructs the prediction results of each frequency component and residual component to obtain the final prediction result, which can effectively reduce the large error in the prediction result caused by the volatility of photovoltaic power generation.

[0064] The decomposition algorithm used in step S201 is specifically:

[0065] S2011, a pair of opposite positive and negative white noises are added to the original power signal of photovoltaic power generation to generate two sets of mixed signals;

[0066] S2012, performing modal decomposition on the two groups of mixed signals respectively to obtain two groups of positive and negative IMF components;

[0067] S2013, repeating steps S2011 and S2012 several times, adding a new normally distributed Gaussian white noise each time;

[0068] S2014, averaging each set of positive and negative IMF components to obtain a plurality of intrinsic mode components IMF and residual components RES.

[0069] Furthermore, the above modal decomposition method is specifically as follows:

[0070] S20121, extracting the upper extreme point and the lower extreme point of the new photovoltaic power generation signal;

[0071] S20122, generating an upper envelope and a lower envelope based on the new photovoltaic power generation signal and the upper extreme point and the lower extreme point, and calculating the mean of the envelope; subtracting the new photovoltaic power generation signal from the mean of the envelope to obtain a detail signal of the local oscillation;

[0072] S20123, judging whether the detail signal satisfies the termination condition of decomposition, such as whether the number of extreme points is less than 3, or whether the maximum and minimum values ​​appear alternately without repeated extreme points, if the termination condition is satisfied; if the detail signal does not satisfy the termination condition, it is regarded as a new photovoltaic power generation signal, and the above steps are repeated to obtain an IMF;

[0073] S20124, repeat step S20123 until the last IMF that meets the termination condition is obtained.

[0074] It should be noted that during the actual operation of a photovoltaic power station, the data recorded in the system usually contains some missing values ​​and / or outliers. Therefore, it is necessary to preprocess these missing values ​​and / or outlier data to improve the data quality during subsequent processing, so as to avoid large deviations in subsequent prediction results.

[0075] Among them, outliers are usually data recorded for system anomalies caused by hardware anomalies in the system (such as poor sensor contact or sensor anomalies, etc.). If the abnormal data in the data is not processed, the abnormal data will participate in the subsequent model training, which will cause the model to train on the outliers, which will lead to certain errors in the model output results.

[0076] Therefore, the embodiment of the present application first checks the abnormal data in the data, takes the detected abnormal data as missing values, and uses the interpolation method to calculate the filling value to fill in the missing value to ensure data integrity.

[0077] As for the real missing values, they can be obtained through statistical counting, and the interpolation method can be used to repair these missing data.

[0078] It should be noted that after the above processing, the data corresponding to the points of missing values ​​are obtained through interpolation, and the data corresponding to the points of abnormal values ​​are first detected and deleted and then interpolated. Therefore, the data at these points are repaired data, which may produce certain deviations. Therefore, it is necessary to evaluate the interpolated data later to determine the deviation of the interpolated data. Therefore, the embodiment of the present application also needs to mark these repaired data points so that these repaired data can be evaluated after the prediction model is established.

[0079] In some embodiments, the determination of the abnormal value of each feature in the photovoltaic data set may be performed by the following method:

[0080] First, all the data of a feature are statistically calculated to obtain the first limit value A 上 , the second limit value A 下 And the limit distance A 间距 ; Among them, the first limit value A 上 The second limit value A is the value at the 25% position of all data. 下 It is the value at 75% of all data, and the limit interval is A 间距 is the difference between the first limit value and the second limit value;

[0081] The upper and lower limits of the outlier are calculated; the upper limit of the outlier is A 上 +1.5A 间距 , the lower limit of the outlier is A 下 -1.5A 间距 ;

[0082] Finally, all data of a feature are substituted into the calculation range formed by the upper and lower limits of the abnormal value, and values ​​above the upper and lower limits of the abnormal value and below the lower limit of the abnormal value are all abnormal values.

