Nutrient solution concentration regulation method and system for potato aeroponics

Abstract

The application discloses a kind of for potato mist culture nutrient solution concentration regulation method and system, it is related to fertilization technical field, including: extracting root system characteristic data and concentration characteristic data;Based on root system characteristic data and concentration characteristic data, establish concentration trend prediction model, and utilize concentration trend prediction model to predict the concentration variation trend of nutrient solution under different conditions;The concentration variation trend of nutrient solution under different conditions is analyzed using feature analysis algorithm, and the key factor that influences the stability of nutrient solution concentration is identified;And based on key factor, dynamically adjust nutrient solution.This application identifies the key factor that influences concentration stability by feature analysis algorithm, and dynamically regulates nutrient solution proportion accordingly, so as to realize the intelligent management and accurate supply of nutrient solution concentration, provide effective support for the whole cycle of potato mist culture nutrition optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fertilization, in particular to a method and system for regulating the concentration of nutrient solution for potato aeroponics. BACKGROUND

[0002] In recent years, soilless culture technology has been widely used, among which aeroponics as a new type of cultivation method gradually shows its unique advantages in potato planting. Aeroponics realizes efficient recycling of nutrient solution by atomizing the nutrient components dissolved in water and spraying them to the plant roots through pipelines. In the process of potato aeroponics, the regulation of the concentration of nutrient solution is crucial because the growth of potato roots and the formation of tubers have strict requirements for the concentration of nutrient solution. Potatoes at different growth stages have different requirements for the concentration of nutrient solution, for example, during the vegetative growth period, lower concentration is beneficial to promote root development, while during the tuber bulking period, appropriate increase in the concentration of nutrient solution can promote the growth of tubers. In order to realize accurate regulation of the concentration, the aeroponics system is usually equipped with real-time monitoring devices, which can dynamically adjust the concentration of nutrient solution according to the growth conditions of potatoes and environmental conditions, to ensure that plants can obtain the best nutrient supply.

[0003] However, the existing technology relies more on manual experience, lacks real-time monitoring of root characteristics and nutrient solution components, resulting in lag in concentration regulation; empirical proportioning is difficult to adapt to the needs of different growth stages, which is easy to cause nutrient imbalance; and it cannot predict the impact of environmental changes on the concentration, resulting in insufficient regulation accuracy, thereby affecting the nutrient supply effect and yield stability of potato aeroponics.

[0004] In view of the problems in the related art, no effective solution has been proposed so far. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a method and system for regulating the concentration of nutrient solution for potato aeroponics, which solves the problems of the prior art that the concentration regulation is lagging due to the reliance on manual experience and the lack of real-time monitoring of root characteristics and nutrient solution components; empirical proportioning is difficult to adapt to the needs of different growth stages, which is easy to cause nutrient imbalance; and it cannot predict the impact of environmental changes on the concentration, resulting in insufficient regulation accuracy, thereby affecting the nutrient supply effect and yield stability of potato aeroponics.

[0006] To achieve the above purpose, the present application is realized by the following technical solutions:

[0007] According to one aspect of the present application, a method for regulating the concentration of nutrient solution for potato aeroponics is provided, comprising:

[0008] The root system temperature distribution image of the potato and the concentration data of each component of the nutrient solution are acquired, and root system feature data and concentration feature data are extracted;

[0009] Based on the root system feature data and the concentration feature data, a concentration trend prediction model is established, and the concentration trend prediction model is used to predict the concentration change trend of the nutrient solution under different conditions;

[0010] The concentration change trend of the nutrient solution under different conditions is analyzed by using a feature analysis algorithm, and the key factors affecting the stability of the concentration of the nutrient solution are identified; and based on the key factors, the nutrient solution is dynamically adjusted.

[0011] Further, the root system temperature distribution image of the potato and the concentration data of each component of the nutrient solution are acquired, and root system feature data and concentration feature data are extracted, including:

[0012] The repeated data, missing values and abnormal values of the acquired root system temperature distribution image and the concentration data of each component of the nutrient solution are denoised, filtered and smoothed to obtain complete root system temperature distribution image and concentration data of each component of the nutrient solution;

[0013] Initialize the data model, extract the high-frequency feature indexes in the temperature distribution image and the concentration data, and sort them;

[0014] Construct an encoding mapping dictionary, and perform differential compression and feature value encoding on the extracted high-frequency features;

[0015] Based on the high-frequency feature sequence, a feature data tree is constructed to record the combination path information of the temperature distribution image and the concentration data;

[0016] From the bottom of the feature data tree, a combination path is generated, key feature patterns are extracted, and the feature set is updated;

[0017] Statistical analysis is performed on the feature set, and root system feature data and concentration feature data are extracted.

[0018] Further, based on the root system feature data and the concentration feature data, a concentration trend prediction model is established, and the concentration trend prediction model is used to predict the concentration change trend of the nutrient solution under different conditions, including:

[0019] The root system feature data and the concentration feature data are divided, and a training set and a validation set are constructed;

[0020] The key parameters of the concentration trend prediction model are initialized, and an initial solution is randomly generated in the parameter search space as the initialization structure configuration of the concentration trend model;

[0021] Input the training set into the initial concentration trend prediction model to calculate errors as initial fitness values of the concentration trend prediction model, and evaluate the performance of the initial concentration trend prediction model by using the validation set, and optimize the parameters of the initial concentration trend prediction model through multiple iterations to obtain a final concentration trend prediction model.

[0022] Optimize the final concentration trend prediction model by using a model optimization algorithm, and predict the concentration change of the nutrient solution under different conditions by using the optimized concentration trend prediction model, and output the concentration change trend of the nutrient solution under different conditions.

