Medicinal material growth illumination optimization method and system based on red light regulation and control

Abstract

The invention relates to the technical field of medicinal material growth illumination optimization methods, in particular to a medicinal material growth illumination optimization method and system based on red light regulation, and the method comprises the following steps: obtaining environmental parameters, and processing the environmental parameters to generate environmental feature data; setting the illumination duration of medicinal material growth, and setting the irradiation intensity of a red light lamp within the illumination duration; within the illumination duration, environmental parameters of medicinal material growth are detected in real time, and the environmental parameters detected in real time are processed to obtain environmental characteristic data; inputting the environment characteristic data into a neural network model of illumination duration, then outputting predicted illumination duration, judging whether the output predicted illumination duration is consistent with the illumination duration of medicinal material growth or not, and if yes, keeping the irradiation intensity of the red light lamp unchanged; if not, updating the irradiation intensity of the red light lamp according to the predicted illumination duration; and according to the updated irradiation intensity of the red light lamp, adjusting the red light lamp in the medicinal material growth environment, and performing real-time optimization control on the medicinal material growth environment.

Description

Technical Field

[0001] The present invention relates to the technical field of lighting optimization methods for medicinal material growth, and in particular to a lighting optimization method for medicinal material growth based on red light regulation and a system thereof. Background Art

[0002] In recent years, with the rapid development of modern agricultural technology, the field of medicinal material cultivation has also ushered in new opportunities and challenges. Traditional medicinal material cultivation methods mainly rely on manual experience and fixed lighting schemes. Although this method is simple and direct, it is difficult to adapt to complex and changing environmental conditions, and it is impossible to carry out refined management based on the characteristics of different medicinal materials. As people's requirements for the quality of medicinal materials continue to increase, the limitations of traditional methods are becoming increasingly prominent.

[0003] In this context, some researchers have begun to try to apply modern lighting technology to medicinal material cultivation. Among them, red light regulation technology has attracted widespread attention due to its significant role in promoting plant photosynthesis. However, existing red light regulation methods still have many shortcomings. First, most systems use fixed lighting schemes and cannot be adjusted in real time according to environmental changes and medicinal material growth conditions. Secondly, the setting of light intensity and duration is often based on rough empirical values ​​and lacks scientific quantitative basis. Furthermore, existing systems generally lack in-depth mining and utilization of historical data, and cannot learn and optimize from long-term planting experience.

[0004] In addition, most of the existing lighting control systems operate in isolation, lacking coordination with other environmental factors (such as temperature and humidity). This single-factor optimization method is difficult to comprehensively improve the growth quality of medicinal materials. At the same time, the scalability and adaptability of the existing system are poor, making it difficult to meet the diverse needs of different types of medicinal materials.

[0005] In the face of these technical problems, the present invention proposes a method and system for optimizing the lighting conditions for medicinal material growth based on red light regulation. The method aims to achieve precise control and dynamic optimization of the lighting conditions for medicinal material growth through intelligent algorithms and real-time environmental monitoring. The core of the present invention is to combine advanced neural network technology with traditional planting experience to build an adaptive lighting regulation system. Summary of the invention

[0006] In the face of these technical problems, the present invention proposes a method and system for optimizing the lighting conditions for medicinal material growth based on red light regulation. The method aims to achieve precise control and dynamic optimization of the lighting conditions for medicinal material growth through intelligent algorithms and real-time environmental monitoring. The core of the present invention is to combine advanced neural network technology with traditional planting experience to build an adaptive lighting regulation system.