[0083] The various data output by the photovoltaic power generation system usually have units, such as irradiance in W / m 2 , temperature is ℃, humidity is %, and air quality index is a dimensionless quantity. The data ranges of various types of data are also very different. If they are directly input into the subsequent prediction model without processing, the large-dimensional data will usually be larger than the small-dimensional data. At this time, the large-dimensional data will be more prominent in the prediction model, which will make it almost impossible for the prediction model to perform machine learning calculations on the small-dimensional data. Therefore, it is necessary to normalize the data to improve the training efficiency of the prediction model and avoid data distribution disorder caused by dimensions.

[0084] It should be further explained that, when the prediction model is subsequently trained, the data in the training data set is regularized using the above method, that is, the data at some times are missing or interpolated due to anomalies, and are not the original real data.

[0085] Therefore, when the photovoltaic power generation data and the meteorological factor data are respectively regularized in the embodiment of the present application, the regularized photovoltaic power generation data and the meteorological factor data are also marked.

[0086] When the photovoltaic power generation data of the photovoltaic power stations grouped in each category and the meteorological factor data at the same time are used as training data sets, and the training data sets are input into the photovoltaic power generation prediction model for calculation, the data at the marked time points are screened out, and the predicted photovoltaic power generation data are used to replace the data initially supplemented by the interpolation method; finally, the photovoltaic power generation data of the photovoltaic power station after replacement and the meteorological factor data at the same time are used as training data sets, and the training data sets are input into the photovoltaic power generation prediction model for secondary calculation; the performance of the trained photovoltaic power generation prediction model is evaluated using the secondary calculation structure at this time.

[0087] Furthermore, since the output of photovoltaic power stations is greatly affected by geographical terrain and climatic conditions, different weather conditions such as wind speed, irradiance, cloud cover, etc. will cause the output of photovoltaic power stations in sub-regions to vary greatly. For example, in areas such as rivers and lakes, beaches, wastelands, saline-alkali land, and sandy land, the temperature, humidity, and air quality index will be different, which will affect the power generation in these areas. Therefore, when predicting power, it is necessary to perform classified prediction based on the location of the photovoltaic power generation unit, which can effectively improve the accuracy of the prediction.

[0088] Therefore, before using the prediction model for prediction calculation, the embodiment of the present application also designs a step of calculating the correlation between the output power and meteorological parameters of each photovoltaic power station in the photovoltaic power station cluster, clustering the photovoltaic power stations, and obtaining the grouping steps of the photovoltaic power stations.

[0089] The specific method is as follows:

[0090] Standardize the power data of each photovoltaic power station and construct an overall data matrix;

[0091] Each photovoltaic power station is regarded as a class, and the distance between each two power stations is calculated using the clustering algorithm to construct a clustering distance matrix. At the same time, the correlation between the output power of each photovoltaic power station and the meteorological parameters is calculated;

[0092] The similarity of the two power stations is calculated based on the distance between the two power stations and the correlation between the output power of the two power stations and the meteorological parameters;

[0093] Merge the two most similar categories as a new category, update the number of categories, use the weighted average method to analyze the average distance between the new categories, and update the cluster distance matrix;

[0094] Continue the previous step until all categories are clustered into one category;

[0095] Finally, the number of classifications is determined to obtain the final clustering results of the photovoltaic power station.

[0096] After the photovoltaic power stations are clustered and grouped using the above method, the data of the corresponding groups are processed separately during subsequent predictions, and finally the data of each group are superimposed to obtain the total data of the entire photovoltaic power station.

[0097] Step S30, determining the target group to which the photovoltaic power station to be diagnosed belongs;

[0098] In this embodiment, one grouping consists of photovoltaic power plants with similar characteristics, so a target grouping matching the photovoltaic power plant to be diagnosed can be searched according to its characteristics.

[0099] Step S40, inputting the meteorological factor data corresponding to the time point to be predicted into the target photovoltaic power prediction model corresponding to the target group group;

[0100] Step S50, comparing the photovoltaic power generation prediction value output by the target photovoltaic power generation prediction model with the photovoltaic power generation power actual value at the photovoltaic power station to be diagnosed at the time point to be predicted, and determining a deviation value;

[0101] In this embodiment, the meteorological factor data corresponding to the time point to be predicted is input into the target photovoltaic power generation prediction model to obtain the photovoltaic power generation prediction value output by the target photovoltaic power generation prediction model. The photovoltaic power generation prediction value is compared with the photovoltaic power generation power measured value of the photovoltaic power station to be diagnosed at the time point to be predicted to determine the deviation value.

