Agricultural light complementary operation and maintenance method, device and equipment for desertification land photovoltaic power station and medium

Abstract

The invention belongs to the technical field of photovoltaic operation and maintenance, and particularly relates to an agricultural and photovoltaic complementary operation and maintenance method, device, equipment and medium for a desertification land photovoltaic power station, and the method comprises the steps: processing image data of a target region through employing a pre-trained image recognition CNN neural network, accurately recognizing a desertification land region, and taking the region as a site selection basis for the construction of the photovoltaic power station. The state of a photovoltaic power station, the growth state of crops and the external environment condition are obtained in real time, and collected data are input into a pre-trained DQN (deep Q network) neural network model. The DQN model outputs a corresponding Q value for each possible operation and maintenance strategy by analyzing the data, and the operation and maintenance strategy corresponding to the optimal Q value is determined as the current optimal operation and maintenance strategy according to the Q values output by the DQN model, so that the comprehensive optimization of the power generation efficiency, the soil improvement effect and the crop yield of the photovoltaic power station is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic operation and maintenance, and specifically relates to an agricultural-photovoltaic complementary operation and maintenance method, device, equipment and medium for a photovoltaic power station in desertified land. Background Art

[0002] In the field of photovoltaic power generation, the operation and maintenance management strategies of onshore photovoltaic power stations have become increasingly mature, but the operation and maintenance management strategies of photovoltaic power stations on desertified land involving agricultural photovoltaic complementarity and soil improvement are still in urgent need of exploration. Take the contiguous desertified land in Zhangwu County, Fuxin City, Liaoning Province as an example. Before the installation of photovoltaic power stations, crops generally suffered from poor harvests and the soil gradually became desertified. With the installation of photovoltaic power stations, the soil moisture content no longer decreases, and how to combine photovoltaic power stations with soil improvement and crop planting to form agricultural photovoltaic complementarity has put forward higher requirements for the operation and maintenance management strategies of photovoltaic power stations. In the existing technology, the operation and maintenance strategies of onshore photovoltaic power stations cannot be directly applied to agricultural photovoltaic complementary photovoltaic power stations in desertified land and other areas. Summary of the invention

[0003] The purpose of the present invention is to provide a method, device, equipment and medium for operation and maintenance of photovoltaic power stations in desertified land to solve the problem in the prior art that the operation and maintenance strategy of photovoltaic power stations on land is not suitable for photovoltaic power stations in desertified land.

[0004] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides an operation and maintenance method for agricultural-photovoltaic complementary power station in desertified land, comprising: Acquire the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, each of which includes a plurality of photovoltaic modules with adjustable inclination angles; The photovoltaic power station status, crop growth status and external environmental conditions are input into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The operation and maintenance strategy corresponding to the optimal Q value is determined as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

[0005] Furthermore, the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; wherein the image recognition CNN neural network is trained in the following manner: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

[0006] Furthermore, the photovoltaic power station status, crop growth status and external environmental conditions are input into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; wherein the DQN neural network model is trained in the following manner: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

[0007] Furthermore, different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization objective in the overall benefit.

[0008] In a second aspect, the present invention provides an agricultural-photovoltaic complementary operation and maintenance device for a photovoltaic power station on desertified land, comprising: The input data acquisition module is used to obtain the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, and each photovoltaic array includes a plurality of photovoltaic modules with adjustable inclination angles; A strategy optimization module is used to input the status of the photovoltaic power station, the growth status of crops and the external environmental conditions into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The strategy determination module is used to determine the operation and maintenance strategy corresponding to the optimal Q value as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

[0009] Furthermore, in the input data acquisition module, the image recognition CNN neural network is trained as follows: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

[0010] Furthermore, in the strategy optimization module, the DQN neural network model is trained as follows: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

[0011] Furthermore, different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization objective in the overall benefit.

[0012] In a third aspect of the present invention, an electronic device is provided, comprising a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the above-mentioned agricultural-photovoltaic complementary operation and maintenance method of a photovoltaic power station on desertified land.

