A synergistic management method for photovoltaic power station and fishery light complementation

Abstract

The application discloses a kind of photovoltaic station fish-light complementary synergic management methods, and the application relates to photovoltaic station operation and maintenance technical field.The method steps include: collecting data;Important features are screened using recursive feature selection method based on random forest, growth data are obtained, and the feature weights of each feature are calculated using TreeSHAP method;The growth deviation degree corresponding to each feature in growth data is calculated, the water biological efficiency index is calculated in combination with feature weight, and the efficiency curve is drawn;The growth trend factor of efficiency curve is calculated based on least square method of sliding window, and the growth trend stability is determined;The inertia weight of particle swarm optimization algorithm is adjusted using growth trend stability, and the optimal state of photovoltaic panel, i.e., the optimal photovoltaic panel inclination and the optimal angle of reflective film, is obtained;The optimal photovoltaic panel inclination and the optimal angle of reflective film are used to predict the power generation of photovoltaic panel, and the optimal power generation of photovoltaic station is obtained, whereby the photovoltaic station is synergically managed, and the operation and maintenance state of photovoltaic station is optimized.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, specifically a collaborative management method for photovoltaic power plants that integrates solar and fish farming. Background Technology

[0002] With the advancement of clean energy strategies and the increasing constraints of land resources, the aquaculture-solar complementary model, which combines photovoltaic power generation with aquaculture, has shown great potential. This model achieves integrated land use by installing photovoltaic panel arrays above the aquaculture water surface, enabling power generation on the water and aquaculture below. It generates green electricity while sustaining aquaculture production, becoming an important development direction for promoting agricultural modernization and energy transition. Efficiently and collaboratively managing photovoltaic power generation and aquaculture, ensuring the healthy growth of aquatic organisms, and maximizing power generation revenue are the core challenges for promoting the high-quality development of the aquaculture-solar complementary industry.

[0003] However, in existing technologies, the configuration of photovoltaic panels in the operation and maintenance of photovoltaic power plants is usually set based on the principle of maximizing static power generation efficiency. For example, a fixed tilt angle is set according to the season and latitude to maximize the annual power generation. Only the basic shading effect on the aquaculture water body is considered, and the needs of aquatic organisms for survival in the underwater aquaculture environment are not fully considered. It is difficult to achieve the joint optimization of photovoltaic panels and the growth of aquatic organisms, resulting in poor operation and maintenance of photovoltaic power plants. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a collaborative management method for photovoltaic power plants that integrates solar and fish farming, thereby resolving the problems existing in the background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a collaborative management method for photovoltaic power stations integrating solar and fish farming, comprising the following steps:

[0006] Step S1: Collect aquatic organism growth data under different living conditions and preprocess it to obtain aquatic organism growth data;

[0007] Step S2: The aquatic organism growth data is filtered by feature importance using a recursive feature selection method based on random forest to obtain the filtered second aquatic organism growth data; the feature weight of each feature in the second aquatic organism growth data is calculated using the TreeSHAP method.

[0008] Step S3: By calculating the deviation between the second aquatic organism growth data and the preset growth threshold range, the growth deviation of each feature in the second aquatic organism growth data is obtained; by combining the growth deviation of each feature and the corresponding feature weight, the aquatic organism efficiency index is calculated; and the aquatic organism efficiency curve is plotted based on the aquatic organism efficiency index.

[0009] Step S4: Calculate the growth trend factor of the aquatic organism efficiency curve using the least squares method based on a sliding window, obtain the optimal survival conditions of aquatic organisms based on the growth trend factor, and calculate the growth trend stability of the aquatic organism efficiency curve corresponding to the optimal survival conditions.

[0010] Step S5: Adjust the inertia in the particle swarm optimization algorithm through the growth trend stability, and obtain the optimal state of the photovoltaic panel through the particle swarm optimization algorithm. The optimal state of the photovoltaic panel includes: the optimal photovoltaic panel tilt angle and the optimal reflective film angle.

[0011] Step S6: Predict the power generation of the photovoltaic panels in the optimal state for each aquaculture cycle, obtain the predicted power generation value, and provide adjustment strategies for aquaculture based on the predicted power generation value to achieve coordinated management of fishery-solar complementarity.

[0012] Preferably, the step of filtering the aquatic organism growth data by feature importance using a recursive feature selection method based on random forest to obtain the filtered second aquatic organism growth data includes the following specific steps:

[0013] The aquatic organism growth data were selected based on feature importance using a recursive feature selection method based on random forest. The aquatic organism growth data is X={ , ,... ,..., }, construct a decision tree set T={ , ,..., ,..., There are M trees in total;

[0014] For each tree and characteristics Calculate the cumulative purity increase for each feature across all split points:

[0015]

[0016] in, This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree. For trees The set of split points, h is In the context of a split point h, I() is an indicator function, i.e., the characteristic index of the split point h. The index p is 1 when it is equal to the index of the aquatic organism growth data feature, and 0 when they are not. The feature index of the split point h, The reduction in Gini impurity at the splitting point h. = , Let the Gini impurity be the parent node of the split point h. The child node of the split point h, Let be the Gini impurity of the child node `child` of the split point `h`. and , which represent the number of samples for the parent and child nodes of the split point h, respectively, and p is the index of the aquatic organism growth data feature;

[0017] Then features The feature importance in the random forest model is as follows:

[0018]

[0019] in, The p-th feature of aquatic organism growth data In a random forest model, feature importance is denoted by M, where M is the total number of decision trees and m is the m-th decision tree. This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree;

[0020] Based on the feature importance of all features in the aquatic organism growth data in the random forest model, the features with the lowest feature importance are removed to form a new feature subset. The random forest model is then retrained using the new feature subset, and the feature importance is repeatedly calculated until the number of remaining features reaches the preset feature quantity threshold. Finally, the filtered second aquatic organism growth data is output.

[0021] Preferably, the step of calculating the feature weight of each feature in the second aquatic organism growth data using the TreeSHAP method includes the following steps:

[0022] The SHAP value of each feature of the second aquatic organism growth data was calculated using the TReeSHAP method:

[0023]

[0024] in, Let be the SHAP value of the p-th feature in the second aquatic organism growth data, M be the total number of decision trees, m be the m-th decision tree, and h be the split node in the decision tree. Indicates that node h is affected by features Number of samples affected This is the p-th feature in the second set of aquatic organism growth data. This represents the total number of samples reaching node h. The reduction in Gini impurity at the split point h;

[0025] Calculate the weight of each feature in the second aquatic organism growth data based on the SHAP value of each feature:

[0026]

[0027] in, Let be the weight of the p-th feature in the second aquatic organism growth data. P1 is the SHAP value of the p-th feature in the second aquatic organism growth data, and P2 is the total number of features in the second aquatic organism growth data.

[0028] Preferably, the step of calculating the deviation between the second aquatic organism growth data and a preset growth threshold range to obtain the growth deviation of each feature in the second aquatic organism growth data includes the following steps:

[0029] The growth deviation of each feature in the second aquatic organism growth data at each moment in each aquaculture cycle is calculated based on the deviation between the second aquatic organism growth data and the preset growth threshold.

