Photovoltaic power prediction method and device for traction power supply system, terminal and medium

Abstract

The invention provides a photovoltaic power prediction method and device for a traction power supply system, a terminal and a medium, and the method comprises the steps: constructing a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long-short-term memory network, and training the CNN-BiLSTM photovoltaic power prediction model through employing the preprocessed historical performance data; introducing an adaptive weight coefficient, fusing a whale optimization algorithm and applying a Levy flight strategy to improve a dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm; and optimizing the hyper-parameters of the CNN-BiLSTM photovoltaic power prediction model by using the improved dung beetle optimization algorithm. The photovoltaic power prediction method and device for the traction power supply system, the terminal and the medium can be used for photovoltaic power prediction of the flexible direct-current traction power supply system, and the problems that an existing prediction model is insufficient in spatial-temporal feature extraction capacity, and a traditional optimization algorithm is prone to falling into local optimum are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic prediction, and in particular relates to a photovoltaic power prediction method, device, terminal and medium for a traction power supply system. Background Art

[0002] In a flexible DC traction power supply system, photovoltaic power prediction technology is the key to ensuring the stable operation of the system and optimizing energy management. Accurate photovoltaic power prediction can alleviate the local power supply tension and over-generation in the flexible DC traction power supply system, and improve the energy utilization efficiency. In the field of time series prediction, deep learning methods have received extensive attention due to their excellent self-learning ability and non-linear fitting ability. Among them, long short-term memory neural networks have been studied more in the field of time series prediction. Long short-term memory neural networks can make full use of historical data for learning and predict future time series, so they are more suitable for photovoltaic power prediction in flexible DC traction power supply systems.

[0003] However, it is difficult for long short-term memory neural networks to mine the coupling relationship between feature input sequences. The combination of convolutional neural networks and long short-term memory neural networks has advantages in extracting high-dimensional data feature information and processing time series, but there are still problems with the difficult selection of hyperparameters for the two network models. In the field of intelligent algorithms, existing heuristic optimization algorithms such as the dung beetle optimization algorithm, northern goshawk optimization algorithm, grey wolf optimization algorithm, sparrow optimization algorithm, whale optimization algorithm and particle swarm optimization algorithm are prone to falling into local minimum problems prematurely, which affects the accuracy of photovoltaic power prediction in flexible DC traction power supply systems. Summary of the Invention

[0004] In view of this, the present invention aims to propose a photovoltaic power prediction method, device, terminal and medium for a traction power supply system to solve the problem of poor accuracy of photovoltaic power prediction in existing flexible DC traction power supply systems.

[0005] To achieve the above object, the technical solution of the present invention is realized as follows: In a first aspect, the present invention provides a photovoltaic power prediction method for a traction power supply system, including: Obtaining historical performance data of a photovoltaic power generation module of a flexible DC traction power supply system and performing preprocessing; Constructing a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and training the CNN-BiLSTM photovoltaic power prediction model using the preprocessed historical performance data to fully mine the spatial features and time features of the historical performance data; Introduce an adaptive weight coefficient, fuse the whale optimization algorithm and apply the Lévy flight strategy to improve the dung beetle optimization algorithm, obtaining an improved dung beetle optimization algorithm; Use the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, so as to realize the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtain an optimized CNN-BiLSTM photovoltaic power prediction model; Use the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

[0006] Furthermore, the introducing an adaptive weight coefficient, fusing the whale optimization algorithm and applying the Lévy flight strategy to improve the dung beetle optimization algorithm, obtaining an improved dung beetle optimization algorithm, includes: Introduce an adaptive weight coefficient during the process of updating the position of the dung ball by the dung beetle in the dung beetle optimization algorithm; In the reproduction and foraging behaviors of the dung beetles in the dung beetle optimization algorithm, incorporate the spiral bubble net hunting strategy in the whale optimization algorithm; In the stealing behavior of the dung beetles in the dung beetle optimization algorithm, introduce the Lévy flight strategy to achieve a comprehensive search of the solvable area for stealing.

[0007] Furthermore, the introducing an adaptive weight coefficient during the process of updating the position of the dung ball by the dung beetle, includes: Initialize the parameters to be optimized by the dung beetles; among them, the parameters to be optimized by the dung beetles include the number of hidden nodes, the training period, and the initial learning rate of the CNN-BiLSTM photovoltaic power prediction model; Divide the dung beetle population size according to the ratio of 6:7:7:10 and generate the initial positions of the dung beetles; among them, the dung ball dung beetles = 1 / 5N, the reproductive dung beetles = 7 / 30N, the foraging dung beetles = 7 / 30N, the stealing dung beetles = 1 / 3N, and N is the size of the dung beetle population; Determine the initial positions of the dung beetles according to the initialization strategy. Before each position update, first detect whether the update of the dung ball position of the dung beetle encounters an obstacle; If no obstacle is encountered, perform position update according to the dung ball update formula of the dung beetle with the introduced adaptive weight coefficient. The formula is as follows: ; ; ;

[0008] In the above formula, is the adaptive weight coefficient, represents the only thet The position of the current iteration, represents the maximum number of iterations of the dung beetle population, represents the worst position of the global dung beetles; is the natural coefficient, taking a value of -1 or 1; is a constant representing the deflection coefficient, ∈(0, 0.2]; is a constant, a random number in (0, 1); If an obstacle is encountered, switch to the dancing behavior formula with an adaptive weight coefficient for position update. The formula is as follows: ;

[0009] In the above formula, is the adaptive weight coefficient, represents the position of the t th iteration of the ball-rolling dung beetle only, represents the deflection angle, , if equals 0, or the position of the dung beetle is not updated.