[0013] In a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the above-mentioned agricultural-photovoltaic complementary operation and maintenance method of a photovoltaic power station on desertified land is implemented.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: In the field of photovoltaic power generation, although the operation and maintenance management strategies of onshore photovoltaic power stations are relatively mature, there are still significant technical gaps for photovoltaic power stations on desertified land, especially the operation and maintenance management strategies that combine agricultural and photovoltaic complementarity and soil improvement. In order to solve the above problems, the present invention proposes an agricultural and photovoltaic complementarity operation and maintenance method for photovoltaic power stations on desertified land. The image data of the target area is processed by using a pre-trained image recognition CNN (convolutional neural network) neural network to accurately identify the desertified land area as the site selection basis for the construction of photovoltaic power stations. The status of the photovoltaic power station, the growth status of crops and the external environmental conditions are obtained in real time, and the collected data are input into the pre-trained DQN (deep Q network) neural network model. The DQN model analyzes these data and outputs a corresponding Q value for each possible operation and maintenance strategy. According to the Q value output by the DQN model, the operation and maintenance strategy corresponding to the optimal Q value is determined as the current optimal operation and maintenance strategy, achieving the comprehensive optimization of photovoltaic power station power generation efficiency, soil improvement effect and crop yield. This scheme accurately identifies the desertified land area through the image recognition CNN neural network, and customizes the photovoltaic power station construction plan based on this, ensuring that the construction of the photovoltaic power station can adapt to the special environment of desertified land. The DQN neural network model is used to input multivariate data such as the status of photovoltaic power stations, the growth status of crops, and external environmental conditions. Through intelligent analysis, the optimal operation and maintenance strategy is provided for the agricultural-photovoltaic complementary system, which takes into account the power generation efficiency of photovoltaic power stations, the growth needs of crops, and the soil improvement effect, thus realizing the intelligent management of agricultural-photovoltaic complementary in the true sense. The problem that the operation and maintenance strategy of traditional onshore photovoltaic power stations is not applicable to direct application on desertified land is solved. The agricultural-photovoltaic complementary operation and maintenance device, electronic device, and computer-readable storage medium of a photovoltaic power station on desertified land provided by the present invention also solve the problems raised in the background technology section.

[0015] Through the operation and maintenance strategy, the power generation of photovoltaic power stations, crop yields and soil improvement effects have been tripled. The stable operation of photovoltaic power stations has increased the supply of clean energy, the increase in crop yields has driven the development of the agricultural economy, and soil improvement is conducive to the restoration and protection of the ecological environment. The three promote each other, forming a virtuous cycle and improving the overall benefits of the system.

[0016] Compared with traditional manual inspections and empirical judgments, the operation and maintenance strategy provided by this solution can monitor and analyze data in real time, discover and solve problems in a timely manner, reduce the need for manual intervention, and thus reduce operation and maintenance costs.

[0017] In summary, the agricultural-photovoltaic complementary operation and maintenance method of the desertified land photovoltaic power station of the present invention, through intelligent identification and decision-making technology, effectively solves the problem that the operation and maintenance strategy of onshore photovoltaic power stations in the prior art is not suitable for agricultural-photovoltaic complementary photovoltaic power stations in desertified land, and realizes the coordinated development of photovoltaic power stations, crop planting and soil improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 This is a flow chart of an agricultural-photovoltaic complementary operation and maintenance method for a photovoltaic power station on desertified land according to an embodiment of the present invention; Figure 2 This is a structural block diagram of an agricultural-photovoltaic complementary operation and maintenance device for a photovoltaic power station in desertified land according to an embodiment of the present invention; Figure 3 The present invention is a structural block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0020] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0021] Example 1 like Figure 1 As shown, a method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land comprises: S1. Obtaining the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, each of which includes a plurality of photovoltaic modules with adjustable inclination angles; Specifically, the image recognition CNN neural network is trained as follows: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

[0022] S2, inputting the photovoltaic power station status, crop growth status and external environmental conditions into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; Specifically, the DQN neural network model is trained as follows: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

[0023] Preferably, different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization target in the overall benefit.

[0024] In a more specific embodiment, when training the DQN (Deep Q-Networks) neural network, the goal is to maximize the benefits of photovoltaic power stations on desertified land, crop planting, and soil improvement.

[0025] The following is the detailed training method of the DQN neural network: (1) Data preparation State space definition: Define the state space of DQN. The state includes multiple indicators such as the current month, the inclination of the photovoltaic module, soil moisture content, soil temperature, soil pH, etc., which serve as the input of the DQN neural network.

[0026] Action space definition: The action space includes all PV inspection methods, crop planting methods, and PV module tilt adjustment strategies. Each action corresponds to a specific PV inspection method, crop planting decision, or PV module tilt adjustment.