[0030]

[0031] in, The bias of the p-th feature in the second aquatic organism growth data. This is the p-th feature in the second set of aquatic organism growth data. The center value of the preset growth threshold interval for the p-th feature in the second aquatic organism growth data is given. This is the upper limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data. This is the lower limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data.

[0032] Preferably, the step of calculating the aquatic organism efficiency index by combining the growth deviation degree of each feature and the corresponding feature weight includes the following steps:

[0033] By combining the growth deviation of each feature with its corresponding feature weight, the biomass efficiency index of the aquaculture process at each moment in each aquaculture cycle is calculated:

[0034]

[0035] in, P2 represents the aquatic organism efficiency index, and P2 represents the total number of features in the second aquatic organism growth data. Let p be the weight of the p-th feature in the second aquatic organism growth data. The deviation of the p-th feature in the second aquatic organism growth data.

[0036] Preferably, the calculation of the growth trend factor of the water production efficiency curve using the least squares method based on a sliding window includes the following specific steps:

[0037] For each window k, the growth slope in window k is calculated using the least squares method:

[0038]

[0039] in, Let the growth slope of the k-th window be . This indicates the start time of the k-th window. Let k be the end time of the k-th window. The average time value for the k-th window. This represents the water production efficiency index at time t under the k-th window. represents the average water production efficiency index of the k-th window, and t represents time t;

[0040] The growth trend factor for each breeding cycle is calculated using the growth slope within window k in each breeding cycle.

[0041]

[0042] in, Let K be the growth trend factor, K be the total number of windows in the breeding cycle, and k be the index of the k-th window. Let be the growth slope of the k-th window.

[0043] Preferably, the step of obtaining the optimal survival conditions of aquatic organisms based on the growth trend factor and calculating the growth trend stability of the aquatic organism performance curve corresponding to the optimal survival conditions includes the following specific steps:

[0044] The optimal survival conditions for aquatic organisms are obtained based on the growth trend factor, and the growth trend stability of the aquatic organism efficiency curve corresponding to the optimal survival conditions is calculated.

[0045] After determining the optimal survival conditions for aquatic products in each aquaculture cycle, the growth trend stability of the aquatic bioefficiency curve under the optimal survival conditions in each aquaculture cycle is calculated:

[0046]

[0047] in, To ensure the stability of the growth trend, K represents the total number of windows in the breeding cycle, and k represents the index of the k-th window. The growth slope of the k-th window under optimal survival conditions. The growth trend factor under optimal survival conditions. To prevent the denominator from being 0, the value is a constant and takes the value of 0. .

[0048] Preferably, adjusting the inertia in the particle swarm optimization algorithm based on the growth trend stability includes the following specific steps:

[0049] The inertia in the particle swarm optimization algorithm is adjusted by the stability of the growth trend.

[0050]

[0051] in, The adjusted inertia value, This is the initial inertia value used in the particle swarm optimization algorithm; the default value is 0.7. The adjustment range is set to 0.15 by default. This is the response coefficient, which defaults to 1. For growth trend stability, This is the threshold for growth trend stability, with a default value of 4.

[0052] Preferably, obtaining the optimal state of the photovoltaic panel through the particle swarm optimization algorithm includes the following specific steps:

[0053] The formulas for calculating the state parameters of photovoltaic panels and the light intensity on the water surface are as follows:

[0054]

[0055] in, The intensity of light on the water surface. Natural light intensity, The base transmittance coefficient has a value range of (0.65, 0.75). The tilt angle influence coefficient has a value range of (-0.015, -0.008). is the reflective film supplementary light coefficient, with a value range of (0.003, 0.008), and FG is the reflective film angle of the photovoltaic panel. The tilt angle of the photovoltaic panel;

[0056] The formulas for calculating the state parameters of photovoltaic panels and water surface temperature are as follows:

[0057]

[0058] in, Water surface temperature Atmospheric temperature, This is the light-to-temperature conversion factor, with a default value of 0.0001. Natural light intensity, The base temperature rise coefficient has a value range of (0.8, 1, 2), and FG is the angle of the reflective film on the photovoltaic panel. The tilt angle of the photovoltaic panel. The tilt angle cooling coefficient has a value range of (-0.02, -0.01). The temperature coefficient of the reflective film ranges from 0.005 to 0.012.

[0059] Each day of each aquaculture cycle, the photovoltaic panels are adjusted based on the day's light intensity, temperature, and optimal survival conditions. The water surface temperature and light intensity obtained after the photovoltaic panel adjustment are compared with the optimal survival conditions to obtain a photothermal adaptation value, which is then used as the fitness value in the particle swarm optimization algorithm.

[0060]

[0061] Where SX is the photothermal adaptation value. The intensity of light on the water surface. For optimal light intensity, Water surface temperature To achieve the optimal water surface temperature;

[0062] Based on the fitness value in the particle swarm optimization algorithm, the velocity and position of each particle in the algorithm are updated:

[0063]

[0064] in, Let wd represent the velocity of the nth particle in the (dd+1)th iteration, and gx be the inertia weight in the particle swarm optimization algorithm, which controls the degree to which the particle retains its historical velocities. This represents the velocity of the wd-th feature of the nth particle in dd iterations. and This represents the individual learning factor and the social learning factor in the particle swarm optimization algorithm, and is usually taken as 2. and All numbers are random to avoid local optima in the particle swarm optimization algorithm. The wd-th eigenvalue represents the wd-th eigenvalue of the historical best position of the n-th particle. Let wd be the wd-th eigenvalue of the currently found global optimum position for all particles in the particle swarm optimization algorithm. This represents the position of the wd-th feature of the nth particle in dd iterations;

[0065]

[0066] in, This represents the position of the wd-th feature of the n-th particle in the (dd+1)-th iteration. This represents the position of the wd-th feature of the nth particle in dd iterations. This represents the velocity of the wd-th feature of the nth particle in the (dd+1)th iteration;

[0067] By continuously updating particle positions, the search continues for particles that maximize the photothermal fit. This process continues until the maximum number of iterations is reached. When the iteration stops or the difference between the global optimal fitness value (i.e., the maximum photothermal fitness value SX) of all particles in adjacent iterations is less than 1× When the particle swarm optimization algorithm is considered to have reached a stable convergence state, it stops iterating. After stopping iteration, the photovoltaic panel tilt angle and reflective film angle corresponding to the photovoltaic panel state parameter with the largest photothermal adaptation value are taken as the optimal photovoltaic panel tilt angle. Optimal reflective film angle .

[0068] Preferably, the step of predicting the power generation of the photovoltaic panels in their optimal state for each breeding cycle to obtain the predicted power generation value includes the following specific steps:

[0069] The photovoltaic panels are adjusted daily during each breeding cycle to achieve optimal performance. Power generation is then predicted based on this optimal state, using the optimal tilt angle to obtain the first predicted power generation value.