[0007] The present invention proposes a method for optimizing illumination of medicinal material growth based on red light regulation, comprising the following steps:

[0008] Acquiring environmental parameters, and processing the environmental parameters to generate environmental characteristic data;

[0009] Set the illumination duration for the growth of medicinal materials, and at the same time set the illumination intensity of the red light fixture within the illumination duration;

[0010] During the illumination period, real-time detection of environmental parameters for the growth of medicinal materials is performed, and the real-time detected environmental parameters are processed to obtain environmental characteristic data;

[0011] After inputting the environmental characteristic data into the neural network model of the illumination duration, the predicted illumination duration is output, and it is determined whether the output predicted illumination duration is consistent with the illumination duration for the growth of the medicinal material. If so, the illumination intensity of the red light fixture is kept unchanged; if not, the illumination intensity of the red light fixture is updated according to the predicted illumination duration;

[0012] According to the updated red light intensity, adjust the red light in the medicinal material growth environment to achieve real-time optimization control of the medicinal material growth environment;

[0013] Among them, the training method of the neural network model of the lighting duration includes: initializing the parameters of the neural network; inputting the empirical data into the neural network model; outputting the predicted lighting duration and the illumination intensity of the red light fixture within the corresponding predicted lighting duration; calculating the empirical error; judging whether the empirical error is greater than a first preset threshold, and if so, updating the parameters of the neural network by back-propagating the empirical error through the gradient descent method; if not, returning to the step of inputting the empirical data into the neural network model.

[0014] Preferably, before the step of inputting the empirical data into the neural network model, the method further includes:

[0015] Reading empirical data, wherein the empirical data is composed of a number of environmental characteristic data and corresponding neural network prediction data of illumination duration;

[0016] The empirical data is obtained through historical planting records, expert knowledge base and experimental data, and is processed through data cleaning and standardization;

[0017] After the step of calculating the empirical error, the method further includes: determining whether the empirical error is greater than a second preset threshold value, and if so, determining a neural network model and parameters of the illumination duration, and outputting the neural network model of the illumination duration; if not, returning to the step of inputting the empirical data into the neural network model;

[0018] Among them, the first preset threshold is smaller than the second preset threshold to achieve fine control and rapid convergence of model training.

[0019] Preferably, the environmental parameters include temperature, humidity and light intensity detected by a photosensor;

[0020] The steps of processing the real-time detected environmental parameters to obtain environmental characteristic data are specifically: normalizing the environmental parameters to obtain environmental characteristic data;

[0021] Among them, the normalization process uses the minimum-maximum scaling method to ensure that all feature data are within the interval [0,1];

[0022] The illumination intensity and illumination duration of the red light fixture meet the following conditions: within the illumination duration, the light is irradiated at an initialized illumination intensity, and after illumination, the initialized illumination intensity is left to stand for a set time. Within the set time, the sum of the temperature and humidity detected in real time is a constant A, the product of the temperature and humidity is a constant B, and the light intensity detected by the photosensor within the set time after illumination is a constant C.

[0023] Among them, the value ranges of constants A, B, and C are dynamically adjusted according to different types of medicinal materials and growth stages.

[0024] Preferably, the loss function of the neural network model training of the illumination duration is:

[0025]

[0026] Among them, L is the loss function value, α is the first hyperparameter, and t i is the illumination duration in the empirical data, is the illumination duration predicted by the neural network model of illumination duration, β is the second hyperparameter, and p i is the light intensity detected by the photosensor in the empirical data, is the light intensity detected by the photosensor predicted by the neural network model of the illumination duration, and T is the total number of empirical data; the values ​​of α and β are determined by the cross-validation method to balance the accuracy of the illumination duration and light intensity prediction;

[0027] The loss function also contains a regularization term to prevent the model from overfitting.

[0028] Preferably, the neural network model of the illumination duration includes:

[0029] The input layer, hidden layer, fully connected layer, and output layer are connected in sequence;

[0030] The input layer is used to sequentially receive the temperature, humidity and light intensity detected by the photosensor;

[0031] The hidden layer receives the temperature, humidity and light intensity detected by the photosensor, and outputs a lighting duration and a corresponding illumination intensity;

[0032] The hidden layer includes a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and a fifth hidden layer connected in sequence;

[0033] The illumination duration output by the first hidden layer has a value range of 0-5 hours, and the illumination intensity output by the second hidden layer has a value range of 0-200;

[0034] The illumination duration output by the first hidden layer is in the range of 6-8 hours, and the illumination intensity output by the third hidden layer is in the range of 210-240;

[0035] The illumination duration output by the fourth hidden layer has a value range of 9-12 hours, and the illumination intensity output by the fifth hidden layer has a value range of 241-300;

[0036] The fully connected layer receives the illumination duration and illumination intensity, and outputs them;

[0037] The output layer receives the illumination duration and illumination intensity, and outputs them;

[0038] Among them, each hidden layer uses a different activation function to enhance the nonlinear fitting ability of the model.