[0027] Reward function design: The reward function reflects power generation, crop yield, and soil improvement effects. Trade-offs can be made based on actual business needs. For example, higher rewards can be given to state-action pairs with high power generation, high crop yield, and good soil improvement effects.

[0028] More specifically, the state space definition is: in the scenario of photovoltaic power station in desertified land, the state space includes all the key factors that affect crop planting, soil improvement and photovoltaic power generation. For example: Current month: Climate conditions in different months (such as temperature, humidity, and duration of sunshine) will affect the growth cycle of crops and the power generation efficiency of photovoltaic modules.

[0029] The inclination of photovoltaic modules: The inclination determines the angle at which the photovoltaic panels receive solar radiation, thus affecting the power generation efficiency. At the same time, the inclination also affects the light conditions and temperature distribution of the soil below.

[0030] Soil moisture content: The moisture content in the soil is crucial to the growth of crops, and it also affects the physical properties of the soil and the foundation stability of photovoltaic panels.

[0031] Soil temperature: Soil temperature affects the growth rate of crops, as well as the activity of microorganisms and the release of nutrients in the soil.

[0032] Soil acidity (pH): Soil directly affects the growth of crops and soil fertility, and needs to be properly adjusted to suit the growth needs of specific crops.

[0033] The above indicators are encoded into numerical or categorical data and used as input features of the DQN neural network.

[0034] In practical applications, these data are also normalized or standardized to ensure the stability and convergence of neural network training.

[0035] More specifically, the action space is defined as The action space includes all photovoltaic inspection methods, crop planting methods and photovoltaic module tilt adjustment strategies. These actions can be concretized into the following forms: Photovoltaic inspection methods: including inspection routes, inspection times, etc.

[0036] Crop planting decisions: including choosing which crops to plant (such as wheat, corn, vegetables, etc.), planting density, planting time, etc.

[0037] PV panel tilt adjustment: According to seasonal changes, changes in solar altitude, and the needs of soil improvement and crop growth, the tilt of PV panels is adjusted to optimize power generation efficiency and light distribution.

[0038] In DQN, each action is represented as a discrete number or encoding. The size of the action space depends on the number of available crop types, planting methods, and tilt adjustment strategies.

[0039] More specifically, reward function design The reward function evaluates the quality of the action based on the current state, the action taken, and the subsequent state. In the scenario of photovoltaic power station operation and maintenance management on desertified land, the reward function comprehensively considers multiple aspects such as power generation, crop yield, and soil improvement effect.

[0040] Power generation reward: rewards are given according to the power generation efficiency of the photovoltaic modules. The higher the power generation, the greater the reward.

[0041] Crop Harvest Rewards: Rewards are given based on the yield and quality of the crops. The better the harvest, the greater the reward.

[0042] Soil improvement effect reward: rewards are given based on the degree of improvement of soil moisture content, temperature, pH, etc. The more obvious the improvement in soil conditions, the greater the reward.

[0043] In addition, in order to encourage the agent to explore new state-action pairs, a small exploration reward term can be added to the reward function. At the same time, in order to prevent the agent from taking actions that are harmful to the system (such as destroying the soil structure, affecting the safety of photovoltaic panels, etc.), negative rewards or penalties can be given in these cases.

[0044] Ultimately, the design of the reward function should be weighed and adjusted according to actual business needs to ensure that DQN can learn operation and maintenance management strategies that are both economically efficient and environmentally sustainable.

[0045] (2) Neural network structure Input layer: receives various indicators in the state space as input.

[0046] Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships. The number of hidden layers and the number of neurons in each layer should be adjusted according to the specific task.

[0047] Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action.

[0048] More specifically, the input layer is the first layer of the neural network, which is responsible for receiving data from the state space. In the scenario of agricultural photovoltaic complementary operation and maintenance management of photovoltaic power stations in desertified land, the input layer contains multiple neurons, each of which corresponds to an indicator in the state space. These indicators include: Current month: This can usually be input by encoding it as a one-hot encoding or a simple integer representation.

[0049] The tilt angle of the PV panel: It can be a continuous value representing the degree of tilt, or it can be divided into several intervals and represented by a one-hot vector.

[0050] Soil moisture content, soil temperature, soil acidity (pH value): These are continuous values ​​that can be used directly as input, but they need to be normalized to an input range that is more suitable for neural networks (such as between 0 and 1).