[0023] Further, optimize the final concentration trend prediction model by using a model optimization algorithm, and predict the concentration change of the nutrient solution under different conditions by using the optimized concentration trend prediction model, and output the concentration change trend of the nutrient solution under different conditions, including:

[0024] Randomly initialize the structure parameters of the concentration trend prediction model as the initial solution of the current optimal concentration trend prediction model;

[0025] Iteratively test the influence of each structure parameter combination on the prediction accuracy, and retain the structure parameter combination of the concentration trend prediction model with the smallest error;

[0026] Determine whether the structure parameter combination of the current concentration trend prediction model meets the error constraint, if not, replace the current concentration trend prediction model and continue to test;

[0027] Adjust the structure parameters of the concentration trend prediction model by using a perturbation algorithm to jump out of the local optimum and continue to perform iterative optimization search;

[0028] Compare the error performance of the concentration trend prediction model before optimization and the concentration trend prediction model after optimization, if the error value of the concentration trend prediction model after optimization is smaller than the error value of the concentration trend prediction model before optimization, update the structure parameters of the concentration trend prediction model after optimization;

[0029] If the maximum number of iterations is reached, output the optimized concentration trend prediction model to predict the concentration change of the nutrient solution under different conditions, and output the concentration change trend of the nutrient solution under different conditions.

[0030] Further, adjust the structure parameters of the concentration trend prediction model by using a perturbation algorithm to jump out of the local optimum and continue to perform iterative optimization search, including:

[0031] Initialize the structure parameters of the concentration trend model, generate a candidate structure set and calculate the prediction error as the fitness;

[0032] Select the target structure combination according to the fitness ranking, perform simulated crossover and replace the current worst structure combination;

[0033] The candidate structure combination set is subjected to mutation operation, and the local search capability is enhanced in combination with a simulated annealing mechanism;

[0034] The fitness of the new structure combination is evaluated, and if it is better than the current optimal structure, it is replaced, otherwise the original structure is retained and iterative search continues;

[0035] The structure disturbance and fitness update process is executed in a loop, constantly jumping out of the local optimum and continuing to perform iterative optimization search.

[0036] Further, the mutation operation is performed on the candidate structure combination set, and the local search capability is enhanced in combination with a simulated annealing mechanism, including:

[0037] The current structure combination is randomly selected and the fitness of the current structure combination is calculated, and a new candidate structure is generated through mutation operation;

[0038] The new fitness value is calculated according to the mutated structure combination, and it is judged whether it is better than the current solution, if the new structure solution is better than the current solution, the current solution is replaced; otherwise, it is determined whether to accept the worst solution through the simulated annealing mechanism;

[0039] The temperature descending operation of simulated annealing is executed, the local solution is continuously optimized, and the local search capability is enhanced.

[0040] Further, the feature analysis algorithm is used to analyze the concentration change trend of the nutrient solution under different conditions, and the key factors affecting the stability of the nutrient solution concentration are identified; and based on the key factors, the nutrient solution is dynamically adjusted, including:

[0041] The concentration data of the nutrient solution under different environmental conditions is obtained and is subjected to Boolean discretization, and a concentration change trend analysis feature set is constructed;

[0042] The candidate set of factors affecting the concentration change trend is initialized, and the initial weight of each factor is set;

[0043] Randomly select concentration data samples under different environmental conditions, and calculate the correlation coefficient of each influencing factor and concentration fluctuation;

[0044] Through the importance evaluation algorithm, the environmental parameters and operating parameters related to the concentration stability are screened out;

[0045] A concentration change regression model under multiple environmental conditions is established to quantify the contribution of each influencing factor to the concentration fluctuation;

[0046] The stability performance of each influencing factor in different environmental combinations is verified, and the core influencing factor set is determined, and the core influencing factor is taken as the key factor affecting the stability of the nutrient solution concentration;

[0047] Based on the key factors, the spray frequency, the nutrient solution supply rhythm and the nutrient solution component proportion are dynamically adjusted, so that the concentration of the nutrient solution in the potato mist culture process is accurately controlled and optimized.

[0048] Further, through the importance evaluation algorithm, the environmental parameters and the operation parameters related to the concentration stability are screened out, including:

[0049] A plurality of candidate parameter combinations are randomly selected in the environmental parameter space as an initial analysis set;

[0050] The influence weight of each parameter combination on the concentration stability of the nutrient solution is calculated as an importance evaluation index;

[0051] The parameter combination having the greatest influence on the concentration stability is selected as a reference group;

[0052] The weight coefficient and the correlation degree score of each parameter are updated, if the influence of the new parameter combination on the stability is better than that of the reference group, the reference group is updated, otherwise, the new candidate parameter is supplemented in the parameter space;

[0053] If the maximum iteration number is reached, the environmental parameters and the operation parameters related to the concentration stability are screened out.

[0054] Further, the influence weight of each parameter combination on the concentration stability of the nutrient solution is calculated as an importance evaluation index, including:

[0055] A mapping relationship model between the concentration change of the nutrient solution and the parameter combination is constructed, and the concentration fluctuation response characteristics are extracted;

[0056] The parameter combinations are disturbed and simulated, and the change trend of the concentration stability index under different parameter combinations is recorded;

[0057] According to the concentration fluctuation characteristics and the parameter combination disturbance results, the influence weight of each parameter combination is calculated as an importance evaluation index.

[0058] According to another aspect of the present application, a nutrient solution concentration control system for potato mist culture is also provided, which comprises:

[0059] A data management module is used to obtain the root temperature distribution image of the potato and the concentration data of each component of the mist culture nutrient solution, and extract the root feature data and the concentration feature data;

[0060] A trend prediction module is used to establish a concentration trend prediction model based on the root feature data and the concentration feature data, and predict the concentration change trend of the nutrient solution under different conditions by using the concentration trend prediction model;

[0061] The trend analysis and regulation module is configured to analyze the concentration change trend of the nutrient solution under different conditions by using a feature analysis algorithm, identify key factors affecting the stability of the concentration of the nutrient solution, and dynamically adjust the nutrient solution based on the key factors.