[0102] Step S60, when the deviation exceeds the threshold, the electrical characteristic parameters of each group of photovoltaic components included in the photovoltaic power station to be diagnosed are input into the fault analysis model, wherein the fault analysis model is obtained based on training an improved BP neural network model, and the BP neural network model is improved by an ant colony algorithm to obtain an improved BP neural network model;

[0103] In this embodiment, if the deviation exceeds the threshold, it means that some solar panels in the photovoltaic power station to be diagnosed are faulty, so it is necessary to perform fault diagnosis through the fault analysis model. Among them, the electrical characteristic parameters input into the fault analysis model include open circuit voltage, short circuit current, maximum power voltage and maximum power current. It should be noted that this is only a schematic description of the electrical characteristic parameters and does not constitute a limitation thereto.

[0104] Further, in one embodiment, the fault analysis model is implemented by the following steps:

[0105] Collect the electrical characteristic parameters and fault type classification results of sample PV modules, some of which are used as training samples and some as test samples;

[0106] Initially set the parameters of the improved BP neural network model;

[0107] The improved BP neural network model with initial parameters is used as the model to be trained, the electrical characteristic parameters in the training sample are used as input, and the fault type classification result is used as output to train the model to be trained, so as to obtain the fault analysis model to be verified;

[0108] The fault analysis model to be verified is verified through the test sample. If the verification passes, the fault analysis model to be verified is used as the fault analysis model.

[0109] In this embodiment, it should be noted that the prior art often uses BP neural network, genetic algorithm, sparrow search algorithm and other methods for fault diagnosis when performing multi-feature fault analysis, but the genetic algorithm belongs to a random algorithm and has a certain dependence on the selection of the initial chromosome population size and quality. In some cases, the initial chromosome population of the genetic algorithm has poor traversal, and in the subsequent optimization process, it is difficult for the algorithm to complete the search of the entire interval. When the traditional BP neural network is used for problem solving, sometimes the differences between data samples are too large, resulting in multiple peaks in function processing, and it is easy to fall into the local optimum and cannot jump out.

[0110] Therefore, the embodiment of the present application further improves the BP neural network algorithm, that is, the BP neural network algorithm is improved through the ant colony algorithm. First, a certain number of different numerical values ​​are selected as the initial population of the ant colony algorithm, and then they participate in the iterative process of the ant colony algorithm in turn. After several iterations of the ant colony algorithm, the chromosome with the best fitness is finally decoded, and the weights and thresholds generated by the decoding are assigned to the BP neural network. This not only shortens the iteration time of the neural network training and converges quickly, but also enhances the ability of the neural network to jump out of local extreme values ​​and achieve higher accuracy.

[0111] Furthermore, in one embodiment, the improved BP neural network model is implemented by the following method:

[0112] Step 1: Initialize the weights and thresholds of the BP neural network model, and set the initial parameters of the BP neural network and the ant colony algorithm;

[0113] Step 2: Assign each ant individual its initial position and corresponding initial velocity, recorded as:

[0114] X(i)=[X1(i),X2(i)...X d (i)],i=1,2...N

[0115] V(i)=[V1(i),V2(i)...V d (i)], i = 1, 2...N;

[0116] Where N is the total number of ants in the ant colony, and d is the number of fault types.

[0117] When the ants move, each position represents a solution to the function problem. The ants continuously update their positions during this process. At the t+1th iteration, their positions are:

[0118] X (t) (i) = X (t-1) (i)+V (t) (i), i=1,2...N;

[0119] Among them, X (t) (i) is the position of the ant in the current iteration, X (t-1) (i) is the position of the ant in the last iteration, V (t) (i) is the speed of the ant in the current iteration.

[0120] In addition, the individual ants in the ant colony arrive at the location they want to reach at the next moment at different speeds, and their speed changes are updated as follows:

[0121] V (t) (i) = V (t-1)(i)e -0.2 +r(X g -X (t-1) (i));

[0122] Among them, r is a random number between 0 and 1, X g is the current global optimal position.