[0051] The main role of the input layer is to convert the raw data into a form that the neural network can process and pass it to the next layer for further processing.

[0052] More specifically, the hidden layer is the part of the neural network located between the input layer and the output layer, responsible for extracting features of the input data and learning complex nonlinear relationships.

[0053] In the hidden layer, an activation function (such as ReLU, Sigmoid, or Tanh) is used to increase the nonlinearity of the network. The choice of activation function will also affect the performance and training efficiency of the network.

[0054] More specifically, the output layer is the last layer of the neural network and is responsible for outputting the Q value of each action. In DQN, the number of neurons in the output layer is equal to the size of the action space.

[0055] Output layer neurons: Each neuron corresponds to an action, and its output value represents the expected return (i.e., Q value) of taking that action in that state. The Q value is obtained by learning the state features of the network and is used to guide the decision-making process of the intelligent agent.

[0056] During training, DQN uses the error between the target Q value (calculated by the target network) and the predicted Q value (calculated by the current network) to update the network weights. This process is implemented through the back-propagation algorithm, which aims to minimize the difference between the predicted Q value and the target Q value.

[0057] In summary, the structural design of the DQN neural network comprehensively considers the characteristics of the input data, task requirements, computing resources and other factors. By adjusting the number of hidden layers, the number of neurons in each layer, and the activation function and other parameters, a neural network model suitable for the operation and maintenance management of photovoltaic power stations on desertified land can be constructed.

[0058] (3) Training process Experience collection: Collect a large amount of state-action-reward-new state four-tuple data by simulating or actually operating photovoltaic power stations and crop planting systems.

[0059] Experience replay: The collected experience data is stored in the experience replay buffer. During the training process, a batch of experience samples are randomly selected from the buffer for training to break the time correlation between the data and improve the training stability.

[0060] Target network: The target network structure is the same as the DQN neural network but the parameter update frequency is lower. The target network is used to calculate the target Q value to stabilize the training process.

[0061] Loss function: The mean square error (MSE) is used as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value.

[0062] Optimization algorithm: Use an optimization algorithm (such as Adam) to update the parameters of the DQN neural network to minimize the loss function.

[0063] More specifically, before training DQN, a large amount of training data needs to be collected first. These data exist in the form of four-tuples, namely (state, action, reward, new state), representing the interaction history of the agent in the environment. Data can be collected by simulating the operation of photovoltaic power plants and crop planting systems. At each time step, the system records the current state (such as month, photovoltaic module inclination, soil indicators, etc.), the actions taken (such as adjusting the inclination, planting crops, etc.), the rewards obtained (such as increased power generation, good crop yields, etc.), and the new state reached after the action is executed. Experience replay breaks the temporal correlation between data by randomizing the presentation of samples. All collected experience data are stored in a buffer, which is a queue of limited size. When new data arrives, if the buffer is full, the earliest data will be replaced. In addition, during the training process, instead of directly using the latest experience data to update the network, a batch of samples are randomly selected from the experience replay buffer for training. The advantage of this is that the temporal correlation between samples is reduced, making the training process more stable.

[0064] The target network is used to calculate the target Q value to stabilize the training process. The structure of the target network is exactly the same as the DQN neural network, but the parameter update frequency is different.

[0065] The parameters of the DQN neural network are updated at each time step according to the gradient descent algorithm, while the parameters of the target network are copied from the DQN neural network at a certain number of steps (such as every 1000 steps). This makes the calculation of the target Q value more stable, which is conducive to the stability of the training process.

[0066] DQN uses mean squared error (MSE) as a loss function to measure the difference between the predicted Q value and the target Q value. For each training sample (state, action, reward, new state), the DQN neural network outputs a predicted Q value, which represents the expected return of taking the action in that state. The target Q value is calculated by the target network, which takes into account future rewards and state transitions. Specifically, the target Q value can be expressed as the current reward plus the discount of the maximum future Q value (i.e., r+γmaxa′Qtarget(s′,a′)), where r is the current reward, γ is the discount factor, s′ is the new state, and a′ is the optional action in the new state.

[0067] The difference between the predicted Q-value and the target Q-value is measured by the MSE loss, i.e., MSE=N1∑i=1N(Q(si,ai)−Qtarget(si,ai))2, where N is the number of samples in the batch.