[0070]

[0071] in, This is the first predicted power generation value. For power generation duration, For the power generation efficiency of photovoltaic panels, This represents the total solar radiation. The tilt efficiency coefficient has a value range of (0.010, 0.025). To achieve the optimal tilt angle for photovoltaic panels, The latitude is the local latitude.

[0072] By determining the optimal reflective film angle, we can obtain the compensation amount for power generation based on the optimal reflective film angle:

[0073]

[0074] Where BC is the compensation amount for power generation based on the optimal reflective film angle. This is the radiation gain coefficient, with a value range of (0.002, 0.003). This is the cooling gain coefficient, with a value ranging from (0.0001, 0.0003). For the optimal reflective film angle, Atmospheric temperature, For reference temperature, the default value is 25℃;

[0075] The final power generation forecast is:

[0076]

[0077] in, This is the predicted value for power generation. BC represents the first predicted power generation value, and BC represents the compensation amount for power generation based on the optimal reflective film angle.

[0078] This invention provides a collaborative management method for photovoltaic power plants that integrates solar and fish farming, involving machine learning and deep learning technologies, and has the following beneficial effects:

[0079] (1) The combination of recursive feature selection in random forest and TreeSHAP method: Recursive feature selection evaluates feature importance through multi-decision tree ensemble, dynamically removes redundant and noisy features, and achieves adaptive compression of high-dimensional biological characteristic data, significantly improving computational efficiency; while TreeSHAP method analyzes the decision mechanism of random forest based on game theory, and accurately reveals the direction and interaction law of the influence of core characteristics on biological efficacy by quantifying the SHAP value of features, forming a closed loop. Recursive screening provides a subset of denoised features for TreeSHAP, while TreeSHAP verifies the scientific nature of feature screening through weight allocation and attribution analysis, together providing a foundation for the construction of aquatic biological efficacy index.

[0080] (2) By combining the growth deviation of each feature with the corresponding TreeSHAP feature weight, the aquatic organism efficiency index is calculated, realizing a dynamic quantitative assessment of the organism's growth state and adaptability to living conditions. The index value reflects the comprehensive growth efficiency level of aquatic organisms under a specific combination of environmental parameters, integrating multidimensional environmental interaction effects and biological response characteristics, and providing a precise feedback indicator that can directly drive parameter adjustment for subsequent particle swarm collaborative optimization.

[0081] (3) By adjusting the inertia weight in the particle swarm optimization (PSO) algorithm based on the stability of the growth trend, the parameter optimization efficiency and convergence stability are significantly improved. When the growth trend stability is high, the inertia weight is increased to enhance the global search capability; when the stability is low, the inertia weight is decreased to enhance the local development accuracy. Attached Figure Description

[0082] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0083] Figure 1 The flowchart illustrates the steps of a collaborative management method for photovoltaic power stations that integrates solar and aquaculture, as proposed in this invention.

[0084] Figure 2 This is a step-by-step diagram of a collaborative management method for photovoltaic power stations that integrates solar and fish farming, as proposed in this invention. Detailed Implementation

[0085] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0086] Please see Figure 1-2 The present invention provides a technical solution: a collaborative management method for photovoltaic power stations that integrates solar and fish farming.

[0087] Step S1: Collect aquatic organism growth data under different living conditions and preprocess it to obtain aquatic organism growth data.

[0088] During the data collection phase, growth process data of aquatic organisms under different living conditions were systematically collected. This included aquatic vital signs (such as shell hardness, body length and weight gain rate, and feeding behavior frequency in shrimp and crabs) and physiological indicators (blood biochemical parameters, stress response level). These different living conditions included: strong light zone 1500 lux / 30℃ (allowable fluctuation range 1400~1600 lux, 29.5~30.5℃), medium light zone 1000 lux / 28℃ (allowable fluctuation range 900~1100 lux, 27.5~28.5℃), and weak light zone 600 lux / 26℃ (allowable fluctuation range 500~700 lux, 25.5~26.5℃).

[0089] To address the issue of missing data due to equipment malfunction, Layda's rule was used to remove data, and the K-nearest neighbor method was used to interpolate and fill in the missing data and the data removed by Layda's rule, ultimately obtaining aquatic organism growth data.

[0090] It should be noted that by using Laida's rule and the "K-nearest neighbor completion method" to remove and fill data, when using Laida's rule (3... The criteria are combined with the "K-Nearest Neighbor Completion Method". For each aquatic growth data point, the mean and standard deviation of the dataset are calculated. When the absolute value of the deviation of a data point from the mean of the feature is greater than 3 times the standard deviation, it is judged as outlier data and removed. Subsequently, the missing values ​​caused by the removal of outliers and the original missing values ​​are uniformly filled using the "K-Nearest Neighbor Completion Method" - for each data point to be filled, its Euclidean distance with other samples in the feature space is calculated, the 5 nearest neighbor samples are selected, and the mean of these nearest neighbor samples on the feature is used as the result of the missing value completion, and finally a complete and reliable aquatic organism growth data is constructed.

[0091] Step S2: The aquatic organism growth data is filtered by feature importance using a recursive feature selection method based on random forest to obtain the filtered second aquatic organism growth data; the feature weight of each feature in the second aquatic organism growth data is calculated using the TreeSHAP method.

[0092] Feature importance screening was performed on aquatic organism growth data under different survival conditions. Aquatic organism production data was used as the independent variable, and aquatic product yield (low, medium, and high categories) as the dependent variable. A recursive feature selection method based on random forest was used to screen the feature importance of the aquatic organism growth data. The aquatic organism growth data is X={ , ,... ,..., }, construct a decision tree set T={ , ,..., ,..., There are a total of M trees (M is usually in the range of 100-500).

[0093] For each tree and characteristics Calculate the cumulative purity increase for each feature across all split points:

[0094]

[0095] in, This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree. For trees The set of split points, h is In the context of a split point h, I() is an indicator function, i.e., the characteristic index of the split point h. The index p is 1 when it is equal to the index of the aquatic organism growth data feature, and 0 when they are not. The feature index of the split point h, The reduction in Gini impurity at the splitting point h. = , Let the Gini impurity be the parent node of the split point h. The child node of the split point h, Let be the Gini impurity of the child node `child` of the split point `h`. and , where h represents the number of samples for the parent and child nodes of the split point, and p is the index of the aquatic organism growth data feature.

[0096] Then features The feature importance in the random forest model is as follows:

[0097]

[0098] in, The p-th feature of aquatic organism growth data In a random forest model, feature importance is denoted by M, where M is the total number of decision trees and m is the m-th decision tree. This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree.

[0099] Based on the feature importance of all features in the aquatic organism growth data in the random forest model, the features with the lowest feature importance are removed to form a new feature subset. The random forest model is retrained with the new feature subset, and the feature importance is repeatedly calculated until the number of remaining features reaches a preset feature quantity threshold (such as 50% of the total number of features in the aquatic organism growth data). Finally, the filtered second aquatic organism growth data is output.