[0010] Furthermore, in the reproduction and foraging behaviors of the dung beetles in the dung beetle optimization algorithm, incorporate the spiral bubble net hunting strategy in the whale optimization algorithm, including: Determine the boundary selection strategy for the dung beetle spawning area. The formula is as follows: ; ; ;

[0011] In the above formula: is the position information of the t th iteration of the th brood ball, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local optimal position of the dung beetles; , represent the upper and lower bounds of the dung beetle reproduction respectively; represents the maximum number of iterations of the dung beetle population; , represent the upper and lower bounds of the optimization parameters; Introduce the spiral bubble net optimization method in the whale optimization algorithm into the dung beetle reproduction formula to perform a fine search around the current optimal solution in a spiral path to achieve a rapid approximation to the potential optimal solution in the local area. The formula is as follows: ;

[0012] In the above formula: is the position information of the t th brood ball at the th iteration, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local dung beetle optimal position; , represent the upper and lower bounds of the optimization parameters; is a constant defining the shape of the logarithmic spiral, is a random number in [-1, 1]; determines the boundary of the best foraging area of the dung beetle, and the formula is as follows: ;

[0013] In the above formula, represents the global dung beetle optimal position; , represent the upper and lower bounds of the dung beetle foraging area respectively; , represent the upper and lower bounds of the optimization parameters; determines the position update of the dung beetle, and the formula is as follows: ;

[0014] In the above formula, represents the position where the th foraging dung beetle is located at the t th iteration; is a random number following a normal distribution; is a random number in (0, 1); is a constant defining the shape of the logarithmic spiral, is a random number in [-1, 1]; , represent the upper and lower bounds of the optimization parameters.

[0015] Furthermore, in the stealing behavior of the dung beetle in the dung beetle optimization algorithm, the Lévy flight strategy is introduced to achieve a comprehensive search of the stealable solution area, including: The position update strategy of the stealing dung beetle focuses on local search near the current optimal solution, and the Lévy flight strategy is introduced into the stealing dung beetle formula, and the formula is as follows: ;

[0016] In the above formula, represents the Only steal the position where the dung beetle is in the t th iteration; represents the local optimal position of the dung beetle; is the global optimal position of the dung beetle; is a random vector of size 1xD that follows a normal distribution, where D represents the dimension of the optimization problem; is a constant with a value of 0.5; is the Lévy flight coefficient; and and are random numbers in the range (0, 1).

[0017] Furthermore, using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model to achieve the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model and obtain the optimized CNN-BiLSTM photovoltaic power prediction model, including: Using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, the CNN-BiLSTM prediction models corresponding to each dung beetle read the data and make predictions, and use the RMSE as the fitness function, and the formula is as follows: ;

[0018] In the above formula, and are respectively T the predicted photovoltaic power value and the true photovoltaic power value at time; represents the total number of test samples in the sample set, is the th dung beetle, is the fitness value of the dung beetle, RMSE is the fitness function; Based on the above formula, calculate the fitness values of each dung beetle and obtain the optimal fitness value of the dung beetle population and the corresponding individual positions; When the iteration number is reached, the dung beetle with the highest fitness value is the optimal hyperparameter of the CNN-BiLSTM prediction model; Substitute the optimal hyperparameters into the CNN-BiLSTM photovoltaic power prediction model to obtain the optimized CNN-BiLSTM photovoltaic power prediction model.

[0019] Furthermore, obtaining and preprocessing the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system includes: Obtaining the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system as the sample set; Normalize the historical performance data in the sample set using Min-max to obtain the preprocessed historical performance data.

[0020] In a second aspect, an embodiment of the present invention further provides a photovoltaic power prediction device for a traction power supply system, including: A processing module, configured to obtain the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and perform preprocessing; A construction module, configured to construct a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and use the preprocessed historical performance data to train the CNN-BiLSTM photovoltaic power prediction model to fully mine the spatial features and time features of the historical performance data; An improvement module, configured to introduce an adaptive weight coefficient, fuse the whale optimization algorithm and apply the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm; An optimization module, configured to use the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, so as to realize the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtain an optimized CNN-BiLSTM photovoltaic power prediction model; A prediction module, configured to use the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

[0021] In a third aspect, an embodiment of the present invention further provides a terminal, including: One or more processors; A storage device, configured to store one or more programs; A display, configured to display the prediction result; When the one or more programs are executed by the one or more processors, the one or more processors implement the photovoltaic power prediction method of the traction power supply system as described above.

[0022] In a fourth aspect, the present invention further provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute the photovoltaic power prediction method of the traction power supply system as described above when executed by a computer processor.

[0023] Compared with the prior art, the photovoltaic power prediction method, device, terminal and medium of the traction power supply system of the present invention have the following advantages: (1) The photovoltaic power prediction method, device, terminal and medium of the traction power supply system described in the present invention can be used for photovoltaic power prediction in a flexible DC traction power supply system, solving the problems of insufficient spatio-temporal feature extraction ability of existing prediction models and the tendency of traditional optimization algorithms to fall into local optima. By combining an improved dung beetle intelligent optimization algorithm, a convolutional neural network and a bidirectional long short-term memory network deep learning model, an efficient and accurate photovoltaic power prediction technology is provided for the flexible DC traction power supply system.