[0068] Optionally, DQN uses an optimization algorithm (such as Adam) to update its parameters to minimize the loss function.

[0069] At each training iteration, the parameters of the DQN neural network are updated as directed by the gradient descent algorithm (or more specifically, the Adam algorithm). These updates are designed to reduce the difference between the predicted Q-values ​​and the target Q-values.

[0070] (4) Multi-objective optimization Specifically, different weights are assigned to power generation, crop yields, and soil improvement effects in the reward function to reflect their relative importance in the overall benefit. According to the output of the DQN neural network during training and the feedback of the reward function, the inclination angle of the photovoltaic module, the maintenance method, and the crop planting strategy are adjusted to achieve multi-objective optimization.

[0071] Power generation weight: Power generation is the main economic indicator of a photovoltaic power station, so its weight is usually high. However, the specific weight value needs to be determined based on actual conditions. For example, factors such as the power generation cost of the power station, market demand, and policy subsidies will affect the relative importance of power generation.

[0072] Crop harvest weight: Crop harvest is an important part of the agro-photovoltaic complementary system, reflecting the contribution of the system to agricultural production. Factors such as crop type, market demand, and price will affect the distribution of its weight.

[0073] Weight of soil improvement effect: Soil improvement is a manifestation of long-term benefits, which is beneficial to improving the fertility and ecological environment of the land. However, since the effect of soil improvement often takes a long time to manifest, its weight is relatively low.

[0074] In actual operation, the distribution of weights can be determined through expert consultation, market research, cost-benefit analysis and other methods.

[0075] Specifically, strategy adjustment is a specific implementation method of multi-objective optimization. According to the output of the DQN neural network and the feedback of the reward function, the inclination angle of the photovoltaic module, the maintenance method and the crop planting strategy are adjusted to achieve multi-objective optimization.

[0076] PV panel tilt adjustment: By adjusting the tilt of the PV panel, the angle at which it receives solar radiation can be optimized, thereby improving power generation efficiency. At the same time, the adjustment of the tilt angle also needs to take into account the needs of crop growth and the effect of soil improvement to ensure the maximization of overall benefits.

[0077] Optimizing maintenance methods: Regular maintenance of photovoltaic modules and agricultural facilities is an important measure to maintain stable operation of the system. By optimizing maintenance methods and cycles, the failure rate can be reduced, equipment utilization can be improved, and operation and maintenance costs can be reduced. In addition, targeted maintenance can be carried out in combination with the conditions of soil and crops during the maintenance process.

[0078] Adjustment of crop planting strategy: Formulate appropriate crop planting strategies based on soil conditions, climate conditions, market demand and other factors. During the planting process, attention should be paid to issues such as crop growth cycle, nutrient requirements, pest and disease control, etc., to ensure high yield and quality of crops. At the same time, the impact of crop planting density and layout on the power generation efficiency of photovoltaic modules should also be considered.

[0079] (5) Monitoring and evaluation Training monitoring: Use tools such as tensorboard to monitor changes in indicators such as loss value and average reward during training.

[0080] It should be noted that during the DQN training process, monitoring the changes in various key indicators is crucial for timely detection of problems and adjustment of training strategies.

[0081] Loss value monitoring: Use visualization tools such as TensorBoard to monitor the change of loss value during training. The loss value reflects the difference between the model's predicted Q value and the target Q value, and the change trend can reflect the convergence of the model and the training effect. If the loss value continues to decrease and gradually stabilizes, the model is gradually converging and learning effective strategies.

[0082] Average reward monitoring: In addition to the loss value, you can also monitor the changes in the average reward. The average reward refers to the average of the total rewards obtained by the agent in a certain period of time, reflecting the performance level of the model under the current strategy. If the average reward gradually increases and tends to stabilize, it means that the model is constantly optimizing its strategy and achieving better performance.

[0083] Monitoring of other indicators: Depending on the specific application scenarios and requirements, you can also monitor other related indicators, such as exploration rate (ε value in ε-greedy strategy), learning rate, training speed, etc.

[0084] Model evaluation: After training, an independent test dataset is used to evaluate the trained DQN model to check whether its performance under different conditions meets the expected goals.

[0085] Test dataset preparation: Prepare an independent test dataset, which is separate from the training dataset, to ensure the objectivity and accuracy of the evaluation results.