[0100] It should be noted that by constructing multiple decision trees, the importance of multidimensional features (such as shell hardness, feeding frequency, blood biochemical parameters, etc.) of aquatic organism growth data is assessed. The contribution of each feature to aquatic production is quantified by using the cumulative reduction in Gini impurity of the feature across all split nodes. By recursively removing the least important features and remodeling, the biometric features sensitive to aquatic production (such as body length and weight gain rate, stress response level, etc.) can be adaptively selected, effectively eliminating redundant noise (such as abnormal body temperature data caused by equipment failure). The original multidimensional features are compressed, which significantly improves the efficiency and interpretability of subsequent biological efficacy index calculations while ensuring model accuracy.

[0101] The random forest model was retrained using the filtered second aquatic organism growth data. Based on the last retrained random forest model (i.e., the model corresponding to the second aquatic organism growth data), the SHAP value of each feature of the second aquatic organism growth data was calculated using the TReeSHAP method.

[0102]

[0103] in, Let be the SHAP value of the p-th feature in the second aquatic organism growth data, M be the total number of decision trees, m be the m-th decision tree, and h be the split node in the decision tree. Indicates that node h is affected by features Number of samples affected This is the p-th feature in the second set of aquatic organism growth data. This represents the total number of samples reaching node h. The reduction in Gini impurity at the split point h.

[0104] Calculate the weight of each feature in the second aquatic organism growth data based on the SHAP value of each feature:

[0105]

[0106] in, Let be the weight of the p-th feature in the second aquatic organism growth data. P1 is the SHAP value of the p-th feature in the second aquatic organism growth data, and P2 is the total number of features in the second aquatic organism growth data.

[0107] It should be noted that this paper adopts the TreeSHAP method to analyze the black-box decision-making mechanism of the random forest model and quantify the influence weight and direction of each survival condition on the growth efficiency of aquatic organisms. Based on the principle of game theory, this method reveals the positive / negative driving effect of core features (such as body length and weight gain rate) on biological efficiency by calculating the expected marginal contribution of features in all decision tree splitting paths (such as the shell hardness SHAP value reaching +0.32 in strong light areas, confirming the positive correlation between light intensity and shell development). By quantifying the contribution of local samples (such as the stress response level SHAP value suddenly dropping to -0.15 on high-temperature days), the nonlinear interaction law between features is dynamically captured. This provides mathematical rigor support for the subsequent construction of an aquatic organism efficiency index with clear biophysical significance, and significantly improves the causal interpretability and regulatory strategy reliability of the biological efficiency index model.

[0108] Step S3: By calculating the deviation between the second aquatic organism growth data and the preset growth threshold range, the growth deviation of each feature in the second aquatic organism growth data is obtained; by combining the growth deviation of each feature and the corresponding feature weight, the aquatic organism efficiency index is calculated; and the aquatic organism efficiency curve is plotted based on the aquatic organism efficiency index.

[0109] The growth deviation of each feature in the second aquatic organism growth data at each moment in each aquaculture cycle is calculated based on the deviation between the second aquatic organism growth data and the preset growth threshold.

[0110]

[0111] in, The bias of the p-th feature in the second aquatic organism growth data. This is the p-th feature in the second set of aquatic organism growth data. The center value of the preset growth threshold interval for the p-th feature in the second aquatic organism growth data is given. This is the upper limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data. This is the lower limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data.

[0112] It should be noted that the growth deviation is used to quantify the standardized deviation of actual physical characteristics of aquatic organisms under specific living conditions from a preset ideal growth threshold range. The numerator is the absolute difference between the measured value of a single physical characteristic (such as the body length gain rate of tilapia) and the center value of the preset optimal range for that characteristic, and the denominator is the total length of the threshold range for that characteristic. For example, on day 15 of the mud crab farming cycle, the measured shell hardness is 75 HA (Herx hardness), and its preset optimal growth threshold range is [78, 82 HA] (i.e., lower threshold 78, upper threshold 82, center value 80). The growth deviation is then 1.25, indicating that the current shell hardness is 25% lower than the ideal center value.

[0113] By combining the growth deviation of each feature with its corresponding feature weight, the biomass efficiency index of the aquaculture process at each moment in each aquaculture cycle is calculated:

[0114]

[0115] in, P2 represents the aquatic organism efficiency index, and P2 represents the total number of features in the second aquatic organism growth data. Let p be the weight of the p-th feature in the second aquatic organism growth data. The deviation of the p-th feature in the second aquatic organism growth data.

[0116] It should be noted that the aquatic organism efficiency index, as a quantitative indicator of the overall growth status of aquatic organisms, integrates the real-time deviation of multidimensional physical characteristics (normalized through an exponential function) and the weights of their corresponding characteristics to calculate the aquatic organism efficiency index. For example, taking the 20th day of mud crab farming as an example, assuming that only two key physical characteristics need to be calculated—shell hardness (characteristic p1) and feeding frequency (characteristic p2)—when the deviation of characteristic p1 is 1.25 with a weight of 0.67, and the deviation of characteristic p2 is 0.167 with a weight of 0.33, the aquatic organism efficiency index value is 0.471, significantly lower than the theoretical optimal value of 1.0, indicating that the current living conditions are causing severe lag in the development of the mud crab's shell.

[0117] Based on the aquatic organism efficiency index, an aquatic organism efficiency curve is plotted: for each aquaculture cycle, the aquatic organism efficiency curve is plotted with time as the horizontal axis and the aquatic organism efficiency index corresponding to each time as the vertical axis.

[0118] Step S4: Calculate the growth trend factor of the aquatic organism efficiency curve using the least squares method based on a sliding window, obtain the optimal survival conditions of aquatic organisms based on the growth trend factor, and calculate the growth trend stability of the aquatic organism efficiency curve corresponding to the optimal survival conditions.

[0119] The growth trend factor of the aquatic bioefficiency curve was calculated using the least squares method based on a sliding window. The total duration of each aquaculture cycle is T, the size of each sliding window is w, the sliding step size is s, and the window number is k.

[0120] For each window k, the growth slope in window k is calculated using the least squares method:

[0121]

[0122] in, Let the growth slope of the k-th window be . This indicates the start time of the k-th window. Let k be the end time of the k-th window. The average time value for the k-th window. This represents the water production efficiency index at time t under the k-th window. Let represent the average water production efficiency index of the k-th window, and t represent time t.

[0123] The growth trend factor for each breeding cycle is calculated using the growth slope within window k in each breeding cycle.

[0124]

[0125] in, Let K be the growth trend factor, K be the total number of windows in the breeding cycle, and k be the index of the k-th window. Let be the growth slope of the k-th window.

[0126] By comparing the growth trend factors of an aquatic product under different survival conditions during the same aquaculture cycle, the survival condition corresponding to the largest growth trend factor is taken as the optimal survival condition for that aquatic product. Therefore, the growth trend factor for the optimal survival condition is: , For optimal light intensity, The optimal water surface temperature is determined. For example, within the same aquaculture cycle, by simultaneously monitoring the aquatic biomass efficiency curves under different survival conditions—1500 lux / 30℃ in the strong light zone (allowable fluctuation range of 1400~1600 lux, 29.5~30.5℃), 1000 lux / 28℃ in the medium light zone (allowable fluctuation range of 900~1100 lux, 27.5~28.5℃), and 600 lux / 26℃ in the weak light zone (allowable fluctuation range of 500~700 lux, 25.5~26.5℃)—the growth trend factor for each zone is calculated using the least squares method based on a sliding window. For example, in the farming cycle of giant freshwater prawns in Jiangsu, the central light area... =0.032 is significantly higher than β=0.018 in the strong light region and β=0.021 in the weak light region, so the optimal survival conditions for this breeding cycle are 1000 lux and 28℃.