[0024] (2) The photovoltaic power prediction method, device, terminal and medium of the traction power supply system described in the present invention collect relevant historical data of photovoltaic power generation modules, including environmental factor data such as light intensity and temperature, and photovoltaic power output data, etc., and perform preprocessing operations such as data cleaning and normalization to improve data quality and lay a foundation for subsequent model training. Then, by constructing a CNN-BiLSTM photovoltaic power prediction model, the spatial correlation features of multi-dimensional data such as light and temperature are effectively extracted using CNN, and the power time series change law is captured in combination with BiLSTM, enabling accurate prediction of photovoltaic power.

[0025] (3) The photovoltaic power prediction method, device, terminal and medium of the traction power supply system described in the present invention optimize the number of hidden layer units and the initial learning rate of the CNN-BiLSTM photovoltaic power prediction model for the flexible DC traction power supply system by using an improved dung beetle intelligent optimization algorithm to solve the problem of model hyperparameter optimization. An optimized CNN-BiLSTM photovoltaic power prediction model is built according to the optimization results, which is beneficial to improving the accuracy and timeliness of photovoltaic power prediction results. In addition, by using the optimized CNN-BiLSTM photovoltaic power prediction model to achieve high-precision photovoltaic power prediction and applying the prediction results to the dynamic energy scheduling of the flexible DC traction power supply system, a reliable basis can be provided for the operation of the power supply system, which helps to improve the stability and economy of the power supply system operation. The present invention has significant technical advantages and application values in the field of photovoltaic power prediction, can effectively solve the key problems in the existing technology, and provides strong support for the efficient operation of the flexible DC traction power supply system.

[0026] (3)The photovoltaic power prediction method, device, terminal, and medium of the traction power supply system according to the present invention improve and optimize the traditional dung beetle algorithm. By introducing an adaptive weight coefficient during the update process of the dung beetle's ball-rolling position, it can be dynamically adjusted according to the execution situation of the dung beetle optimization algorithm. Initially, a larger activity range is given to the dung beetle to enhance the global search ability, and then it gradually decreases in the later stage, enabling it to perform fine searches in the local area, thereby improving the search efficiency. At the same time, in the reproduction and foraging behaviors of the dung beetle, the spiral bubble net hunting strategy in the whale optimization algorithm is incorporated, mimicking the spiral movement pattern of whales when hunting, to more effectively search in the solution space and enhance the global search ability of the algorithm, which helps to jump out of the local optimal solution. For the stealing dung beetle, the Lévy flight strategy is introduced. This random walking method has the characteristic of longer step-length jumps, which can further expand the search range and improve the global exploration ability of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings: Figure 1 It is a schematic flowchart of the photovoltaic power prediction method for the traction power supply system according to Embodiment 1 of the present invention; Figure 2 It is a schematic flowchart of the improved dung beetle optimization algorithm in the photovoltaic power prediction method for the traction power supply system according to Embodiment 1 of the present invention; Figure 3 It is a schematic flowchart of the training process of the CNN-BiLSTM photovoltaic power prediction model in the photovoltaic power prediction method for the traction power supply system according to Embodiment 1 of the present invention; Figure 4 It is a schematic diagram of the convergence curve of the benchmark test function in the photovoltaic power prediction method for the traction power supply system according to Embodiment 1 of the present invention; Figure 5 It is a schematic structural diagram of the photovoltaic power prediction device for the traction power supply system according to Embodiment 2 of the present invention; Figure 6 It is a schematic structural diagram of the photovoltaic power prediction terminal for the traction power supply system provided by Embodiment 3 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The present invention will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Additionally, it should be noted that for the sake of description, only the parts related to the present invention rather than all the structures are shown in the drawings.

[0029] Embodiment 1 Figure 1Schematic diagram of the process of photovoltaic power prediction method for the traction power supply system described in Embodiment 1 of the present invention. This embodiment can be used for photovoltaic power prediction of a flexible DC traction power supply system. Refer to Figure 1 , which specifically includes the following steps: Step 101: Obtain the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and perform preprocessing.

[0030] Existing prediction models have problems with insufficient spatio-temporal feature extraction ability. Therefore, in this embodiment, a CNN-BiLSTM photovoltaic power prediction model is built. To train the above prediction model, relevant historical data of the photovoltaic power generation module needs to be collected first, including environmental factor data such as light intensity and temperature, and photovoltaic power output data, and preprocessing operations such as data cleaning and normalization are performed on them to improve the data quality and lay a foundation for subsequent model training.

[0031] Since datasets with different dimensions need to be input in the network model during the process of photovoltaic power prediction of the flexible DC traction power supply system, in order to eliminate the influence of different dimensions of input feature sequences on the prediction accuracy and prediction speed, this embodiment uses Min-max to normalize the data, which specifically includes the following steps: Step 1011: Obtain the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and use it as a sample set.

[0032] Among them, the historical performance data includes the photovoltaic power output data of the photovoltaic module and its corresponding environmental factors. For example, the four environmental factors of photovoltaic module temperature, temperature, light intensity, and air pressure are used as four input features. Those skilled in the art can also adjust the number of input features according to actual needs, which will not be elaborated here. In addition, the sample set can also be divided into a training sample set and a test sample set to facilitate subsequent training and testing of the model.

[0033] Step 1012: Use Min-max to normalize the historical performance data in the sample set to obtain the preprocessed historical performance data. The formula is as follows: '

[0034] In the above formula, is the normalized data, is the multi-load data in the sample set, and are respectively 's maximum and minimum values.

[0035] In the actual application process, in this embodiment, by performing a linear transformation on the historical performance data, the numerical value can be mapped into the interval [0, 1], thereby eliminating the influence of data dimensions and making different indicators comparable.