[0086] Performance indicator evaluation: According to the specific application scenario and requirements, select appropriate performance indicators for evaluation, including accuracy, average reward, cumulative reward, completion time, etc.

[0087] (6) Practical application The trained DQN model is deployed to the photovoltaic power station's agricultural-photovoltaic complementary operation and maintenance management system, and the optimal crop planting method and photovoltaic module tilt adjustment strategy are output based on real-time environmental data.

[0088] Regularly update data and retrain models to adapt to environmental changes and new business needs.

[0089] S3. Determine the operation and maintenance strategy corresponding to the optimal Q value as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

[0090] In an optional embodiment, the present solution proposes a method for operation and maintenance management of agricultural-photovoltaic complementary photovoltaic power stations in desertified land.

[0091] During the infrastructure construction phase, drones were used to take on-site photos to accurately identify desertified land, avoid expropriating basic farmland and forest land within the red line, set up soil moisture monitoring devices in each photovoltaic array, and record soil indicators such as soil moisture content of photovoltaic modules of all photovoltaic arrays at different inclination angles. After completing the record, adaptive planting of agricultural crops was carried out according to soil indicators, and the optimal crop planting method under different months, different photovoltaic module inclination angles, and different soil indicators was learned through neural networks. Multi-objective optimization was carried out based on the learning results. By reasonably planning the maintenance method of photovoltaic power stations and adjusting photovoltaic modules to the appropriate angle, it was ensured that conditions such as light were conducive to soil improvement and crop planting, and the optimal crops under different soil conditions were selected and planted, so as to maximize the benefits of photovoltaic power stations in desertified land, crop planting, and soil improvement.

[0092] First, determine the characteristics and classification of desertified land required in the image recognition algorithm. Determine the number of neurons and the number of neural network layers required for the neural network in the image recognition algorithm, and establish the neural network. Train the neural network with real desertified land images. After each training, adjust the training parameters to achieve the optimal training effect of the neural network. Then use new images to test the trained neural network. The trained neural network can be tested with image sets of desertified land of different blocks, sizes, colors, and inside and outside the red line. In this way, the ability of this scheme to identify desertified land inside and outside the red line can be tested. Image recognition CNN neural network, this scheme involves adjusting the training parameters, such as the number of layers and the number of neurons. Secondly, determine the characteristics and classification of different photovoltaic module inclinations, soil indicators and crops. The characteristics and classification of different photovoltaic module inclinations, soil indicators and crops refer to the use of different photovoltaic module inclinations, different soil indicators and different crops as state quantities when training the neural network. Its feature, i.e., feature, refers to the feature of each state quantity, and classification refers to the type of state quantity (for example, the crops planted are eggplant, potato, corn, etc.); determine the number of neurons and the number of neural network layers required in the deep reinforcement learning algorithm, and establish a deep-Q-network (DQN) neural network. The DQN neural network is trained by taking different photovoltaic module inclinations, different soil indicators and different crops as state quantities, different photovoltaic power station operation and maintenance methods, and drip irrigation frequency of crop drip irrigation belts as action quantities, and the power generation of photovoltaic power stations on desertified land, soil improvement effects and crop harvests as reward equations. After each training, the training parameters are adjusted to achieve the optimal training effect of the neural network, that is, under different photovoltaic module inclinations, different soil indicators and different crop planting conditions, the optimal photovoltaic power station operation and maintenance methods, the drip irrigation frequency of crop drip irrigation belts, the power generation of photovoltaic power stations on desertified land, the optimal soil improvement effects and the optimal crop harvests are obtained.

[0093] The power generation pursues the best inclination angle and the best maintenance route. The best soil improvement effect pursues the best soil moisture content, which requires adjusting the inclination angle of photovoltaic modules and the maintenance time. The best crop yield requires soil drip irrigation, and the drip irrigation belt will affect the maintenance route. Therefore, this plan is a multi-objective optimization.

[0094] Afterwards, the trained neural network was tested using different photovoltaic power station operation and maintenance methods and crop drip irrigation frequencies. Based on the test results, the neural network learning rate, learning step and other related parameters were adjusted, and learning training and testing were carried out again.

[0095] According to the tested neural network, combined with uncertain factors such as sudden failures, the location of the photovoltaic array in the desertified land and road conditions, the optimal maintenance and fault elimination route is planned, and the angle of the photovoltaic modules in the corresponding area and the drip irrigation frequency of the crop drip irrigation belt are adjusted. The soil improvement effect and crop planting income are checked regularly to maximize the power generation of photovoltaic power stations in desertified land, soil improvement effect and crop yield.