[0127] After determining the optimal survival conditions for aquatic products in each aquaculture cycle, the growth trend stability of the aquatic bioefficiency curve under the optimal survival conditions in each aquaculture cycle is calculated:

[0128]

[0129] in, To ensure the stability of the growth trend, K represents the total number of windows in the breeding cycle, and k represents the index of the k-th window. The growth slope of the k-th window under optimal survival conditions. The growth trend factor under optimal survival conditions. To prevent the denominator from being 0, the value is a constant and takes the value of 0. .

[0130] It should be noted that growth trend stability is a core indicator reflecting the stability of the growth process of aquatic organisms by quantifying the fluctuation of the growth trend of their bioefficiency index curve, under the premise that the optimal living conditions of the aquatic organisms remain unchanged. This indicator directly quantifies the fluctuation intensity of the overall performance of organisms under optimal conditions by calculating the standard deviation of the growth slope of the window. The higher the growth trend stability (e.g., the growth curve is uniform and stable), the less the organism is affected by the living conditions. A low growth trend stability value indicates strong growth trend fluctuation, indicating that the organism is greatly affected by the living conditions and is more sensitive to the growth environment.

[0131] Step S5: Adjust the inertia in the particle swarm optimization algorithm by the growth trend stability, and obtain the optimal state of the photovoltaic panel by the particle swarm optimization algorithm. The optimal state of the photovoltaic panel includes: the optimal photovoltaic panel tilt angle and the optimal reflective film angle.

[0132] The inertia in the particle swarm optimization algorithm is adjusted by the stability of the growth trend. Higher growth trend stability (e.g., a uniform and stable growth curve) indicates that the organism is less affected by living conditions. This can increase the inertia in the particle swarm optimization algorithm, enhance its global search capability, explore a wider range of photovoltaic panel states (tilt angle, reflective film angle), find the global optimum, and avoid getting trapped in local optima. Conversely, lower growth trend stability indicates that the organism is more affected by living conditions and more sensitive to the growth environment (e.g., drastic fluctuations in the growth curve). This can reduce the inertia in the particle swarm optimization algorithm, strengthen its local search accuracy, and quickly converge to a photovoltaic panel state that suits the organism's needs, avoiding deviations from the organism's optimal living conditions due to excessive fluctuations in photovoltaic panel state parameters.

[0133] The inertia in the particle swarm optimization algorithm is adjusted by the stability of the growth trend.

[0134]

[0135] in, The adjusted inertia value, This is the initial inertia value used in the particle swarm optimization algorithm; the default value is 0.7. The adjustment range is set to 0.15 by default. This is the response coefficient, which defaults to 1. For growth trend stability, This is the threshold for growth trend stability, with a default value of 4.

[0136] It should be noted that when the Trend value is 8, the gx value is 0.85; when the Trend value is 2, the gx value is 0.5554. This ensures that the higher the stability of the growth trend, the higher the adjusted inertia value, and the lower the stability of the growth trend, the lower the adjusted inertia value.

[0137] Each particle in the particle swarm optimization algorithm corresponds to the state parameters [QJ, FG] of the photovoltaic panel in the solar-fishery complementary system, where QJ is the tilt angle of the photovoltaic panel (23.5°). QJ 40°), FG is the angle of the reflective film on the photovoltaic panel (30°). FG 40°), particle swarm size N (default value is 60), maximum number of iterations (The default value is 400).

[0138] It should be noted that the photovoltaic panel tilt angle (QJ) and reflective film angle (FG) were chosen as the core control parameters for particle swarm optimization because they have the ability to independently, directly, and precisely control the water surface survival conditions. The tilt angle (QJ) of the photovoltaic panel dominates the macroscopic light transmission adjustment: by changing the tilt angle of the photovoltaic panel, the overall shading rate of the photovoltaic array and the light transmission path of the water layer can be systematically adjusted, directly affecting the basic light intensity of the water surface. For example, every 5° increase in tilt angle can increase the illuminance of the low-light area by 8%-12%, meeting the global light intensity requirements for biological growth. The reflective film angle (FG) achieves local light intensity compensation: by finely adjusting the reflection angle of the reflective film, the irradiance of specific aquaculture areas can be enhanced in a targeted manner. For example, every 10° increase in FG can provide an additional 30-50 lux of light to the medium-light area, accurately correcting local photothermal deviations. The two control the "light transmission baseline" and "light supplementation increment" respectively, and both are low-cost parameters that can be adjusted in real time by existing photovoltaic support systems (modification cost < 1 / 3 of traditional array spacing adjustment). Compared with structural parameters such as height and spacing, they have minute-level dynamic response capabilities, ensuring that the particle swarm algorithm can quickly converge to the optimal photothermal adaptation state.

[0139] The formulas for calculating the state parameters of photovoltaic panels and the light intensity on the water surface are as follows:

[0140]

[0141] in, The intensity of light on the water surface. Natural light intensity, The base transmittance coefficient has a value range of (0.65, 0.75). The tilt angle influence coefficient has a value range of (-0.015, -0.008). is the reflective film supplementary light coefficient, with a value range of (0.003, 0.008), and FG is the reflective film angle of the photovoltaic panel. The tilt angle of the photovoltaic panel.

[0142] The formulas for calculating the state parameters of photovoltaic panels and water surface temperature are as follows:

[0143]

[0144] in, Water surface temperature Atmospheric temperature, This is the light-to-temperature conversion factor, with a default value of 0.0001. Natural light intensity, The base temperature rise coefficient has a value range of (0.8, 1, 2), and FG is the angle of the reflective film on the photovoltaic panel. The tilt angle of the photovoltaic panel. The tilt angle cooling coefficient has a value range of (-0.02, -0.01). The temperature coefficient of the reflective film ranges from 0.005 to 0.012.

[0145] Each day of each aquaculture cycle, the photovoltaic panels are adjusted based on the day's light intensity, temperature, and optimal survival conditions. The water surface temperature and light intensity obtained after the photovoltaic panel adjustment are compared with the optimal survival conditions to obtain a photothermal adaptation value, which is then used as the fitness value in the particle swarm optimization algorithm.

[0146]

[0147] Where SX is the photothermal adaptation value. The intensity of light on the water surface. For optimal light intensity, Water surface temperature The optimal water surface temperature.