[0036] Step 102: Construct a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and train the CNN-BiLSTM photovoltaic power prediction model using the preprocessed historical performance data to fully exploit the spatial and temporal features of the historical performance data.

[0037] This embodiment combines the advantages of a convolutional neural network (CNN) and a bidirectional long short-term memory network (BiLSTM), and can construct a more powerful and comprehensive prediction model. By mining data from both time and space perspectives, it realizes the efficient analysis and prediction of the time series data of the historical performance data of photovoltaic power generation modules. Among them, the CNN part is responsible for extracting the spatial features of the data. It identifies and compresses the local features in the input data through convolutional layers and pooling layers, thereby retaining key information and improving the generalization ability of the model. The BiLSTM focuses on the extraction of temporal features. It processes data in both forward and backward directions and can capture the dynamic temporal relationships in sequence data.

[0038] In the actual application process, by combining CNN and BiLSTM, their respective advantages can be fully utilized while compensating for each other's deficiencies. CNN can extract the spatial features in the input data and effectively extract the spatial correlation features of multi-dimensional data such as light and temperature, while BiLSTM can capture the temporal features in the input data and capture the power time series change rules, so as to achieve accurate prediction of photovoltaic power. Therefore, this embodiment constructs a CNN-BiLSTM photovoltaic power prediction model for a flexible DC traction power supply system by combining the two.

[0039] Exemplarily, the CNN-BiLSTM photovoltaic power prediction model includes: an input layer, a CNN layer, a BiLSTM layer, and an output layer; among them, the CNN layer includes a convolutional layer, a pooling layer, and a Flatten layer; the BiLSTM layer includes a forward LSTM network and a backward LSTM network.

[0040] Step 103: Introduce an adaptive weight coefficient, fuse the whale optimization algorithm, and apply the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm.

[0041] Since heuristic optimization algorithms such as the dung beetle optimization algorithm, the northern goshawk optimization algorithm, the grey wolf optimization algorithm, the sparrow optimization algorithm, the whale optimization algorithm, and the particle swarm optimization algorithm are prone to falling into local minima prematurely, traditional optimization algorithms have the problem of being prone to falling into local optima when performing hyperparameter optimization on the prediction model, which affects the prediction accuracy of the prediction model.

[0042] Therefore, to address the problem of optimizing model hyperparameters, this embodiment improves and optimizes the traditional dung beetle algorithm. Specifically, the improvement lies in introducing an adaptive weight coefficient during the update of the dung beetle's ball-rolling position. This coefficient can be dynamically adjusted according to the execution of the dung beetle optimization algorithm. Initially, a larger activity range is given to the dung beetle to enhance the global search ability, and it gradually decreases later, enabling it to perform fine searches in local areas, thereby improving the search efficiency. At the same time, in the reproduction and foraging behaviors of the dung beetle, the spiral bubble net hunting strategy in the whale optimization algorithm is incorporated, mimicking the spiral movement pattern of whales during predation, to more effectively search in the solution space, enhance the global search ability of the algorithm, and help jump out of local optimal solutions. For the kleptoparasitic dung beetle, the Lévy flight strategy is introduced. This random walk method has the characteristic of longer step-length jumps, which can further expand the search range and improve the global exploration ability of the algorithm.

[0043] Specifically, Figure 2 is a schematic flow diagram of the improved dung beetle optimization algorithm in the photovoltaic power prediction method for the traction power supply system described in Embodiment 1 of the present invention. Refer to Figure 2 This improved dung beetle optimization algorithm specifically includes the following steps: Step 1031: Introduce an adaptive weight coefficient during the update of the dung beetle's ball-rolling position in the dung beetle optimization algorithm.

[0044] Since the dung beetle's ball-rolling behavior can be divided into two types: obstacle-free mode and obstacle mode. When the dung beetle encounters an obstacle while rolling, it uses a dancing behavior to find a new rolling direction. Therefore, the initial position of the dung beetle can be determined according to the initialization strategy. Before each position update, first detect whether the dung beetle's ball-rolling position update encounters an obstacle.

[0045] First, initialize the parameters to be optimized by the dung beetle; among them, the parameters to be optimized by the dung beetle include the number of hidden nodes of the CNN-BiLSTM photovoltaic power prediction model, the training period, and the initial learning rate.

[0046] Secondly, divide the dung beetle population size in the ratio of 6:7:7:10 and generate the initial positions of the dung beetles; among them, the ball-rolling dung beetles = 1 / 5N, the reproductive dung beetles = 7 / 30N, the foraging dung beetles = 7 / 30N, the kleptoparasitic dung beetles = 1 / 3N, and N is the size of the dung beetle population.

[0047] Finally, determine the initial position of the dung beetle according to the initialization strategy. Before each position update, first detect whether the dung beetle's ball-rolling position update encounters an obstacle; If no obstacle is encountered, perform position update according to the dung beetle ball-rolling update formula with the introduced adaptive weight coefficient. The formula is as follows: ; ; ;

[0048] In the above formula, is the adaptive weight coefficient, represents the position of the t th iteration of the th dung beetle rolling a ball, represents the maximum number of iterations of the dung beetle population, represents the worst position of the global dung beetles; is the natural coefficient, taking a value of -1 or 1; is a constant representing the deflection coefficient, ∈(0, 0.2]; is a constant, a random number in (0,1). In this embodiment

[0049] If an obstacle is encountered, switch to the dancing behavior formula introducing the adaptive weight coefficient for position update. The formula is as follows: ;

[0050] In the above formula, is the adaptive weight coefficient, represents the position of the t th iteration of the th dung beetle rolling a ball, represents the deflection angle, if or is equal to 0, the position of the dung beetle is not updated.