[0039] Preferably, the method further comprises the following steps:

[0040] Regularly retrain the neural network model of light duration to adapt to seasonal changes in the medicinal material growing environment;

[0041] Through cloud servers, we conduct in-depth mining of long-term accumulated planting data to optimize the structure and parameters of the neural network model;

[0042] Provides a human-computer interaction interface that allows growers to manually adjust system parameters and view historical data.

[0043] Preferably, the method further comprises the following steps:

[0044] Outputting the predicted illumination duration and the corresponding predicted illumination intensity to the light sensor component;

[0045] The light sensor assembly receives the predicted illumination duration and the corresponding predicted illumination intensity, and determines the illumination intensity according to the predicted illumination intensity and determines the illumination duration according to the predicted illumination duration;

[0046] The medicinal material growth lighting process regulated by red light is executed according to the illumination intensity and illumination duration.

[0047] A medicinal material growth illumination optimization system based on red light regulation for executing the method described in the claim, the system comprising:

[0048] A sensor component for acquiring environmental parameters;

[0049] Red light fixtures for providing adjustable lighting;

[0050] A neural network model of light duration, installed in the planting space, for receiving environmental parameters detected by the sensor assembly and generating a neural network model of light duration;

[0051] Communication module, used for data transmission;

[0052] The cloud server is used to receive the predicted illumination duration and the corresponding predicted illumination intensity, and send the illumination intensity and illumination duration to the red light fixture through the communication module.

[0053] Preferably, the system further comprises a data storage and analysis module for: storing long-term accumulated planting data; performing in-depth mining on the planting data; and optimizing the structure and parameters of the neural network model according to the analysis results.

[0054] Preferably, the system further comprises a model retraining module and a human-computer interaction interface, wherein:

[0055] The model retraining module is used to regularly retrain the light duration neural network model to adapt the model to seasonal changes in the medicinal material growth environment and update the model parameters to improve the prediction accuracy;

[0056] The human-computer interaction interface is used to allow growers to manually adjust system parameters, display historical data and forecast results, and provide real-time feedback on the system operating status.

[0057] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0058] First, the present invention realizes dynamic optimization of lighting conditions through real-time environmental monitoring and neural network model prediction. This method can not only cope with instantaneous changes in the environment, but also automatically adjust the lighting strategy according to the growth stage of the medicinal materials, thereby significantly improving the accuracy and flexibility of lighting management.

[0059] Secondly, the neural network model of the present invention adopts a multi-layer structure design, which can simultaneously optimize the two key parameters of light duration and intensity. This multi-objective optimization method effectively solves the limitations of traditional single parameter regulation and can more comprehensively meet the light requirements of different medicinal materials at different growth stages.

[0060] Furthermore, the present invention introduces cloud data analysis and model retraining mechanisms, which enable the system to continuously learn and optimize from historical data. This continuous learning capability allows the system to adapt to seasonal changes and the needs of different types of medicinal materials, greatly improving the versatility and long-term effectiveness of the system.

[0061] In addition, the human-machine interaction interface of the present invention provides the grower with the flexibility of manual adjustment, realizing the organic combination of artificial intelligence and human experience. This design not only improves the reliability of the system, but also provides valuable feedback information for further optimizing the model.

[0062] Finally, the system architecture of the present invention has good scalability. By adding new sensors or adjusting the neural network structure, the system can easily adapt to new environmental factors or types of medicinal materials, laying a foundation for future technology upgrades and application expansion.