[0148] It should be noted that the fitness value (SX) in the particle swarm optimization algorithm quantifies the comprehensive matching degree between the actual water surface environment generated under the control of the current state parameters of the photovoltaic panel (tilt angle QJ, reflective film angle FG) and the optimal growth requirements (optimal light intensity, optimal water temperature) of aquatic organisms. This value integrates the deviations of both light intensity and water temperature in a normalized form. For example: when 1020 lux (optimal value 1000 lux) and If the temperature is 28.5℃ (the optimal value is 28℃), then SX is 0.962. The closer SX is to 1, the better the living conditions match the optimal needs of the organism (e.g., 0.962 indicates a matching degree of 96.2%).

[0149] In particle swarm optimization (PSO), the fitness value (photothermal adaptation value SX) is the core indicator for evaluating the quality of a particle's position. The particle position corresponds to the state parameters of the photovoltaic panel (panel tilt angle, reflective film angle), and the quality of the position is determined by its corresponding fitness value. The closer SX is to 1, the closer the state of the photovoltaic panel at that position is to the optimal survival conditions for aquatic organisms. Velocity determines the magnitude and direction of particle position updates. In the velocity update formula, the individual best position (pbest) is the position with the highest fitness value in the history of each particle, and the global best position (gbest) is the position with the highest fitness value among all particles. Velocity updates need to combine the current velocity, the deviation between the individual best and the current position, and the deviation between the global best and the current position. Therefore, the fitness value indirectly affects the velocity calculation by determining pbest and gbest. Position updates are based on the current position with the updated velocity. In other words, the fitness value determines the direction of iterative optimization of particle positions by guiding the direction and magnitude of the velocity. The global optimal fitness value is calculated as follows: In each iteration, the current fitness value of all particles is compared with the historical global optimal fitness value, and the maximum value between the two is taken as the current global optimal fitness value. This process is repeated until the iteration stops (reaching the maximum number of iterations or the difference between the global optimal fitness values ​​of adjacent iterations is less than 1 × 10⁻⁶). -4 The global optimal fitness value at this point is the final result. Based on the fitness value in the particle swarm optimization algorithm, the velocity and position of each particle in the algorithm are updated:

[0150]

[0151] in, Let wd represent the velocity of the nth particle in the (dd+1)th iteration, and gx be the inertia weight in the particle swarm optimization algorithm, which controls the degree to which the particle retains its historical velocities. This represents the velocity of the wd-th feature of the nth particle in dd iterations. and This represents the individual learning factor and the social learning factor in the particle swarm optimization algorithm, and is usually taken as 2. and All numbers are random to avoid local optima in the particle swarm optimization algorithm. The wd-th eigenvalue represents the wd-th eigenvalue of the historical best position of the n-th particle. Let wd be the wd-th eigenvalue of the currently found global optimum position for all particles in the particle swarm optimization algorithm. This represents the position of the wd-th feature of the nth particle in dd iterations.

[0152]

[0153] in, This represents the position of the wd-th feature of the n-th particle in the (dd+1)-th iteration. This represents the position of the wd-th feature of the nth particle in dd iterations. This represents the velocity of the wd-th feature of the n-th particle in the (dd+1)-th iteration.

[0154] By continuously updating particle positions, the search continues for particles that maximize the photothermal fit. This process continues until the maximum number of iterations is reached. When the iteration stops or the difference between the global optimal fitness values ​​(i.e., the maximum photothermal fitness value SX) of all particles in adjacent iterations is less than 1× When the particle swarm optimization algorithm is considered to have reached a stable convergence state, iteration is stopped. After iteration stops, the photovoltaic panel tilt angle and reflective film angle corresponding to the photovoltaic panel state parameter with the largest photothermal adaptation value are taken as the optimal photovoltaic panel tilt angle. Optimal reflective film angle .

[0155] Step S6: Predict the power generation of the photovoltaic panels in the optimal state for each aquaculture cycle, obtain the predicted power generation value, and provide adjustment strategies for aquaculture based on the predicted power generation value to achieve coordinated management of fishery-solar complementarity.

[0156] The photovoltaic panels are adjusted daily during each breeding cycle to achieve optimal performance. Power generation is then predicted based on this optimal state, using the optimal tilt angle to obtain the first predicted power generation value.

[0157]

[0158] in, This is the first predicted power generation value. For power generation duration, For the power generation efficiency of photovoltaic panels, This represents the total solar radiation. The tilt efficiency coefficient has a value range of (0.010, 0.025). To achieve the optimal tilt angle for photovoltaic panels, The latitude is the local latitude.

[0159] It should be noted that the first power generation prediction is based on quantifying the gain of the optimal photovoltaic panel tilt angle on the basic power generation efficiency, predicting the theoretical power generation before the addition of a reflective film for compensation. For example: the photovoltaic panel power generation efficiency is 18.5%, and the total solar radiation is 5.2 kWh / With a power generation duration of 10 hours, an optimal photovoltaic panel tilt angle of 38°, a local latitude of 34°, and a tilt efficiency coefficient of 0.020, the first predicted power generation value is 10.39 kWh.

[0160] By determining the optimal reflective film angle, we can obtain the compensation amount for power generation based on the optimal reflective film angle:

[0161]

[0162] Where BC is the compensation amount for power generation based on the optimal reflective film angle. This is the radiation gain coefficient, with a value range of (0.002, 0.003). This is the cooling gain coefficient, with a value ranging from (0.0001, 0.0003). For the optimal reflective film angle, Atmospheric temperature, For reference temperature, the default is 25℃.

[0163] It should be noted that the optimal reflective film angle has a certain compensation effect on power generation. To increase radiation gain, the reflective film directly reflects sunlight onto the photovoltaic panel surface, increasing the effective power generation radiation. To reduce heat buildup in photovoltaic panels and improve power generation efficiency, reflective films are used to reduce heat accumulation in the panels. The maximum value is 35°C to avoid excessive contribution of extreme high temperature to BC. For example, when the optimal reflective film angle is 40°, the atmospheric temperature on that day is 35°C, the radiation gain coefficient is 0.003, and the cooling gain coefficient is 0.0003, then the compensation amount of the optimal reflective film angle to the power generation is 0.24. This value (0.24) means that because the reflective film is adjusted to 40°, the predicted value of the first power generation on that day is increased by an additional 24%.

[0164] The final power generation forecast is:

[0165]

[0166] in, This is the predicted value for power generation. BC represents the first predicted power generation value, and BC represents the compensation amount for power generation based on the optimal reflective film angle.

[0167] In the solar-aquaculture complementary management system, aquaculture adjustment strategies are dynamically generated based on predicted power generation values.

[0168] In scenarios with ample power (predicted power generation exceeding the threshold), priority is given to enhancing oxygenation, automatically increasing the operating power of aerators (up to +30%), extending equipment working hours to 16 hours per day, accelerating the recovery of dissolved oxygen in the water, and precharging surplus power to the energy storage system (reserving ≥20% capacity) to provide emergency support for sudden rainy weather.

[0169] In scenarios of power shortage (predicted power generation is below the threshold), non-core sensors (such as ultraviolet radiation monitoring) are shut down, and power is concentrated to ensure the basic operation of aerators. If necessary, energy storage batteries can be used to supplement power for 2-4 hours.