[0123] It can be understood that in the above model, the ant individuals are constantly moving, and their positions and speeds are gradually updated to obtain the distance between the current position and the destination. The size of this distance is selected as the fitness, and the fitness is used to evaluate the quality of the solution corresponding to the current position. After that, the positions of all ant individuals are compared, and the ant individual with the best fitness is selected. This iterative process will be performed several times until the end condition is met, and the position result obtained in the process will be delivered to the ant individual, that is, the global optimal position X is updated accordingly. g , to proceed to the next stage of iteration.

[0124] Step 3: Calculate the fitness of each ant individual and update the global optimal fitness;

[0125] Step 4: When the maximum number of iterations is greater than the first number of iterations, arrange the ant individuals in descending order of fitness and execute step 5. If it is less than the maximum number of iterations, return to step 3 and iterate in a loop;

[0126] Step 5: Update the position of each ant and calculate the position center of the ant colony;

[0127] Step 6, determine whether the number of iterations is greater than the second number of iterations. If not, go to step 6 and loop the calculation process. If satisfied, go to step 7;

[0128] Step 7: Rank all ant individuals according to fitness, select the top N ants as the initial chromosomes of the genetic algorithm, and perform encoding operations, where N is set according to actual needs;

[0129] Step 8: Select, inherit and mutate the chromosomes;

[0130] Step 9: Calculate the fitness of all ant individuals. If the fitness meets the requirements, proceed to step 10, otherwise return to step 8 for a loop.

[0131] Step 10: Complete the decoding of the chromosome, assign the optimal weight and threshold to the BP neural network model, and obtain the improved BP neural network model.

[0132] It should be noted that the improved BP neural network algorithm adopted in the embodiment of the present application, by adding an ant colony algorithm, adopts the fitness ranking method in the ant colony algorithm to select the final retained population, thereby giving full play to the characteristics of the ant colony algorithm's fast global optimization and short algorithm iteration time, reducing the initial population number of the genetic algorithm to a certain extent, and greatly reducing the time taken for the genetic algorithm to optimize.

[0133] Specifically, according to the change law of characteristic parameters when a photovoltaic array fails, the embodiment of the present application selects open circuit voltage, short circuit current, maximum power voltage and maximum power current as fault type data, that is, the number of input layer nodes in the improved BP neural network model is m=4, and the output layer is required to represent 5 fault states, that is, the number of output layer nodes is n=5, and the number of hidden layer nodes is selected between [4, 13], preferably 10.

[0134] Step S70: determining the fault type of each group of photovoltaic components based on the fault type classification result matrix output by the fault analysis model.

[0135] In this embodiment, each element in the fault type classification result matrix represents the fault type of the corresponding photovoltaic module.

[0136] In the embodiment of the present application, the photovoltaic power stations in the new energy station are clustered, and a prediction model corresponding to each group grouping is established; the prediction value is determined based on the prediction model corresponding to the photovoltaic power station to be diagnosed, and compared with the measured value; when the deviation exceeds the threshold, the electrical characteristic parameters of each group of photovoltaic components in the photovoltaic power station to be diagnosed are input into the fault analysis model to obtain the fault type of each group of photovoltaic components. Through the embodiment of the present application, the BP neural network algorithm is improved by the ant colony algorithm, and the fitness ranking method in the ant colony algorithm is used to select the final retained group, which can reduce the initial population of the genetic algorithm to a certain extent, reduce the time used for the genetic algorithm to find the best, and improve the classification efficiency of fault diagnosis.

[0137] In a second aspect, an embodiment of the present application provides a new energy station operation and maintenance fault diagnosis device, which may be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.

[0138] Reference Figure 2 , Figure 2 The hardware structure diagram of the new energy station operation and maintenance fault diagnosis device involved in the embodiment of the present application is shown in FIG. In the embodiment of the present application, the new energy station operation and maintenance fault diagnosis device may include a processor, a memory, a communication interface, and a communication bus.

[0139] The communication bus may be of any type and is used to interconnect the processor, the memory, and the communication interface.

[0140] The communication interface includes input / output (I / O) interface, physical interface and logical interface, etc., which are used to realize the interconnection of devices inside the new energy station operation and maintenance fault diagnosis equipment, and the interface used to realize the interconnection between the new energy station operation and maintenance fault diagnosis equipment and other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber optic interface, an ATM interface, etc.; the user device can be a display, a keyboard, etc.