[0062] The present application has the following advantages:

[0063] 1. The present application can accurately obtain the characteristics of potato root system and nutrient solution concentration, establish a concentration trend prediction model, and realize early prediction of the concentration change of the nutrient solution under different environmental conditions. By using a feature analysis algorithm to identify key factors affecting the stability of the concentration, and dynamically adjusting the ratio of the nutrient solution, the scientificity and pertinence of the concentration regulation are improved, thereby realizing intelligent management and accurate supply of the concentration of the nutrient solution, and providing effective support for the nutrient optimization of the whole cycle of potato mist culture.

[0064] 2. The present application constructs a concentration trend prediction model, and realizes self-adaptive adjustment of the structure parameters by combining with a model optimization algorithm, thereby improving the accuracy and stability of the concentration change trend prediction. By using a disturbance mechanism and a simulated annealing strategy, the local optimum is effectively avoided, and the concentration of the nutrient solution is dynamically predicted and responded under various environmental conditions, thereby providing systematic support for realizing accurate regulation of the concentration of the nutrient solution, and being suitable for the concentration management demand in the complex environment of potato mist culture.

[0065] 3. The present application identifies key factors affecting the stability of the concentration of the nutrient solution by using a feature analysis algorithm, realizes accurate modeling and quantitative explanation of the concentration change trend by combining data analysis and regression modeling under multiple environmental conditions, dynamically adjusts the spraying frequency, supply rhythm and nutrient ratio, and helps to realize accurate regulation of the concentration of the nutrient solution, thereby improving the systematicness and responsiveness of the concentration regulation process, and providing stable and reliable nutrient supply strategy support for potato mist culture. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0067] Figure 1 is a flow chart of a nutrient solution concentration regulation method for potato mist culture according to an embodiment of the present application;

[0068] Figure 2 is a principle block diagram of a nutrient solution concentration regulation system for potato mist culture according to an embodiment of the present application.

[0069] In the drawings:

[0070] 1, data management module; 2, trend prediction module; 3, trend analysis and regulation module. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0072] In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more. In addition, the terms "first", "second", "third" and the like are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0073] According to the embodiments of the present application, a nutrient solution concentration regulation method and system for potato mist culture are provided.

[0074] The present application will be further described in conjunction with the drawings and specific embodiments. Figure 1 As shown in the drawings, the nutrient solution concentration regulation method for potato mist culture according to the embodiments of the present application comprises:

[0075] S1, obtaining the root temperature distribution image of potato and the concentration data of each component of the mist culture nutrient solution, and extracting the root feature data and the concentration feature data;

[0076] Specifically, the concentration data of each component can be obtained by using a nutrient solution component detection device (including a spectrometer, an electrochemical sensor, etc.), and the three-dimensional structure image of the root system can be obtained by using a radar detector, and the temperature distribution image of the root system can be obtained by using an infrared camera.

[0077] Specifically, the concentration data includes:

[0078] Macronutrients: nitrogen (N), phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), sulfur (S), etc. ,

[0079] Micronutrients: iron (Fe), zinc (Zn), manganese (Mn), copper (Cu), boron (B), molybdenum (Mo), etc.

[0080] Auxiliary indicators: EC value (conductivity), pH value, etc.

[0081] ​​​​​​Specifically, the concentration characteristic data includes concentration variation trend (time series), fluctuation amplitude and stability index of different element concentrations, correlation between components (such as the interaction between potassium and calcium), concentration regulation sensitivity (response intensity of concentration change to root system state), and stage concentration characteristics (characteristic mode of different growth stages).

[0082] Specifically, the root system characteristic data includes root system temperature distribution (average value, gradient, fluctuation), root system vitality index (respiration rate related temperature characteristics), root system area temperature and humidity coupling characteristics, root system spatial distribution characteristics (root length density, coverage range), and root system temperature and nutrient solution concentration response characteristics (temperature and absorption rate relationship).

[0083] S2, based on the root system characteristic data and the concentration characteristic data, a concentration trend prediction model (i.e. a deep belief network model DBN) is established, and the concentration trend prediction model is used to predict the concentration variation trend of the nutrient solution under different conditions;

[0084] S3, using a feature analysis algorithm to analyze the concentration variation trend of the nutrient solution under different conditions, identifying the key factors affecting the stability of the nutrient solution concentration; and based on the key factors, dynamically adjusting the nutrient solution.

[0085] Specifically, the key factors affecting the stability of the nutrient solution concentration include environmental parameter type key factors and operation parameter type key factors. The environmental parameter type key factors include root system temperature distribution, air humidity and temperature, light intensity and photoperiod, mist chamber gas composition (such as 、 ) concentration, pH value and EC value of the nutrient solution itself, etc. The operation parameter type key factors include spray frequency and duration, replenishment period and replenishment amount, nutrient solution formula adjustment strategy, and nutrient solution circulation system settings, etc.

[0086] In this optional embodiment, the root system temperature distribution image of the potato and the concentration data of each component of the nutrient solution are obtained, and the root system characteristic data and the concentration characteristic data are extracted, including:

[0087] The repeated data, missing values and abnormal values of the obtained root system temperature distribution image and the concentration data of each component of the nutrient solution are denoised, filtered and smoothed to obtain complete root system temperature distribution image and concentration data of each component of the nutrient solution;

[0088] Initialize the data model, extract high-frequency characteristic indexes in the temperature distribution image and the concentration data and sort them;

[0089] Construct an encoding mapping dictionary to differentially compress and encode the extracted high-frequency features;

[0090] constructing a feature data tree based on the high-frequency feature sequence, recording the combined path information of the temperature distribution image and the concentration data;

[0091] traversing from the bottom of the feature data tree, generating a combined path, extracting a key feature mode, and updating the feature set;

[0092] performing statistical analysis on the feature set to extract root system feature data and concentration feature data.