[0170] In a power balance scenario (predicted power generation is in the middle range), a steady-state maintenance strategy is employed to maintain the photovoltaic panel tilt angle and reflective film angle at their optimal values, and to monitor dissolved oxygen (DO) and temperature fluctuations in the water every 8 hours.

[0171] This paper proposes a dynamic collaborative optimization method for photovoltaic panels in a fishery-solar hybrid system that integrates a biological growth response feedback mechanism. By constructing a biological efficiency-driven collaborative optimization framework, a quantitative correlation model between survival condition factors and the growth efficiency of aquatic organisms is established, and the state parameters of the photovoltaic panels are dynamically adjusted based on real-time biological responses.

[0172] The combination of recursive feature selection from random forests and TreeSHAP significantly improves computational efficiency by using multi-decision-tree ensembles to evaluate feature importance and dynamically eliminate redundant and noisy features, thus achieving adaptive compression of high-dimensional biological characteristic data. TreeSHAP, based on game theory to analyze the decision-making mechanism of random forests, precisely reveals the direction and interaction patterns of the core characteristics' influence on biological efficacy by quantifying feature SHAP values. The two methods form a closed loop. Recursive selection provides a denoised feature subset for TreeSHAP, while TreeSHAP, through weight allocation and attribution analysis, verifies the scientific validity of the feature selection, jointly providing a foundation for constructing the aquatic biological efficacy index.

[0173] By combining the growth deviation of each feature with the corresponding TreeSHAP feature weight, an aquatic organism efficiency index is calculated, enabling a dynamic quantitative assessment of the organism's growth state and adaptability to living conditions. The index value reflects the comprehensive growth efficiency level of aquatic organisms under a specific combination of environmental parameters, integrating multidimensional environmental interaction effects and biological response characteristics, and providing a precise feedback indicator that can directly drive parameter adjustment for subsequent particle swarm optimization.

[0174] By adjusting the inertia weights in the Particle Swarm Optimization (PSO) algorithm based on the stability of the growth trend, the parameter optimization efficiency and convergence stability are significantly improved. When the growth trend stability is high, the inertia weights are increased to enhance the global search capability; when the stability is low, the inertia weights are decreased to enhance the accuracy of local exploration.

[0175] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0176] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A collaborative management method for photovoltaic power stations integrating fisheries and solar power, characterized in that: Includes the following steps: Step S1: Collect aquatic organism growth data under different living conditions and preprocess it to obtain aquatic organism growth data; Step S2: The aquatic organism growth data is filtered for feature importance using a recursive feature selection method based on random forest to obtain the filtered second aquatic organism growth data; the feature weight of each feature in the second aquatic organism growth data is calculated using the TreeSHAP method, including the following specific steps: The aquatic organism growth data were selected based on feature importance using a recursive feature selection method based on random forest. The aquatic organism growth data is X={ , ,... ,..., }, construct a decision tree set T={ , ,..., ,..., There are a total of Tree; For each tree and characteristics Calculate the cumulative purity increase for each feature across all split points: ； in, This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree. For trees The set of split points, h is In the context of a split point h, I() is an indicator function, i.e., the characteristic index of the split point h. The index p is 1 when it is equal to the index of the aquatic organism growth data feature, and 0 when they are not. The feature index of the split point h, The reduction in Gini impurity at the splitting point h. , Let the Gini impurity be the parent node of the split point h. The child node of the split point h, Let be the Gini impurity of the child node `child` of the split point `h`. and , which represent the number of samples for the parent and child nodes of the split point h, respectively, and p is the index of the aquatic organism growth data feature; Then features The feature importance in the random forest model is as follows: ； in, The p-th feature of aquatic organism growth data In a random forest model, feature importance is denoted by M, where M is the total number of decision trees and m is the m-th decision tree. This represents the p-th feature in the aquatic organism growth data. The importance of features in the m-th decision tree; Based on the feature importance of all features in the aquatic organism growth data in the random forest model, the features with the lowest feature importance are removed to form a new feature subset. The random forest model is retrained with the new feature subset, and the feature importance is repeatedly calculated until the number of remaining features reaches the preset feature number threshold. The second set of aquatic organism growth data after filtering is then output. The SHAP value of each feature of the second aquatic organism growth data was calculated using the TReeSHAP method: ； in, Let be the SHAP value of the p-th feature in the second aquatic organism growth data, M be the total number of decision trees, m be the m-th decision tree, and h be the split node in the decision tree. Indicates that node h is affected by features Number of samples affected This is the p-th feature in the second set of aquatic organism growth data. This represents the total number of samples reaching node h. The reduction in Gini impurity at the split point h; Calculate the weight of each feature in the second aquatic organism growth data based on the SHAP value of each feature: ； in, Let be the weight of the p-th feature in the second aquatic organism growth data. P1 is the SHAP value of the p-th feature in the second aquatic organism growth data, and P2 is the total number of features in the second aquatic organism growth data. Step S3: By calculating the deviation between the second aquatic organism growth data and the preset growth threshold range, the growth deviation of each feature in the second aquatic organism growth data is obtained; by combining the growth deviation of each feature and the corresponding feature weight, the aquatic organism efficiency index is calculated; based on the aquatic organism efficiency index, the aquatic organism efficiency curve is plotted, including the following steps: The growth deviation of each feature in the second aquatic organism growth data at each moment in each aquaculture cycle is calculated based on the deviation between the second aquatic organism growth data and the preset growth threshold. ； in, The bias of the p-th feature in the second aquatic organism growth data. This is the p-th feature in the second set of aquatic organism growth data. The center value of the preset growth threshold interval for the p-th feature in the second aquatic organism growth data is given. This is the upper limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data. The lower limit of the preset growth threshold range for the p-th feature in the second aquatic organism growth data; Step S4: Calculate the growth trend factor of the aquatic organism efficiency curve using the least squares method based on a sliding window, obtain the optimal survival conditions of aquatic organisms based on the growth trend factor, and calculate the growth trend stability of the aquatic organism efficiency curve corresponding to the optimal survival conditions. Step S5: Adjust the inertia in the particle swarm optimization algorithm through the growth trend stability, and obtain the optimal state of the photovoltaic panel through the particle swarm optimization algorithm. The optimal state of the photovoltaic panel includes: the optimal photovoltaic panel tilt angle and the optimal reflective film angle. Step S6: Predict the power generation of the photovoltaic panels in the optimal state for each aquaculture cycle, obtain the predicted power generation value, and provide adjustment strategies for aquaculture based on the predicted power generation value to achieve coordinated management of fishery-solar complementarity.

2. The collaborative management method for photovoltaic power stations integrating fisheries and solar power as described in claim 1, characterized in that: The process of calculating the aquatic organism efficiency index by combining the growth deviation degree of each feature and its corresponding feature weight includes the following steps: By combining the growth deviation of each feature with its corresponding feature weight, the biomass efficiency index of the aquaculture process at each moment in each aquaculture cycle is calculated: ； in, P2 represents the aquatic organism efficiency index, and P2 represents the total number of features in the second aquatic organism growth data. Let p be the weight of the p-th feature in the second aquatic organism growth data. The deviation of the p-th feature in the second aquatic organism growth data.