[0141] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0142] The processor may be a general-purpose processor, which may call the new energy station operation and maintenance fault diagnosis program stored in the memory and execute the new energy station operation and maintenance fault diagnosis method provided in the embodiment of the present application. For example, the general-purpose processor may be a central processing unit (CPU). The method executed when the new energy station operation and maintenance fault diagnosis program is called may refer to the various embodiments of the new energy station operation and maintenance fault diagnosis method of the present application, which will not be repeated here.

[0143] Those skilled in the art will understand that Figure 2 The hardware structure shown in the figure does not constitute a limitation on the present application, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0144] In a third aspect, an embodiment of the present application also provides a computer-readable storage medium.

[0145] The computer-readable storage medium of the present application stores a new energy station operation and maintenance fault diagnosis program, wherein when the new energy station operation and maintenance fault diagnosis program is executed by a processor, the steps of the new energy station operation and maintenance fault diagnosis method as described above are implemented.

[0146] Among them, the method implemented when the new energy station operation and maintenance fault diagnosis program is executed can refer to the various embodiments of the new energy station operation and maintenance fault diagnosis method of the present application, and will not be repeated here.

[0147] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0148] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.

[0149] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.

[0150] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or. For example, A / B can mean A or B. The “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0151] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0152] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present application.

[0153] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A new energy station operation and maintenance fault diagnosis method, characterized in that: The new energy station operation and maintenance fault diagnosis method comprises: Cluster multiple photovoltaic power stations in the new energy station according to similarity to obtain multiple groupings; For each cluster group, the photovoltaic power generation power and corresponding meteorological factor data of each photovoltaic power station included in the cluster group are obtained and used as training data, and a photovoltaic power generation power prediction model corresponding to the cluster group is established based on the training data; Determine the target group to which the photovoltaic power station to be diagnosed belongs; Input the meteorological factor data corresponding to the time point to be predicted into the target photovoltaic power prediction model corresponding to the target group group; Compare the photovoltaic power generation power prediction value output by the target photovoltaic power generation power prediction model with the photovoltaic power generation power actual measurement value of the photovoltaic power station to be diagnosed at the time point to be predicted, and determine the deviation value; When the deviation exceeds the threshold, the electrical characteristic parameters of each group of photovoltaic components included in the photovoltaic power station to be diagnosed are input into the fault analysis model, wherein the fault analysis model is obtained based on training the improved BP neural network model, and the BP neural network model is improved by the ant colony algorithm to obtain the improved BP neural network model; Based on the fault type classification result matrix output by the fault analysis model, the fault type of each group of photovoltaic modules is determined.

2. The new energy station operation and maintenance fault diagnosis method according to claim 1 is characterized in that: The fault analysis model is implemented by the following steps: Collect the electrical characteristic parameters and fault type classification results of sample PV modules, some of which are used as training samples and some as test samples; Initially set the parameters of the improved BP neural network model; The improved BP neural network model with initial parameters is used as the model to be trained, the electrical characteristic parameters in the training sample are used as input, and the fault type classification result is used as output to train the model to be trained, so as to obtain the fault analysis model to be verified; The fault analysis model to be verified is verified through the test sample. If the verification passes, the fault analysis model to be verified is used as the fault analysis model.

3. The new energy station operation and maintenance fault diagnosis method according to claim 2 is characterized in that: The improved BP neural network model is implemented by the following method: Step 1: Initialize the weights and thresholds of the BP neural network model, and set the initial parameters of the BP neural network and the ant colony algorithm; Step 2: Assign each ant individual an initial position and a corresponding initial velocity; Step 3: Calculate the fitness of each ant individual and update the global optimal fitness; Step 4: When the maximum number of iterations is greater than the first number of iterations, arrange the ant individuals in descending order of fitness and execute step 5. If it is less than the maximum number of iterations, return to step 3 and iterate in a loop; Step 5: Update the position of each ant and calculate the position center of the ant colony; Step 6, determine whether the number of iterations is greater than the second number of iterations. If not, go to step 6 and loop the calculation process. If satisfied, go to step 7; Step 7: Rank all ant individuals according to their fitness, select the top N ants as the initial chromosomes of the genetic algorithm, and perform encoding operations; Step 8: Select, inherit and mutate the chromosomes; Step 9: Calculate the fitness of all ant individuals. If the fitness meets the requirements, proceed to step 10, otherwise return to step 8 for a loop. Step 10: Complete the decoding of the chromosome, assign the optimal weights and thresholds to the BP neural network model, and obtain an improved BP neural network model.