[0063] In general, the invention solves the problems of low efficiency and insufficient precision in traditional methods through the innovation of algorithm mechanism, and realizes the synergistic effect of environmental monitoring, data analysis, model prediction and light control. This all-round optimization can not only significantly improve the yield and quality of medicinal materials, but also reduce energy consumption, providing a new technical path for modern and intelligent medicinal material planting. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a logic block diagram of the entire system of the present invention.

[0065] Figure 2 It is a logic block diagram of the overall method of the present invention.

[0066] Figure 3 This is a logic block diagram for training the neural network model of the present invention. DETAILED DESCRIPTION

[0067] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects thereof are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0068] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0069] See also Figure 1-3 The present invention relates to a method and system for optimizing illumination of medicinal material growth based on red light regulation, which is mainly used in the field of smart agriculture, especially in the medicinal material planting industry. First, let the present invention elaborate on the core content of the present invention.

[0070] The method for optimizing illumination of medicinal material growth based on red light regulation of the present invention comprises the following steps:

[0071] First, environmental parameters are acquired, and the environmental parameters are processed to generate environmental characteristic data;

[0072] Secondly, set the illumination duration for the growth of medicinal materials, and at the same time set the illumination intensity of the red light fixture within the illumination duration;

[0073] Then, during the illumination period, the environmental parameters for the growth of the medicinal materials are detected in real time, and the environmental parameters detected in real time are processed to obtain environmental characteristic data;

[0074] Next, the environmental characteristic data is input into the neural network model of the illumination duration, and the predicted illumination duration is outputted. It is determined whether the output predicted illumination duration is consistent with the illumination duration for the growth of the medicinal material. If so, the illumination intensity of the red light fixture is kept unchanged; if not, the illumination intensity of the red light fixture is updated according to the predicted illumination duration;

[0075] Finally, according to the updated red light lamp irradiation intensity, the red light lamp in the medicinal material growth environment is adjusted to achieve real-time optimization control of the medicinal material growth environment.

[0076] In one embodiment of the present invention, the acquisition of environmental parameters is completed by the sensor assembly 1. The sensor assembly 1 includes a temperature sensor, a humidity sensor, and a light intensity sensor. These sensors monitor the growth environment of medicinal materials in real time and provide accurate environmental data for the system. Preferably, the accuracy of the temperature sensor is ±0.1°C, the accuracy of the humidity sensor is ±2%RH, and the accuracy of the light intensity sensor is ±3%. Such high-precision sensors can ensure that the system obtains accurate environmental data, thereby providing a reliable basis for subsequent optimization control.

[0077] Next, the system will set the initial illumination duration and the illumination intensity of the red light fixture 2 according to the growth characteristics of different medicinal materials. For example, for shade-loving plants such as ginseng, the initial illumination duration is set to 8 hours and the illumination intensity is set to 50%. For light-loving plants such as angelica, the initial illumination duration is set to 12 hours and the illumination intensity is set to 80%. This personalized initial setting can provide a more suitable growth environment for different types of medicinal materials.

[0078] During the set lighting duration, the system will continuously monitor changes in environmental parameters. This real-time monitoring method enables the system to quickly respond to environmental changes and adjust the lighting strategy in a timely manner. For example, if the temperature is detected to rise suddenly above 30°C, the system will reduce the red light intensity to 40% accordingly to prevent the plants from suffering from heat stress. This dynamic adjustment method can provide the most suitable growth environment for plants, thereby improving the yield and quality of medicinal materials.

[0079] The training method of the neural network model is as follows:

[0080] First, initialize the parameters of the neural network;

[0081] Secondly, the empirical data is input into the neural network model;

[0082] Then, the predicted illumination duration and the illumination intensity of the red light fixture within the corresponding predicted illumination duration are output;

[0083] Next, calculate the empirical error;

[0084] Finally, determine whether the empirical error is greater than the first preset threshold. If so, back-propagate the empirical error through the gradient descent method to update the parameters of the neural network; if not, return to the step of inputting the empirical data into the neural network model.

[0085] In another embodiment of the present invention, the training of the neural network model is an iterative process. First, the present invention randomly initializes the weights and biases of the network. For example, the present invention uses the Xavier initialization method, which can make the output variance of each layer as equal as possible, which helps to solve the gradient vanishing problem of deep networks.