3. The collaborative management method for photovoltaic power stations integrating fisheries and solar power as described in claim 2, characterized in that: The calculation of the growth trend factor of the water production efficiency curve using the least squares method based on a sliding window includes the following specific steps: For each window k, the growth slope in window k is calculated using the least squares method: ； in, Let the growth slope of the k-th window be . This indicates the start time of the k-th window. Let k be the end time of the k-th window. The average time value for the k-th window. This represents the water production efficiency index at time t under the k-th window. represents the average water production efficiency index of the k-th window, and t represents time t; The growth trend factor for each breeding cycle is calculated using the growth slope within window k in each breeding cycle. ； in, Let K be the growth trend factor, K be the total number of windows in the breeding cycle, and k be the index of the k-th window. Let be the growth slope of the k-th window.

4. The collaborative management method for photovoltaic power station fishery-solar complementary systems according to claim 3, characterized in that: The process of obtaining the optimal survival conditions for aquatic organisms based on the growth trend factor and calculating the growth trend stability of the aquatic organism performance curve corresponding to the optimal survival conditions includes the following specific steps: The optimal survival conditions for aquatic organisms are obtained based on the growth trend factor, and the growth trend stability of the aquatic organism efficiency curve corresponding to the optimal survival conditions is calculated. After determining the optimal survival conditions for aquatic products in each aquaculture cycle, the growth trend stability of the aquatic bioefficiency curve under the optimal survival conditions in each aquaculture cycle is calculated: ； in, To ensure the stability of the growth trend, K represents the total number of windows in the breeding cycle, and k represents the index of the k-th window. The growth slope of the k-th window under optimal survival conditions. The growth trend factor under optimal survival conditions. To prevent the denominator from being 0, the value is a constant and takes the value of 0. .

5. The collaborative management method for photovoltaic power station fishery-solar complementary systems according to claim 4, characterized in that: The adjustment of inertia in the particle swarm optimization algorithm based on the stability of the growth trend includes the following specific steps: The inertia in the particle swarm optimization algorithm is adjusted by the stability of the growth trend. ； in, The adjusted inertia value, This is the initial inertia value used in the particle swarm optimization algorithm; the default value is 0.

7. The adjustment range is set to 0.15 by default. This is the response coefficient, which defaults to 1. For growth trend stability, This is the threshold for growth trend stability, with a default value of 4.

6. The collaborative management method for photovoltaic power station fishery-solar complementary systems according to claim 5, characterized in that: The process of obtaining the optimal state of the photovoltaic panel using the particle swarm optimization algorithm includes the following specific steps: The formulas for calculating the state parameters of photovoltaic panels and the light intensity on the water surface are as follows: ； in, The intensity of light on the water surface. Natural light intensity, The base transmittance coefficient has a value range of (0.65, 0.75). The tilt angle influence coefficient has a value range of (-0.015, -0.008). is the reflective film supplementary light coefficient, with a value range of (0.003, 0.008), and FG is the reflective film angle of the photovoltaic panel. The tilt angle of the photovoltaic panel; The formulas for calculating the state parameters of photovoltaic panels and water surface temperature are as follows: ； in, Water surface temperature Atmospheric temperature, This is the light-to-temperature conversion factor, with a default value of 0.0001. Natural light intensity, The base temperature rise coefficient has a value range of (0.8, 1, 2). Where is the tilt angle of the photovoltaic panel, and FG is the angle of the reflective film on the photovoltaic panel. The tilt angle cooling coefficient has a value range of (-0.02, -0.01). The temperature coefficient of the reflective film ranges from 0.005 to 0.

012. Each day of each aquaculture cycle, the photovoltaic panels are adjusted based on the day's light intensity, temperature, and optimal survival conditions. The water surface temperature and light intensity obtained after the photovoltaic panel adjustment are compared with the optimal survival conditions to obtain a photothermal adaptation value, which is then used as the fitness value in the particle swarm optimization algorithm. ； Where SX is the photothermal adaptation value. The intensity of light on the water surface. For optimal light intensity, Water surface temperature To achieve the optimal water surface temperature; Based on the fitness value in the particle swarm optimization algorithm, the velocity and position of each particle in the algorithm are updated: ； in, Let wd represent the velocity of the nth particle in the (dd+1)th iteration, and gx be the inertia weight in the particle swarm optimization algorithm, which controls the degree to which the particle retains its historical velocities. This represents the velocity of the wd-th feature of the nth particle in dd iterations. and This represents the individual learning factor and the social learning factor in the particle swarm optimization algorithm, and is usually taken as 2. and All numbers are random to avoid local optima in the particle swarm optimization algorithm. The wd-th eigenvalue represents the wd-th eigenvalue of the historical best position of the n-th particle. Let wd be the wd-th eigenvalue of the currently found global optimum position for all particles in the particle swarm optimization algorithm. This represents the position of the wd-th feature of the nth particle in dd iterations; ； in, This represents the position of the wd-th feature of the n-th particle in the (dd+1)-th iteration. This represents the position of the wd-th feature of the nth particle in dd iterations. This represents the velocity of the wd-th feature of the nth particle in the (dd+1)th iteration; By continuously updating particle positions, the search continues for particles that maximize the photothermal fit. This process continues until the maximum number of iterations is reached. When the iteration stops or the difference between the global optimal fitness value (i.e., the maximum photothermal fitness value SX) of all particles in adjacent iterations is less than 1× When the particle swarm optimization algorithm is considered to have reached a stable convergence state, it stops iterating. After stopping iteration, the photovoltaic panel tilt angle and reflective film angle corresponding to the photovoltaic panel state parameter with the largest photothermal adaptation value are taken as the optimal photovoltaic panel tilt angle. Optimal reflective film angle .

7. The collaborative management method for photovoltaic power station fishery-solar complementary systems according to claim 6, characterized in that: The process of predicting the power generation of photovoltaic panels under optimal conditions for each breeding cycle, to obtain the predicted power generation value, includes the following specific steps: The photovoltaic panels are adjusted daily during each breeding cycle to achieve optimal performance. Power generation is then predicted based on this optimal state, using the optimal tilt angle to obtain the first predicted power generation value. ； in, This is the first predicted power generation value. For power generation duration, For the power generation efficiency of photovoltaic panels, This represents the total solar radiation. The tilt efficiency coefficient has a value range of (0.010, 0.025). To achieve the optimal tilt angle for photovoltaic panels, The latitude is the local latitude. By determining the optimal reflective film angle, we can obtain the compensation amount for power generation based on the optimal reflective film angle: ； Where BC is the compensation amount for power generation based on the optimal reflective film angle. This is the radiation gain coefficient, with a value range of (0.002, 0.003). This is the cooling gain coefficient, with a value ranging from (0.0001, 0.0003). For the optimal reflective film angle, Atmospheric temperature, For reference temperature, the default value is 25℃; The final power generation forecast is: ； in, This is the predicted value for power generation. BC represents the first predicted power generation value, and BC represents the compensation amount for power generation based on the optimal reflective film angle.

Images