[0093] Specifically, first, the temperature distribution image of potato root system and the concentration data of each component of the nutrient solution are obtained, and the data is denoised, filtered and smoothed to remove duplicate data, missing values and outliers, ensuring the integrity of the data. Then, the data model is initialized, the high-frequency feature indicators in the temperature distribution and concentration data are extracted, and they are sorted according to importance. An encoding mapping dictionary is constructed to compress the high-frequency feature difference and encode the feature value. Then, a feature data tree is constructed based on the high-frequency feature sequence to record the combined path information of the temperature and concentration data. By traversing the bottom of the feature data tree, a combined path is generated and a key feature mode is extracted, and finally the feature set is statistically analyzed to extract root system feature data and concentration feature data, providing a basis for further regulation and prediction. Thus, the data processing accuracy and model prediction ability are improved.

[0094] In this optional embodiment, based on the root system feature data and the concentration feature data, a concentration trend prediction model is established, and the concentration trend prediction model is used to predict the concentration change trend of the nutrient solution under different conditions, including:

[0095] Divide the root system feature data and the concentration feature data, and construct a training set and a validation set;

[0096] Initialize the key parameters of the concentration trend prediction model, and randomly generate an initial solution in the parameter search space as the initial structure configuration of the concentration trend model;

[0097] Input the training set into the initial concentration trend prediction model to calculate the error as the initial fitness value of the concentration trend prediction model, and use the validation set to evaluate the performance of the initial concentration trend prediction model, and optimize the parameters of the initial concentration trend prediction model through multiple iterations to obtain the final concentration trend prediction model;

[0098] Optimize the final concentration trend prediction model using a model optimization algorithm, and use the optimized concentration trend prediction model to predict the concentration change of the nutrient solution under different conditions, and output the concentration change trend of the nutrient solution under different conditions.

[0099] Specifically, first, the root system characteristic data and the concentration characteristic data are divided to construct a training set and a validation set to ensure the representativeness and integrity of the data set. Then, the key parameters of the concentration trend prediction model are initialized, and an initial solution is randomly generated in the parameter search space as the initial configuration of the concentration trend model. The training set is input into the initial model, the error is calculated, and the initial fitness value is obtained, and then the initial model performance is evaluated using the validation set. Through multiple iterations of optimizing the initial model parameters, the final concentration trend prediction model is gradually obtained. Finally, the final model is further optimized using a model optimization algorithm, and the concentration change of the nutrient solution under different conditions is predicted using the optimized model, and the expected concentration change trend is output to provide data support for precise regulation. Thus, the concentration change trend under different conditions can be accurately predicted, and the regulation efficiency and stability of the system are improved.

[0100] In this optional embodiment, the final concentration trend prediction model is optimized using a model optimization algorithm, and the optimized concentration trend prediction model is used to predict the concentration change of the nutrient solution under different conditions, and the concentration change trend of the nutrient solution under different conditions is output, including:

[0101] Randomly initializing the structure parameters of the concentration trend prediction model as the initial solution of the current optimal concentration trend prediction model;

[0102] Iteratively testing the influence of each structure parameter combination on the prediction accuracy, and retaining the structure parameter combination of the concentration trend prediction model with the smallest error;

[0103] Judging whether the structure parameter combination of the current concentration trend prediction model meets the error constraint, if not, replacing the current concentration trend prediction model and continuing to test;

[0104] Adjusting the structure parameters of the concentration trend prediction model using a perturbation algorithm to jump out of the local optimum and continue to perform iterative optimization search;

[0105] Comparing the error performance of the concentration trend prediction model before optimization and the concentration trend prediction model after optimization, if the error value of the concentration trend prediction model after optimization is smaller than the error value of the concentration trend prediction model before optimization, updating the structure parameters of the concentration trend prediction model after optimization;

[0106] If the maximum number of iterations is reached, the optimized concentration trend prediction model is output to predict the concentration change of the nutrient solution under different conditions, and the concentration change trend of the nutrient solution under different conditions is output.

[0107] Specifically, first, the structure parameters of the concentration trend prediction model are randomly initialized as the initial solution of the current optimal model. Then, the influence of each combination of structure parameters on the prediction accuracy is iteratively tested, and the combination of structure parameters with the smallest error is retained. Next, it is judged whether the error of the current model meets the constraint condition. If not, the current model is replaced and the test is continued. The model structure parameters are adjusted using a perturbation algorithm to avoid falling into a local optimal solution, and the iterative optimization search is continued. During the optimization process, the error performance of the model before and after optimization is compared. If the error after optimization is smaller, the model parameters are updated. If the maximum number of iterations is reached, the optimized model is output, and the concentration change of the nutrient solution under different conditions is predicted using the model, and the corresponding concentration change trend is output. Thus, the concentration change trend can be accurately predicted under different environmental conditions, and the accuracy and stability of the system are improved.

[0108] Specifically, the model optimization algorithm is a greedy algorithm, which is an optimization strategy that always selects the current optimal solution at each iteration to quickly approach the overall optimal solution. In the present application, the greedy algorithm is used for optimization of the concentration trend prediction model: first, the model structure parameters are randomly initialized, and in each iteration, the parameter combination with the smallest prediction error is selected as the current optimal solution; if a new parameter combination is found that can further reduce the error, the model structure is immediately updated. Through continuous iteration and local perturbation search, the model can gradually approach the optimal state. Thus, the optimization efficiency of the model is improved, the search redundancy is reduced, and the prediction accuracy is significantly improved.