[0051] In the actual application process, in this embodiment, the adaptive weight coefficient can be dynamically adjusted according to the execution situation of the dung beetle optimization algorithm, and can be dynamically adjusted according to the preset weight coefficient adjustment strategy and the current number of iterations . Specifically, as the iteration progresses, the adaptive weight coefficient gradually decreases, so that the improved dung beetle optimization algorithm has a strong global search ability in the early stage and gradually enhances the local search ability in the later stage.

[0052] Step 1032, in the reproduction and foraging behaviors of the dung beetles in the dung beetle optimization algorithm, incorporate the spiral bubble net hunting strategy in the whale optimization algorithm.

[0053] Since the spiral bubble net hunting strategy in the whale optimization algorithm mimics the spiral movement pattern of whales during predation, it can search more effectively in the solution space and improve the global search ability of the algorithm. Therefore, in this embodiment, the whale optimization algorithm is incorporated into the dung beetle optimization algorithm to further enhance the global search ability of the dung beetle optimization algorithm.

[0054] First, determine the boundary selection strategy for the dung beetle spawning area, and the formula is as follows: ; ; ;

[0055] In the above formula: is the position information of the t th iteration of the st brood ball, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local dung beetle optimal position; , respectively represent the upper and lower bounds of dung beetle reproduction; represents the maximum number of iterations of the dung beetle population; , represent the upper and lower bounds of the optimization parameters.

[0056] Secondly, introduce the spiral bubble net optimization method in the whale optimization algorithm into the dung beetle reproduction formula to perform a fine search around the current optimal solution in a spiral path, so as to quickly approach the potential optimal solution in the local area. The formula is as follows: ;

[0057] In the above formula: is the position information of the t th iteration of the st brood ball, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local dung beetle optimal position; , represent the upper and lower bounds of the optimization parameters; is a constant defining the logarithmic spiral shape, is a random number in [-1, 1].

[0058] Thirdly, determine the boundary of the best foraging area of the dung beetle, and the formula is as follows: ;

[0059] In the above formula, represents the global optimal position of the dung beetle; and represent the upper and lower bounds of the foraging area of the dung beetle respectively; and represent the upper and lower bounds of the optimization parameters.

[0060] Finally, determine the position update of the dung beetle, and the formula is as follows: ;

[0061] In the above formula, represents the position of the rd foraging dung beetle at the t th iteration; is a random number obeying the normal distribution; is a random number in (0, 1); is a constant defining the shape of the logarithmic spiral, is a random number in [-1, 1]; and represent the upper and lower bounds of the optimization parameters.

[0062] In the actual application process, in this embodiment, by introducing the spiral bubble net optimization method in the whale optimization algorithm into the breeding dung beetle formula, a fine search is carried out around the current optimal solution in a spiral path, and it is also possible to quickly approach the potential optimal solution in the local area.

[0063] Step 1033, in the stealing behavior of the dung beetle in the dung beetle optimization algorithm, introduce the Lévy flight strategy to achieve a comprehensive search of the solvable area of stealing.

[0064] Since the position update strategy of the stealing dung beetle focuses on local search near the current optimal solution, and the Lévy flight is a random search strategy that realizes efficient exploration of the solution space by simulating the foraging behavior of animals. The introduction of this strategy enables the algorithm to better conduct a comprehensive search of the solvable area, improving the global search ability of the algorithm and the ability to jump out of the local optimal solution. Therefore, in this embodiment, by introducing the Lévy flight strategy into the stealing dung beetle formula, the algorithm can better conduct a comprehensive search of the solvable area of stealing.

[0065] Specifically, focus the position update strategy of the stealing dung beetle on local search near the current optimal solution, and introduce the Lévy flight strategy into the stealing dung beetle formula. The formula is as follows: ; ;

[0066] The position update strategy of the stealing dung beetle focuses on local search near the current optimal solution. The Lévy flight strategy is introduced into the stealing dung beetle formula, and the formula is as follows: ;

[0067] In the above formula, represents the position where the th iteration of the stealing dung beetle is located; t represents the local optimal position of the dung beetle; is the global optimal position of the dung beetle; is a random vector of size 1xD that follows a normal distribution, where D represents the dimension of the optimization problem; is a constant with a value of 0.5; is the Lévy flight coefficient; is the Lévy flight coefficient; , , are random numbers in the range (0, 1).

[0068] Although there are cases in the prior art where an optimization algorithm is combined with a neural network, the main improvement point of this embodiment lies in improving the dung beetle optimization algorithm through the introduction of an adaptive weight coefficient, the fusion of the whale optimization algorithm, and the application of the Lévy flight strategy. These improvements make the dung beetle optimization algorithm have significant advantages in global search ability and jumping out of local optimal solutions.

[0069] In addition, the prediction method described in this embodiment specifically combines the innovation of the application scenario, that is, it is applied to a flexible DC traction power supply system, which can effectively improve the prediction accuracy. This case is aimed at the photovoltaic power prediction of a flexible DC traction power supply system, which is an application scenario with specific requirements and challenges. The prediction method described in this embodiment addresses the problems existing in photovoltaic power prediction, realizes optimization through innovative strategies, simulates the foraging behavior of animals to efficiently explore the solution space, enhances the global search ability and the ability to jump out of local optimal solutions, and solves the defect that traditional optimization algorithms are prone to fall into local optimal problems when applied to photovoltaic power prediction models, which is beneficial to improving the accuracy of photovoltaic power prediction.