[0096] Uncertainties are reflected in the training process, and relevant uncertainties are included in the training set.

[0097] In some other optional embodiments, an operation and maintenance method is also provided, including: First, build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm. Determine the number of neurons and neural network layers required for the image recognition algorithm, and establish a (CNN) neural network. Train the neural network with real desertified land images. After each training session, adjust the training parameters to achieve the optimal training effect for the neural network. Then use image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line to test the trained neural network, and adjust the neural network parameters according to the test results. After the adjustment, train and test again, and iterate until the optimal effect is achieved.

[0098] Secondly, build a training environment for the neural network of the deep reinforcement learning algorithm. Mathematical modeling is performed on the internal facilities, array positions, road conditions, soil indicators, etc. of the photovoltaic power station, and the mixed-integer linear programming (MILP) model is used as the final form for linearization. Analyze and model the uncertainties related to photovoltaic power station operation and maintenance, soil improvement, crop planting, etc. (light and temperature, soil moisture content, photovoltaic component fault location, fault time, etc.), and establish the distribution curve of uncertainty based on historical data to clarify the correlation between the uncertainties (different uncertainties affect each other); determine the state space (state variables), action space (control variables), and reward function of the neural network, and build a mathematical environment that can represent the photovoltaic power station; use Python to build a "smart brain" based on the deep reinforcement learning algorithm, and connect the mathematical environment through the Gurobi planning optimization operator; train the DQN neural network in the "smart brain" through historical data, and achieve the optimal training effect by adjusting the training parameters; use real-time data to test the trained DQN and use the real environment to verify the simulation results; actually apply this strategy in photovoltaic power stations to maximize the benefits of agricultural-photovoltaic complementarity.

[0099] Uncertainty is included in every training set, that is, the training environment is different each time you train.

[0100] Because the existing onshore photovoltaic power station operation and maintenance strategies cannot be directly applied to the photovoltaic power stations for agricultural and photovoltaic complementary use in areas such as desertified land. In addition, the existing drone technology cannot accurately identify whether the required land is within the red line, and needs to be compared with the forestry and grassland three-adjustment map. The present invention can eliminate the land within the red line through learning, accurately select the desertified land, and avoid disruptive problems. The existing technology cannot effectively combine soil improvement, crop planting and photovoltaic power station operation and maintenance, and cannot achieve multi-objective optimization tasks. However, through the scheme of the present invention, the power generation of photovoltaic power stations on desertified land, the optimal soil improvement effect and the optimal crop yield can be obtained.

[0101] Example 2 like Figure 2 As shown, based on the same inventive concept as the above embodiment, the present invention also provides an agricultural-photovoltaic complementary operation and maintenance device for a photovoltaic power station in desertified land, comprising: The input data acquisition module is used to obtain the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, and each photovoltaic array includes a plurality of photovoltaic modules with adjustable inclination angles; A strategy optimization module is used to input the status of the photovoltaic power station, the growth status of crops and the external environmental conditions into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The strategy determination module is used to determine the operation and maintenance strategy corresponding to the optimal Q value as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

[0102] Furthermore, in the input data acquisition module, the image recognition CNN neural network is trained as follows: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

[0103] Furthermore, in the strategy optimization module, the DQN neural network model is trained as follows: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

[0104] Furthermore, different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization objective in the overall benefit.

[0105] Example 3 like Figure 3 As shown, the present invention also provides an electronic device 100 for implementing a method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station in desertified land according to Embodiment 1; The electronic device 100 includes a memory 101 , at least one processor 102 , a computer program 103 stored in the memory 101 and executable on the at least one processor 102 , and at least one communication bus 104 .