[0109] In this optional embodiment, the structure parameters of the concentration trend prediction model are adjusted using a perturbation algorithm to jump out of the local optimum and continue the iterative optimization search, including:

[0110] The structure parameters of the concentration trend model are initialized to generate a candidate structure set and calculate the prediction error as the fitness;

[0111] According to the fitness, the target structure combination is selected, simulated crossover is performed, and the current worst-performing structure combination is replaced;

[0112] The candidate structure combination set is subjected to mutation operation, and the local search ability is enhanced in combination with the simulated annealing mechanism;

[0113] Specifically, the formula for mutation operation on the candidate structure combination set is:

[0114] ;

[0115] In the formula, levy Levy represents the mutation value, controls the change range of the model structure configuration, and determines the formation of the new candidate structure combination set after perturbation operation; uis a random variable that obeys the normal distribution, representing the randomness of “prediction error” or “model bias”, and affecting the disturbance amplitude of the current candidate structure combination set; v is a random variable that obeys the normal distribution, indicating the randomness of the “model structure configuration” and affecting the jump degree of the variation of the current candidate structure combination set; β express levy The parameter of the distribution, usually set to 1.5, controls the jumpiness of the mutation operation. Smaller values ​​will lead to more local searches. Γ(·) represents the Gamma function, a mathematical function used to normalize the distribution process and represents the adjustment coefficient at different mutation stages.

[0116] Evaluate the fitness of the new structure combination. If it is better than the current optimal structure, replace it. Otherwise, retain the original structure and continue the iterative search.

[0117] The structural perturbation and fitness update process is executed cyclically, constantly jumping out of the local optimum and continuing to perform iterative optimization search.

[0118] Specifically, the structural parameters of the concentration trend prediction model are first initialized, multiple candidate structure sets are generated, and the prediction error of each candidate structure is calculated as the fitness value. The candidate structures are sorted according to the fitness value, the target structure combination is selected, and the structure combination with the worst performance is replaced by a simulated crossover operation. Next, the candidate structure set is mutated, and the simulated annealing mechanism is combined to enhance the local search capability, thereby avoiding falling into the local optimal solution. The fit of the new structure combination is evaluated. If it is better than the current optimal structure, it is replaced; otherwise, the original structure is retained and the iterative search continues. By continuously executing the structural perturbation and fitness update process, the local optimal solution is jumped out, and the iterative optimization search is continued to obtain the optimal concentration trend prediction model. Therefore, through perturbation and mutation operations, the model avoids the local optimal problem, improves the prediction accuracy and the global optimization capability of the model.

[0119] Specifically, the perturbation algorithm is a transit search algorithm, a heuristic global optimization method suitable for escaping local optima in complex search spaces. In the present invention, the algorithm generates a set of candidate structures by initializing structural parameters and calculates their prediction error as fitness. It is then sorted by fitness, simulated crossover and mutation operations are performed, and a simulated annealing mechanism is introduced to enhance local search capabilities. Through repeated perturbations and fitness updates, the search for more optimal structural combinations is continued, avoiding falling into local minima. This enhances the globality and adaptability of the model structure search and improves the accuracy and robustness of the concentration trend prediction model.

[0120] In this optional embodiment, performing a mutation operation on the candidate structure combination set and combining a simulated annealing mechanism to enhance the local search capability includes:

[0121] Randomly select a current structure combination and calculate the fitness of the current structure combination, generate a new candidate structure through mutation operation;

[0122] Calculate a new fitness value according to the mutated structure combination, judge whether it is better than the current solution, if the new structure solution is better than the current solution, replace the current solution, otherwise, determine whether to accept the worst solution through the simulated annealing mechanism;

[0123] Perform the temperature descending operation of simulated annealing, continuously optimize the local solution and enhance the local search capability.

[0124] Specifically, the application expands the population diversity by introducing cross mutation operation on the basis of the eclipse search algorithm, so as to prevent the result from falling into a local optimal solution; in addition, the idea of simulated annealing and variable neighborhood search algorithm is used to efficiently search the binary features, and meanwhile, the control parameters are set to balance the global search and local search capabilities, so as to improve the convergence precision.

[0125] In the optional embodiment, the feature analysis algorithm is used to analyze the concentration change trend of the nutrient solution under different conditions, identify the key factors affecting the stability of the concentration of the nutrient solution, and dynamically adjust the nutrient solution based on the key factors, including:

[0126] Obtain the concentration data of the nutrient solution under different environmental conditions and perform Boolean discretization to construct a concentration change trend analysis feature set;

[0127] Initialize a candidate set of factors affecting the concentration change trend, and set the initial weight of each factor;

[0128] Randomly select concentration data samples under different environmental conditions, and calculate the correlation coefficient of each influencing factor and concentration fluctuation;

[0129] Through the importance evaluation algorithm, the environmental parameters and operation parameters related to the concentration stability are screened out;

[0130] Establish a concentration change regression model under multiple environmental conditions to quantify the contribution of each influencing factor to the concentration fluctuation;

[0131] Verify the stability performance of each influencing factor in different environmental combinations, determine the core influencing factor set, and take the core influencing factor as the key factor affecting the stability of the concentration of the nutrient solution;

[0132] Based on the key factors, the spray frequency, nutrient solution supply rhythm and nutrient solution component proportion are dynamically adjusted to realize precise control and optimization of the concentration of the nutrient solution in the potato mist culture process.

[0133] Specifically, first, the concentration data of the nutrient solution under different environmental conditions is obtained, and Boolean discretization processing is performed to construct a concentration change trend analysis feature set. Then, the candidate set of influencing factors is initialized, the initial weight is set, the correlation between each candidate factor and the concentration fluctuation is calculated by randomly sampling, the environmental and operating parameters significantly related to the concentration stability are selected by using the importance evaluation algorithm, and a concentration change regression model is constructed to quantify the contribution of each factor to the concentration fluctuation. Further, the stability performance of each factor under multiple environmental combinations is evaluated to finally determine the core influencing factor set. Based on the identified key factors, the spray frequency, the nutrient solution supply rhythm and the ratio are dynamically adjusted to realize precise control of the concentration of the nutrient solution in the potato aeroponics system. Thus, the core driving factor identification and control strategy optimization of the concentration stability are realized.