[0070] Step 104: Use the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, so as to realize the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtain an optimized CNN-BiLSTM photovoltaic power prediction model.

[0071] Figure 3 This is a schematic diagram of the training process of the CNN-BiLSTM photovoltaic power prediction model in the photovoltaic power prediction method of the traction power supply system according to Embodiment 1 of the present invention. SeeFigure 3 。

[0072] First, use the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model. The CNN-BiLSTM prediction models corresponding to each dung beetle read the data and make predictions. The RMSE of the prediction results is used as the fitness function, and the formula is as follows:

[0073] In the above formula, and are respectively T the predicted value of photovoltaic power and the true value of photovoltaic power at time represents the total number of test samples in the sample set, is the rd dung beetle, is the fitness value of the dung beetle, RMSE is the fitness function.

[0074] Secondly, based on the above formula, calculate the fitness value of each dung beetle, that is , and obtain the optimal fitness value of the dung beetle population and the corresponding individual position; when the iteration number is reached, the dung beetle with the highest fitness value is the optimal hyperparameter of the CNN-BiLSTM prediction model.

[0075] Finally, substitute the optimal hyperparameters into the CNN-BiLSTM photovoltaic power prediction model to obtain an optimized CNN-BiLSTM photovoltaic power prediction model.

[0076] The improved dung beetle optimization algorithm (IDBO) can be compared with the dung beetle optimization algorithm (DBO), northern goshawk optimization algorithm (NGO), grey wolf optimization algorithm (GWO), sparrow search algorithm (SSA), whale optimization algorithm (WOA) and particle swarm optimization algorithm (PSO) based on existing experimental methods, and the F1, F2 (unimodal functions) and F3, F4 (multimodal functions) in common benchmark test functions are used to evaluate the search ability.

[0077] Specifically, the experimental population size can be set to 30 and the maximum number of iterations to 500. Each algorithm runs independently 30 times, and the average value and standard deviation of the optimal values are calculated. Among them, the formulas of each benchmark test function are as follows: , dimension 30, range [-100, 100]; , dimension 30, range [-10, 10]; , dimension 30, range [-500, 500]; , Dimension 30, range [-32, 32].

[0078] Figure 4 It is a schematic diagram of the convergence curve of the benchmark test function in the photovoltaic power prediction method of the traction power supply system described in Embodiment 1 of the present invention. See Figure 4 , The experimental results show that IDBO is significantly superior to the other six algorithms in terms of convergence speed and stability. As can be seen from the above, the improved dung beetle optimization algorithm in this embodiment has significant advantages in global search ability and jumping out of local optimal solutions. Therefore, through experimental verification, the improved dung beetle optimization algorithm (DBO) combined with bidirectional long short-term memory network (BiLSTM) shows significant performance improvement in the photovoltaic power prediction task. Compared with traditional optimization methods, it has higher prediction accuracy and faster convergence speed.

[0079] In the actual application process, by using the above improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, the improved dung beetle optimization algorithm can show stronger search ability and faster convergence speed when solving the hyperparameter optimization problem of the CNN-BiLSTM photovoltaic power prediction model, and shows significant performance improvement in the photovoltaic power prediction task, with higher prediction accuracy and faster convergence speed compared with the existing technology. Therefore, constructing a CNN-BiLSTM photovoltaic power prediction model with hyperparameters obtained by the improved dung beetle optimization algorithm can more accurately predict the output power of the photovoltaic system, thereby providing a more reliable energy management strategy for the flexible DC traction power supply system.

[0080] Step 105: Use the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

[0081] In the actual application process, using the optimized CNN-BiLSTM model to achieve high-precision photovoltaic power prediction and applying the prediction results to the dynamic energy scheduling of the flexible DC traction power supply system can provide a reliable basis for system operation and help improve the stability and economy of system operation.

[0082] The prediction method described in this embodiment has significant technical advantages and application values in the field of photovoltaic power prediction, can effectively solve the key problems in the existing technology, and provides strong support for the efficient operation of the flexible DC traction power supply system.

[0083] Embodiment 2 Figure 5 It is a schematic diagram of the structure of the photovoltaic power prediction device of the traction power supply system described in Embodiment 2 of the present invention. Figure 5 A block diagram of an exemplary apparatus suitable for use in implementing embodiments of the present invention is shown. Figure 5The displayed device is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present invention. For example, Figure 5 As shown, this prediction device includes: A processing module 201, configured to obtain historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and perform preprocessing.

[0084] A construction module 202, configured to construct a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and use the preprocessed historical performance data to train the CNN-BiLSTM photovoltaic power prediction model, so as to fully mine the spatial features and time features of the historical performance data.

[0085] An improvement module 203, configured to introduce an adaptive weight coefficient, fuse the whale optimization algorithm, and apply the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm.

[0086] An optimization module 204, configured to use the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, so as to realize the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtain an optimized CNN-BiLSTM photovoltaic power prediction model.

[0087] A prediction module 205, configured to use the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

[0088] For the photovoltaic power prediction device described in this embodiment, by constructing a CNN-BiLSTM photovoltaic power prediction model, the spatial features and time features of the historical performance data of the photovoltaic power generation module can be fully mined. At the same time, by using the improved dung beetle optimization algorithm to optimize the photovoltaic power prediction model, high-precision photovoltaic power prediction can be realized, providing a reliable basis for the operation of the traction power supply system, and helping to improve the stability and economy of the operation of the traction power supply system.

[0089] Based on the above embodiments, the processing module further includes: A data acquisition unit, configured to obtain historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and use it as a sample set.