4. The new energy station operation and maintenance fault diagnosis method according to claim 1 is characterized in that: The photovoltaic power generation prediction model corresponding to the grouping based on the training data includes: The training data set is input into the photovoltaic power generation prediction model for calculation, the performance of the photovoltaic power generation prediction model is evaluated, and the photovoltaic power generation prediction model with the optimal parameters is determined.

5. The new energy station operation and maintenance fault diagnosis method according to claim 4 is characterized in that: The step of inputting the training data set into the photovoltaic power prediction model for calculation comprises: Inputting the training data set into the first power prediction model to obtain an initial prediction value; Subtracting the initial predicted value from the actual power data to obtain a correction data set; The correction data set is input into the photovoltaic power error correction model to obtain the correction prediction value; The initial prediction value is superimposed with the corrected prediction value to obtain the corrected photovoltaic power generation power.

6. The new energy station operation and maintenance fault diagnosis method according to claim 5 is characterized in that: Inputting the training data set into the first power prediction model to obtain an initial prediction value comprises: The historical photovoltaic power data is decomposed using a decomposition algorithm to obtain multiple intrinsic mode components IMF and residual components RES; Use the training sets of each intrinsic mode component IMF and residual component RES to establish their own SVM regression models; The penalty parameter c and kernel function parameter g of each SVM regression model are optimized by using the sparrow search algorithm, and the prediction model of each intrinsic mode component IMF and residual component RES is obtained, and the test set is predicted at the same time; The photovoltaic power generation values ​​predicted by the prediction models of each intrinsic mode component IMF and the residual component RES are superimposed to obtain the initial prediction value.

7. The new energy station operation and maintenance fault diagnosis method according to claim 5 is characterized in that: The photovoltaic power error correction model adopts the following method: Initialize the algorithm parameters, including the number of ant colonies, number of iterations, initial positions and initial speeds of ants, and spatial dimensions; Construct the network topology and obtain the parameters that need to be optimized; The mean absolute error (MAE) is used as the fitness function, and the fitness value is calculated according to the initial position of the ant, so as to obtain the initial global optimal and individual optimal values ​​of the ant; Use the obtained fitness value to update the ant speed and position, calculate the fitness value again based on the updated ant position, and find the individual optimal position of the ant and the global optimal position; When the number of iterations reaches the maximum value, the iteration process ends, and the found hyperparameters are brought into the prediction model to predict the output results; if not, return to the previous step to continue the iteration.

8. The new energy station operation and maintenance fault diagnosis method according to any one of claims 1 to 7, characterized in that: The multiple photovoltaic power stations in the new energy station are clustered according to similarity to obtain multiple groupings including: Group each photovoltaic power station in the new energy site as a group; The distance between each two cluster groups is calculated using clustering algorithm; Calculating the correlation of each cluster grouping based on the output power of the photovoltaic power plants included in each cluster grouping and the meteorological parameters; The similarity between each two groupings is calculated based on the distance between each two groupings and the correlation between each two groupings; The two cluster groups with the highest similarity are merged as a new cluster group, and the step of calculating the distance between each two cluster groups by using the clustering algorithm is returned until the number of cluster groups reaches a preset number.

9. A new energy station operation and maintenance fault diagnosis device, characterized in that: The new energy station operation and maintenance fault diagnosis device includes a processor, a memory, and a new energy station operation and maintenance fault diagnosis program stored in the memory and executable by the processor, wherein when the new energy station operation and maintenance fault diagnosis program is executed by the processor, the steps of the new energy station operation and maintenance fault diagnosis method as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a new energy station operation and maintenance fault diagnosis program, wherein when the new energy station operation and maintenance fault diagnosis program is executed by a processor, the steps of the new energy station operation and maintenance fault diagnosis method as described in any one of claims 1 to 8 are implemented.

Images