[0134] Specifically, the feature analysis algorithm is a dendritic cell algorithm, which is a heuristic learning algorithm simulating the biological immune system and has strong pattern recognition and classification ability. In the present application, the dendritic cell algorithm is used to analyze the concentration change trend of the nutrient solution under different environmental conditions and identify the key factors affecting the concentration stability. Through the algorithm, the system can extract high-correlation factors from the concentration data and evaluate their influence on the concentration fluctuation. Finally, by screening the parameters related to the stability, the nutrient solution ratio and control strategy are dynamically adjusted. Thus, the analysis accuracy of the concentration change is improved, which helps to precisely adjust the nutrient solution ratio and optimize the growth environment.

[0135] In this optional embodiment, the environmental parameters and operating parameters related to the concentration stability are selected by the importance evaluation algorithm, including:

[0136] Randomly selecting several candidate parameter combinations in the environmental parameter space as the initial analysis set;

[0137] Calculating the influence weight of each parameter combination on the concentration stability of the nutrient solution as the importance evaluation index;

[0138] Selecting the parameter combination with the greatest influence on the concentration stability as the reference group;

[0139] Updating the weight coefficient and correlation degree score of each parameter. If a new parameter combination is found to have a better influence on the stability than the reference group, the reference group is updated, otherwise, it is eliminated and new candidate parameters are supplemented in the parameter space;

[0140] If the maximum number of iterations is reached, the environmental parameters and operating parameters related to the concentration stability are selected.

[0141] Specifically, first, a plurality of candidate parameter combinations are randomly selected in the environmental parameter space to form an initial analysis set, and the influence weight of each group on the nutrient solution concentration stability is calculated as an importance evaluation index. Then, the parameter combination with the greatest current influence is selected as the reference group, and the parameter weight and correlation degree are dynamically updated. If the influence of the new combination on the concentration stability is better than that of the reference group, the new combination replaces the reference group and the iteration continues; otherwise, the combination is eliminated, and new candidate parameters are added in the parameter space to maintain the exploration ability. Finally, when the maximum number of iterations is reached, the core environmental parameters and operating parameters related to the concentration stability are selected. Thus, the key influencing factors are efficiently identified, and quantifiable and traceable decision-making basis is provided for subsequent nutrient solution regulation.

[0142] Specifically, the importance evaluation algorithm is a black hole algorithm, which is a global optimization algorithm based on gravitational attraction mechanism, simulating the process of stars moving around black holes and being gradually absorbed. In the present application, the black hole algorithm randomly selects candidate parameter combinations in the environmental parameter space, calculates their influence weight on the concentration stability of the nutrient solution, and constantly updates the current optimal parameter set. If the new combination is better than the reference combination, the new combination replaces the reference combination; otherwise, the combination is eliminated, and new parameter candidates are added, and the iteration is performed until the maximum number of iterations is reached. Thus, the accuracy and reliability of the stability analysis are improved.

[0143] In this optional embodiment, calculating the influence weight of each parameter combination on the concentration stability of the nutrient solution as an importance evaluation index includes:

[0144] A mapping relationship model between the concentration change of the nutrient solution and the parameter combination is constructed, and the concentration fluctuation response characteristics are extracted;

[0145] The parameter combinations are disturbed and simulated, and the change trend of the concentration stability index under different parameter combinations is recorded;

[0146] According to the concentration fluctuation characteristics and the parameter combination disturbance results, the influence weight of each parameter combination is calculated as an importance evaluation index.

[0147] According to another embodiment of the present application, as Figure 2 shown, a nutrient solution concentration regulation system for potato fog culture is also provided, which comprises:

[0148] A data management module 1 is used to obtain the root temperature distribution image of the potato and the concentration data of each component of the fog culture nutrient solution, and extract the root feature data and the concentration feature data;

[0149] A trend prediction module 2 is used to establish a concentration trend prediction model based on the root feature data and the concentration feature data, and predict the concentration change trend of the nutrient solution under different conditions using the concentration trend prediction model;

[0150] A trend analysis and regulation module 3 is configured to analyze the concentration variation trend of the nutrient solution under different conditions by using a feature analysis algorithm, identify key factors affecting the stability of the concentration of the nutrient solution, and dynamically adjust the nutrient solution based on the key factors.

[0151] In order to facilitate the understanding of the above technical solutions of the present application, the following will be described in detail for the nutrient solution concentration regulation of the present application in the actual process of potato mist culture.

[0152] I. Root temperature and nutrient solution concentration data acquisition and feature extraction:

[0153] 1) Obtain the root temperature distribution image: use infrared thermal imaging technology to collect the root zone temperature distribution image of potatoes in real time. Through image processing technology, the temperature feature data of the root system is extracted. Assuming that in a certain collection, the temperature range of the root zone is 18°C to 22°C, it can be divided by region and the root system distribution in different temperature ranges is counted.

[0154] 2) Obtain the concentration data of the nutrient solution: use conductivity meter and pH meter and other equipment to measure the concentration of each component in the nutrient solution in real time. Select to measure nitrogen ( ), potassium ( ) and pH value, assuming that the concentration data is as follows:

[0155] Nitrogen concentration ( ): 150 ppm.

[0156] Potassium concentration ( ): 200 ppm.

[0157] pH value: 5.8.

[0158] 3) Extract root feature data and concentration feature data:

[0159] Process the obtained root temperature distribution image to extract the temperature fluctuation range and root distribution density and other features:

[0160] Root temperature fluctuation amplitude: ±2°C.

[0161] Root distribution density: the root distribution density in the root zone is 0.8 root / cm².

[0162] Perform fluctuation analysis on the concentration data to calculate the following concentration features:

[0163] Nitrogen concentration fluctuation amplitude: ±10 ppm.

[0164] Potassium concentration fluctuation amplitude: ±15 ppm.

[0165] pH value stability: the fluctuation range is 5.6-6.0.