[0090] A normalization unit, configured to perform normalization processing on the historical performance data in the sample set by using Min-max to obtain preprocessed historical performance data.

[0091] Based on the above embodiments, the improvement module further includes: An update unit is used to introduce an adaptive weight coefficient during the process of updating the dung beetle rolling ball position in the dung beetle optimization algorithm.

[0092] An incorporation unit is used to incorporate the spiral bubble net hunting strategy in the whale optimization algorithm into the reproduction and foraging behaviors of dung beetles in the dung beetle optimization algorithm.

[0093] An introduction unit is used to introduce the Lévy flight strategy in the stealing behavior of dung beetles in the dung beetle optimization algorithm to achieve a comprehensive search for the solvable area of stealing.

[0094] The photovoltaic power prediction device of the traction power supply system provided by the embodiments of the present invention can execute the photovoltaic power prediction method of the traction power supply system provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0095] Embodiment III Figure 6 It is a schematic structural diagram of a photovoltaic power prediction terminal of a traction power supply system provided by Embodiment III of the present invention; Figure 6 It shows a block diagram of an exemplary terminal suitable for implementing the embodiments of the present invention. Figure 6 The displayed terminal is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0096] As Figure 6 shown, the terminal 12 is presented in the form of a general computing device. The components of the terminal 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0097] The bus 18 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the multiple bus structures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0098] The terminal 12 typically includes a variety of computer system readable media. These media can be any available media accessible by the terminal 12, including volatile and non - volatile media, removable and non - removable media.

[0099] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Terminal 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 can be used for reading and writing on non-removable, non-volatile magnetic media ( Figure 6 not shown, commonly referred to as a "hard disk drive"). Although Figure 6 not shown in the figure, a disk drive for reading and writing on removable non-volatile disks (such as "floppy disks"), and an optical disk drive for reading and writing on removable non-volatile optical disks (such as CD-ROM, DVD-ROM or other optical media) can be provided. In these cases, each drive can be connected to bus 18 through one or more data media interfaces. Memory 28 may include at least one program product having a set (such as at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0100] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. Program modules 42 generally perform the functions and / or methods in the embodiments described in the present invention.

[0101] Terminal 12 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the terminal 12, and / or communicate with any device that enables the terminal 12 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through an input / output (I / O) interface 22. Also, terminal 12 can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN) and / or a public network, such as the Internet) through a network adapter 20. As shown in the figure, network adapter 20 communicates with other modules of terminal 12 through bus 18. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with terminal 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0102] Processing unit 16 executes various functional applications and data processing by running programs stored in system memory 28, such as implementing the photovoltaic power prediction method of the traction power supply system provided by the embodiments of the present invention.

[0103] Example 4 Example 4 of the present invention further provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute the photovoltaic power prediction method of any traction power supply system provided in the above embodiments when executed by a computer processor.

[0104] The computer storage medium of the embodiments of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection with one or more wires, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device.

[0105] The computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries the computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and this computer-readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0106] The program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0107] Computer program code for performing the operations of the present invention may be written in one or more programming languages or combinations thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0108] Note that the above is only a preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments may be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A photovoltaic power prediction method for a traction power supply system, characterized in that, Including: Obtaining historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and performing preprocessing; Constructing a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and training the CNN-BiLSTM photovoltaic power prediction model using the preprocessed historical performance data to fully mine the spatial and temporal features of the historical performance data; Introducing an adaptive weight coefficient, integrating the whale optimization algorithm, and applying the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm; Using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model to achieve the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtaining an optimized CNN-BiLSTM photovoltaic power prediction model; Using the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

2. The method according to claim 1, wherein The introducing an adaptive weight coefficient, integrating the whale optimization algorithm, and applying the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm includes: Introducing an adaptive weight coefficient during the process of updating the position of the dung ball of the dung beetle optimization algorithm; Incorporating the spiral bubble net hunting strategy in the whale optimization algorithm into the reproduction and foraging behaviors of the dung beetle in the dung beetle optimization algorithm; Introducing the Lévy flight strategy in the stealing behavior of the dung beetle in the dung beetle optimization algorithm to achieve a comprehensive search of the solvable area of stealing.

3. The method according to claim 2, wherein The introducing an adaptive weight coefficient during the process of updating the position of the dung ball of the dung beetle optimization algorithm includes: Initializing the parameters to be optimized by the dung beetle; among them, the parameters to be optimized by the dung beetle include the number of hidden nodes, the training cycle, and the initial learning rate of the CNN-BiLSTM photovoltaic power prediction model; Dividing the dung beetle population size according to the ratio of 6:7:7:10 and generating the initial position of the dung beetle; among them, the dung ball dung beetle = 1 / 5N, the reproductive dung beetle = 7 / 30N, the foraging dung beetle = 7 / 30N, the stealing dung beetle = 1 / 3N, and N is the size of the dung beetle population; Determining the initial position of the dung beetle according to the initialization strategy, and detecting whether the position update of the dung ball of the dung beetle encounters an obstacle before each position update; If no obstacle is encountered, the position is updated according to the dung ball update formula with an introduced adaptive weight coefficient, and the formula is as follows: ; ; ; In the above formula, is the adaptive weight coefficient, represents the position of the t th iteration of the only rolling dung beetle, represents the maximum number of iterations of the dung beetle population, represents the worst position of the global dung beetle; is the natural coefficient, taking a value of -1 or 1; represents the constant of the deflection coefficient, ∈(0, 0.2]; represents a constant, a random number in (0,1); If an obstacle is encountered, switch to the dancing behavior formula with an introduced adaptive weight coefficient for position update, and the formula is as follows: ; In the above formula, is the adaptive weight coefficient, represents the position of the t th iteration of the only rolling dung beetle, represents the deflection angle, if or is equal to 0, the position of the dung beetle is not updated.