[0106] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the agricultural-photovoltaic complementary operation and maintenance method of a photovoltaic power station on desertified land in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0107] The memory 101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0108] At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0109] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for operation and maintenance of a photovoltaic power station on desertified land with complementary agriculture and photovoltaic power. The processor 102 can execute the plurality of instructions to implement: Acquire the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, each of which includes a plurality of photovoltaic modules with adjustable inclination angles; The photovoltaic power station status, crop growth status and external environmental conditions are input into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The operation and maintenance strategy corresponding to the optimal Q value is determined as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

[0110] Example 4 If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0111] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land, characterized in that: include: Acquire the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, each of which includes a plurality of photovoltaic modules with adjustable inclination angles; The photovoltaic power station status, crop growth status and external environmental conditions are input into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The operation and maintenance strategy corresponding to the optimal Q value is determined as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

2. The method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land according to claim 1, characterized in that: The image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; wherein the image recognition CNN neural network is trained in the following manner: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

3. The method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land according to claim 1, characterized in that: The photovoltaic power station status, crop growth status and external environmental conditions are input into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; wherein the DQN neural network model is trained in the following manner: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

4. The method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land according to claim 3 is characterized in that: Different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization objective in the overall benefit.

5. A photovoltaic power station with agricultural and photovoltaic complementary operation and maintenance device for desertified land, characterized in that: include: The input data acquisition module is used to obtain the status of the photovoltaic power station, the growth status of crops and the external environmental conditions; wherein the photovoltaic power station is built in the desertified land area of ​​the target area, and the image data of the target area is identified by a pre-trained image recognition CNN neural network to determine the desertified land area; the photovoltaic power station includes a plurality of photovoltaic arrays, and each photovoltaic array includes a plurality of photovoltaic modules with adjustable inclination angles; A strategy optimization module is used to input the status of the photovoltaic power station, the growth status of crops and the external environmental conditions into a pre-trained DQN neural network model, and the DQN neural network model outputs the Q value corresponding to each operation and maintenance strategy; The strategy determination module is used to determine the operation and maintenance strategy corresponding to the optimal Q value as the optimal operation and maintenance strategy; wherein the operation and maintenance strategy includes the maintenance method of the photovoltaic power station, the inclination angle of the photovoltaic components and the crop planting method.

6. The agricultural-photovoltaic complementary operation and maintenance device for a photovoltaic power station in desertified land according to claim 5, characterized in that: In the input data acquisition module, the image recognition CNN neural network is trained as follows: Build a desertified land image recognition training environment to determine the desertified land features and classifications required in the image recognition algorithm; Determine the number of neurons and neural network layers required in the image recognition algorithm, and establish a CNN neural network; The neural network is trained by using real desertified land images; after each training session, the training parameters are adjusted to achieve the optimal training effect for the neural network; After the training is completed, the trained neural network is tested using image sets of desertified land in different blocks, sizes, colors, and inside and outside the red line. The neural network parameters are adjusted according to the test results, and training and testing are performed again after the adjustment, iterating until the optimal effect is achieved.

7. The method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land according to claim 5, characterized in that: In the strategy optimization module, the DQN neural network model is trained as follows: Define the state space; the state includes the normal state and the current month in the emergency situation, the inclination angle of the photovoltaic module, the soil moisture content, the soil temperature, and the soil pH index; Define an action space; the action space includes all crop planting methods, PV power station maintenance methods, and PV module tilt adjustment strategies. Each action corresponds to a specific crop planting decision, PV power station maintenance, or PV module tilt adjustment. Design a reward function; the reward function is used to reflect the power generation, crop yield and soil improvement effects; Determine the network structure of the DQN neural network model, including: Input layer: receives various indicators in the state space as input; Hidden layers: Set multiple hidden layers to extract features and learn complex nonlinear relationships; Output layer: The output layer contains neurons equal to the size of the action space, and the output of each neuron represents the Q value of the corresponding action; The training process is as follows: by simulating photovoltaic power stations and crop planting systems, collect state-action-reward-new state four-tuple data for subsequent training of the DQN neural network; store the collected experience data in the experience replay buffer; during the training process, randomly select a batch of experience samples from the buffer for training; the target network calculates the target Q value; uses the mean square error as the loss function to measure the difference between the Q value output by the DQN neural network and the target Q value; uses the optimization algorithm to update the parameters of the DQN neural network to minimize the loss function.

8. The method for agricultural-photovoltaic complementary operation and maintenance of a photovoltaic power station on desertified land according to claim 7, characterized in that: Different weights are assigned to power generation, crop yield and soil improvement effect in the reward function, and the weights are used to reflect the relative importance of each optimization objective in the overall benefit.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the agricultural-photovoltaic complementary operation and maintenance method of a photovoltaic power station on desertified land as claimed in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, the agricultural-photovoltaic complementary operation and maintenance method of a photovoltaic power station in desertified land is implemented as described in any one of claims 1 to 4.

Images