[0166] 2. Construction and prediction of concentration trend prediction model based on deep belief network (DBN):

[0167] 1) Data Preparation: Based on the collected root temperature distribution and nutrient solution concentration fluctuation characteristics, a dataset was created and fed into the Deep Belief Network (DBN) model. Assuming a training set and a validation set, the dataset size was 1,000 sets of samples, each containing root temperature, concentration fluctuation amplitude, and pH value.

[0168] 2) Model Building: We used a Deep Belief Network (DBN) for training, setting the network to 3 layers with 100 neurons per layer. Initially, we set the learning rate to 0.01 and the number of iterations to 500. We trained the network and obtained the optimal model parameters using backpropagation, greedy algorithms, and gradient descent optimization.

[0169] 3) Assume that under certain prediction conditions, the root temperature is 20°C, the nitrogen concentration is 160 ppm, the potassium concentration is 220 ppm, and the pH is 5.9. The DBN model is used to predict the output of the prediction results as follows:

[0170] Predicted nitrogen concentration: 155 ppm (fluctuation range is ±10 ppm).

[0171] Predicted potassium concentration: 210 ppm (fluctuation range is ±12 ppm).

[0172] Predicted pH value: 5.85 (fluctuation range: ±0.1).

[0173] 3. Concentration Trend Analysis, Key Factor Identification, and Dynamic Allocation Mechanism:

[0174] 1) Feature analysis and influencing factor identification: Use feature analysis algorithms (such as information gain, decision tree algorithm, dendritic cell algorithm, etc.) to analyze the prediction results and identify the most critical influencing factors. Suppose that during the analysis, the following influencing factors are obtained:

[0175] The pH value has a greater impact on concentration stability, with a weight of 0.45.

[0176] The influence weight of root temperature fluctuation on concentration change is 0.35.

[0177] The influence weight of the fluctuation of potassium concentration in the nutrient solution on the concentration change is 0.2.

[0178] It can be concluded that pH value is the most critical factor affecting concentration stability.

[0179] 2) Dynamic allocation strategy: Based on the prediction results and key factor identification results, the system automatically adjusts the spray frequency and nutrient solution ratio. When the root temperature rises, increase the Concentration, reduction Concentration, and the pH is adjusted to the target interval (5.8~6.0) by buffer, as shown in Table 1:

[0180] Table 1, dynamic deployment strategy

[0181]

[0182] Therefore, at a temperature of 22°C, the potassium concentration is increased to 230 ppm to promote root absorption; the nitrogen concentration is adjusted to 150 ppm to avoid excess; and the pH value is maintained in the appropriate range of 5.8-5.9.

[0183] In summary, by means of the above technical solutions of the present application, the present application realizes self-adaptive adjustment of structural parameters by constructing a concentration trend prediction model and combining a model optimization algorithm, thereby improving the accuracy and stability of concentration change trend prediction, effectively avoiding being trapped in local optimum by using a disturbance mechanism and a simulated annealing strategy, dynamically predicting and responding to nutrient solution concentration under various environmental conditions, and thereby providing systematic support for precise control of nutrient solution concentration, which is suitable for concentration management requirements in the complex environment of potato hydroponics. The present application identifies key factors affecting the stability of nutrient solution concentration by a feature analysis algorithm, realizes precise modeling and quantitative interpretation of the concentration change trend by combining data analysis and regression modeling under multiple environmental conditions, dynamically adjusts the spraying frequency, supply rhythm and nutrient ratio, which is helpful to realize precise control of nutrient solution concentration, thereby improving the systematization and responsiveness of the concentration control process, and providing stable and reliable nutrient supply strategy support for potato hydroponics.

[0184] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for controlling the concentration of a nutrient solution for potato mist culture, characterized by, The method comprises the following steps: Obtain the root temperature distribution image of potatoes and the concentration data of each component of the nutrient solution, and extract the root feature data and the concentration feature data; Based on the root feature data and the concentration feature data, a concentration trend prediction model is established, and the concentration trend prediction model is used to predict the concentration trend of the nutrient solution under different conditions; Using a feature analysis algorithm, the concentration trend of the nutrient solution under different conditions is analyzed, and the key factors affecting the stability of the concentration of the nutrient solution are identified; and based on the key factors, the nutrient solution is dynamically adjusted; The method comprises the following steps: Denoising, filtering and smoothing the repeated data, missing values and outliers of the obtained root temperature distribution image and the concentration data of each component of the nutrient solution to obtain complete root temperature distribution image and concentration data of each component of the nutrient solution; Initialize the data model, extract the high-frequency feature index in the temperature distribution image and the concentration data, and sort them; Construct an encoding mapping dictionary to perform differential compression and feature value encoding on the extracted high-frequency features; Based on the high-frequency feature sequence, a feature data tree is constructed to record the combined path information of the temperature distribution image and the concentration data; From the bottom of the feature data tree, generate a combined path, extract a key feature mode, and update the feature set; Perform statistical analysis on the feature set to extract the root feature data and the concentration feature data.

2. The method for controlling the concentration of a nutrient solution for potato mist culture according to claim 1, wherein The method comprises the following steps: Divide the root feature data and the concentration feature data, and construct a training set and a validation set; Initialize the key parameters of the concentration trend prediction model, and randomly generate an initial solution in the parameter search space as the initial structure configuration of the concentration trend model; Input the training set into the initial concentration trend prediction model to calculate the error as the initial fitness value of the concentration trend prediction model, and evaluate the performance of the initial concentration trend prediction model using the validation set. Through multiple iterations, the parameters of the initial concentration trend prediction model are optimized to obtain the final concentration trend prediction model; Use a model optimization algorithm to optimize the final concentration trend prediction model, and use the optimized concentration trend prediction model to predict the concentration change of the nutrient solution under different conditions, and output the concentration trend of the nutrient solution under different conditions.