4. The method according to claim 3, characterized in that The incorporating the spiral bubble net hunting strategy in the whale optimization algorithm into the reproduction and foraging behaviors of the dung beetle in the dung beetle optimization algorithm includes: Determining the boundary selection strategy of the dung beetle spawning area, and the formula is as follows: ; ; ; In the above formula: is the position information of the t th brooding ball at the th iteration, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local optimal position of the dung beetle; , represent the upper and lower bounds of dung beetle reproduction respectively; represents the maximum number of iterations of the dung beetle population; , represent the upper and lower bounds of the optimization parameters; Introducing the spiral bubble net optimization method in the whale optimization algorithm into the formula of the reproductive dung beetle to perform a fine search around the current optimal solution in a spiral path to achieve a rapid approximation to the potential optimal solution in the local area, and the formula is as follows: ; In the above formula: is the position information of the t th brooding ball at the th iteration, and represent two independent random vectors of size 1xD, where D represents the dimension of the optimization problem; represents the local optimal position of the dung beetle; , represent the upper and lower bounds of the optimization parameters; is a constant defining the shape of the logarithmic spiral, is a random number in [-1, 1]; Determining the boundary of the best foraging area of the dung beetle, and the formula is as follows: ; In the above formula, represents the global optimal position of the dung beetle; and represent the upper and lower bounds of the foraging area of the dung beetle, respectively; and represent the upper and lower bounds of the optimization parameters; Determining the position update of the dung beetle, and the formula is as follows: ; In the above formula, represents the position where the th dung beetle forages only at the t th iteration; is a random number that follows a normal distribution; is a random number in (0, 1); is a constant that defines the shape of the logarithmic spiral, is a random number in [-1, 1]; , represent the upper and lower bounds of the optimization parameters.

5. The method according to claim 4, wherein In the stealing behavior of the dung beetle in the dung beetle optimization algorithm, a Lévy flight strategy is introduced to achieve a comprehensive search of the stealable solution area, including: The position update strategy of the stealing dung beetle focuses on local search near the current optimal solution, and the Lévy flight strategy is introduced into the stealing dung beetle formula. The formula is as follows: ; In the above formula, represents the position where the th iteration of only stealing dung beetles is located; t ; represents the optimal position of the local dung beetle; is the optimal position of the global dung beetle; is a random vector of size 1xD that follows a normal distribution, where D represents the dimension of the optimization problem; is a constant with a value of 0.5; is the Lévy flight coefficient; , , are random numbers in the range (0, 1).

6. The method according to claim 1, characterized in that, Using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model to achieve the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtaining the optimized CNN-BiLSTM photovoltaic power prediction model, including: Using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, the CNN-BiLSTM prediction models corresponding to each dung beetle read data and make predictions, and the RMSE is used as the fitness function, and the formula is as follows: ; In the above formula, and are respectively T the predicted value and the true value of the photovoltaic power at a certain moment; represents the total number of test samples in the sample set, is the th dung beetle, the fitness value of the dung beetle, RMSE is the fitness function; Based on the above formula, calculate the fitness values of each dung beetle, and obtain the optimal fitness value of the dung beetle population and the corresponding individual position; When the number of iterations is reached, the dung beetle with the highest fitness value is the optimal hyperparameter of the CNN-BiLSTM prediction model; Substitute the optimal hyperparameters into the CNN-BiLSTM photovoltaic power prediction model to obtain the optimized CNN-BiLSTM photovoltaic power prediction model.

7. The method according to claim 1, characterized in that Obtaining and preprocessing the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system, including: Obtain the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system and use it as a sample set; Use Min-max to normalize the historical performance data in the sample set to obtain the preprocessed historical performance data.

8. A photovoltaic power prediction device for a traction power supply system, characterized in that, Including: A processing module for obtaining and preprocessing the historical performance data of the photovoltaic power generation module of the flexible DC traction power supply system; A construction module for constructing a CNN-BiLSTM photovoltaic power prediction model based on a convolutional neural network and a bidirectional long short-term memory network, and training the CNN-BiLSTM photovoltaic power prediction model using the preprocessed historical performance data to fully mine the spatial and temporal features of the historical performance data; An improvement module for introducing an adaptive weight coefficient, fusing the whale optimization algorithm and applying the Lévy flight strategy to improve the dung beetle optimization algorithm to obtain an improved dung beetle optimization algorithm; An optimization module for using the improved dung beetle optimization algorithm to optimize the hyperparameters of the CNN-BiLSTM photovoltaic power prediction model to achieve the adaptive selection of the optimal hyperparameters of the CNN-BiLSTM photovoltaic power prediction model, and obtaining the optimized CNN-BiLSTM photovoltaic power prediction model; A prediction module for using the optimized CNN-BiLSTM photovoltaic power prediction model to predict the photovoltaic power of the photovoltaic power generation module of the flexible DC traction power supply system.

9. A terminal, including: One or more processors; A storage device for storing one or more programs; A display for displaying the prediction results; When the one or more programs are executed by the one or more processors, the one or more processors implement the photovoltaic power prediction method of the traction power supply system according to any one of claims 1-7.

10. A storage medium containing computer-executable instructions, which are used to execute the photovoltaic power prediction method of the traction power supply system as described in any one of claims 1-7 when executed by a computer processor.

Images