3. The method for controlling the concentration of a nutrient solution for aeroponics of potatoes according to claim 2, characterized in that, The method comprises the following steps: Randomly initialize the structure parameters of the concentration trend prediction model as the initial solution of the current optimal concentration trend prediction model; Iteratively test the influence of each structure parameter combination on the prediction accuracy, and retain the structure parameter combination of the concentration trend prediction model with the smallest error; Determine whether the structure parameter combination of the current concentration trend prediction model meets the error constraint, if not, replace the current concentration trend prediction model and continue to test; Adjusting the structural parameters of the concentration trend prediction model using a perturbation algorithm, jumping out of the local optimum and continuing to perform iterative optimization search; Comparing the error performance of the concentration trend prediction model before optimization and the concentration trend prediction model after optimization, if the error value of the concentration trend prediction model after optimization is less than the error value of the concentration trend prediction model before optimization, then updating the structural parameters of the concentration trend prediction model after optimization; If the maximum number of iterations is reached, output the optimized concentration trend prediction model to predict the concentration change of the nutrient solution under different conditions, and output the concentration change trend of the nutrient solution under different conditions.

4. The method of claim 3, wherein the concentration of the nutrient solution is controlled by the amount of the nutrient solution supplied to the potato plant. The adjusting the structural parameters of the concentration trend prediction model using a perturbation algorithm, jumping out of the local optimum and continuing to perform iterative optimization search includes: Initializing the structural parameters of the concentration trend model, generating a candidate structure set and calculating its prediction error as fitness; Selecting the target structure combination according to the fitness ranking, performing simulated crossover and replacing the current worst structure combination; Performing mutation operation on the candidate structure combination set, and combining with the simulated annealing mechanism to enhance the local search ability; Evaluating the fitting degree of the new structure combination, if better than the current optimal structure, then replace, otherwise keep the original structure and continue iterative search; Looping the structure disturbance and fitness updating process, constantly jumping out of the local optimum and continuing to perform iterative optimization search.

5. The method of claim 4, wherein the concentration of the nutrient solution is controlled by the amount of the nutrient solution supplied to the potato plant. The mutation operation on the candidate structure combination set, and combining with the simulated annealing mechanism to enhance the local search ability includes: Randomly selecting the current structure combination and calculating the fitness of the current structure combination, generating a new candidate structure through mutation operation; According to the structure combination after mutation, calculate the new fitness value, judge whether it is better than the current solution, if the new structure solution is better than the current solution, replace the current solution; Otherwise, decide whether to accept the worst solution through the simulated annealing mechanism; Performing temperature descending operation of simulated annealing, continuously optimizing local solution and enhancing local search ability.

6. The method for controlling the concentration of a nutrient solution for aeroponics of potatoes according to claim 1, characterized in that, The using feature analysis algorithm to analyze the concentration change trend of the nutrient solution under different conditions, identifying the key factors affecting the stability of the nutrient solution concentration; And based on the key factors, dynamically adjusting the nutrient solution includes: Obtaining the concentration data of the nutrient solution under different environmental conditions and performing Boolean discretization, constructing the concentration change trend analysis feature set; Initializing the candidate set of factors affecting the concentration change trend, and setting the initial weight of each factor; Randomly selecting concentration data samples under different environmental conditions, calculating the correlation coefficient of each influencing factor and concentration fluctuation; Through the importance evaluation algorithm, screening out the environmental parameters and operation parameters related to the concentration stability; Establishing a concentration change regression model under multiple environmental conditions to quantify the contribution of each influencing factor to the concentration fluctuation; Verify the stability performance of each influencing factor in different environmental combinations, determine the core influencing factor set, and take the core influencing factor as the key factor affecting the stability of the nutrient solution concentration; Based on the key factors, dynamically adjusting the spray frequency, nutrient solution supply rhythm and nutrient solution component proportion, realizing the precise regulation and optimization of the nutrient solution concentration in the potato mist culture process.

7. The method of claim 6, wherein the concentration of the nutrient solution is controlled by the amount of the nutrient solution supplied to the potato plant. The through the importance evaluation algorithm, screening out the environmental parameters and operation parameters related to the concentration stability includes: Randomly select several candidate parameter combinations in the environmental parameter space as the initial analysis set; Calculate the influence weight of each parameter combination on the concentration stability of the nutrient solution as an importance evaluation index; Select the parameter combination that has the greatest impact on the concentration stability as the reference group; Update the weight coefficient and correlation degree score of each parameter. If a new parameter combination is found to have a better impact on stability than the reference group, update the reference group. Otherwise, eliminate it and supplement new candidate parameters in the parameter space; If the maximum number of iterations is reached, the environmental parameters and operating parameters related to the concentration stability are selected.

8. The method of claim 7, wherein the concentration of the nutrient solution is controlled by the amount of the nutrient solution supplied to the potato plant. The calculation of the influence weight of each parameter combination on the concentration stability of the nutrient solution as an importance evaluation index includes: Construct a mapping relationship model between the concentration change of the nutrient solution and the parameter combination, and extract the concentration fluctuation response characteristics; Perform perturbation simulation on each parameter combination and record the change trend of the concentration stability index under different parameter combinations; According to the concentration fluctuation characteristics and the parameter combination perturbation results, calculate the influence weight of each parameter combination as an importance evaluation index.

9. A nutrient solution concentration regulating system for potato mist culture for implementing the nutrient solution concentration regulating method for potato mist culture according to any one of claims 1 to 8, characterized by, The system comprises: A data management module for obtaining the root temperature distribution image of potatoes and the concentration data of each component of the nutrient solution, and extracting root feature data and concentration feature data; A trend prediction module for establishing a concentration trend prediction model based on the root feature data and the concentration feature data, and predicting the concentration change trend of the nutrient solution under different conditions using the concentration trend prediction model; A trend analysis and control module for analyzing the concentration change trend of the nutrient solution under different conditions using a feature analysis algorithm, identifying the key factors affecting the concentration stability of the nutrient solution, and dynamically adjusting the nutrient solution based on the key factors.

Images