[0086] Then, the present invention inputs a large amount of historical planting data collected into the model. These data include records of optimal lighting strategies under different environmental conditions. For example, the present invention has such a record: under the conditions of temperature 25°C, humidity 60%, and initial light intensity 70%, the optimal lighting duration is 10 hours.

[0087] Next, the model will output its prediction of the duration and intensity of illumination. The present invention compares this prediction result with the actual optimal value and calculates the error. If the error is greater than a preset threshold (for example, the present invention sets the threshold to 0.1, i.e., allows 10% error), then the present invention uses the back propagation algorithm to update the parameters of the model.

[0088] Preferably, the present invention uses Adam optimizer for parameter update. Adam optimizer combines the advantages of momentum method and adaptive learning rate method, and can converge to the global optimal solution faster. The present invention sets the initial learning rate to 0.001, β1 to 0.9, and β2 to 0.999. The selection of these parameters is based on the empirical values ​​obtained from a large number of experiments, which can achieve good results in most cases.

[0089] In a specific embodiment of the present invention, the source and processing method of the empirical data are as follows:

[0090] Before the step of inputting the empirical data into the neural network model, it also includes: reading the empirical data, wherein the empirical data is composed of a number of environmental characteristic data and the corresponding neural network prediction data of the lighting duration; wherein the empirical data is obtained through a comprehensive combination of historical planting records, expert knowledge base and experimental data, and undergoes data cleaning and standardization processing.

[0091] In practical applications, the present invention will collect a large amount of raw data. For example, a ginseng planting base has environmental data and corresponding growth records for the past five years. These data include daily temperature, humidity, light intensity, and corresponding ginseng growth conditions (such as plant height, leaf area, root weight, etc.).

[0092] However, these raw data have problems such as noise and missing values. Therefore, the present invention needs to perform data cleaning. For example, the present invention uses the moving average method to smooth the temperature data and the interpolation method to fill in the missing humidity data. For obviously abnormal data (such as the temperature suddenly reaching 50°C), the present invention will directly delete it or replace it with a random value within the normal range.

[0093] In addition, the present invention also requires data standardization. For example, the present invention uses the Z-score standardization method to convert all feature data into a distribution with a mean of 0 and a standard deviation of 1. This processing method can eliminate the dimensional differences between different features, allowing the neural network to better learn patterns in the data.

[0094] After the step of calculating the empirical error, the method further includes: determining whether the empirical error is greater than a second preset threshold value, and if so, determining a neural network model and parameters of the illumination duration, and outputting the neural network model of the illumination duration; if not, returning to the step of inputting the empirical data into the neural network model;

[0095] Among them, the first preset threshold is smaller than the second preset threshold to achieve fine control and rapid convergence of model training.

[0096] In a specific embodiment of the present invention, the specific content and processing method of the environmental parameters are as follows:

[0097] The environmental parameters include temperature, humidity and light intensity detected by a photosensor; the step of processing the environmental parameters detected in real time to obtain environmental characteristic data is specifically as follows: normalizing the environmental parameters to obtain environmental characteristic data; wherein the normalization process uses a minimum-maximum scaling method to ensure that all characteristic data are within the interval [0,1].

[0098] In a specific embodiment of the present invention, the present invention uses the minimum-maximum scaling method for normalization processing. The advantage of this method is that it can retain the distribution characteristics of the original data while mapping all data to the [0,1] interval.

[0099] For example, assuming that the temperature data collected by the present invention ranges from 10°C to 35°C, the present invention can use the following formula for normalization:

[0100]

[0101] In this way, 10℃ will be mapped to 0, 35℃ will be mapped to 1, and other temperature values ​​will be mapped between 0 and 1. The same method is also applicable to humidity and light intensity data. This processing method allows parameters of different scales to be compared and calculated in the same model, improving the training efficiency and prediction accuracy of the model.

[0102] The illumination intensity and illumination duration of the red light fixture meet the following conditions: within the illumination duration, the light is irradiated at an initialized illumination intensity, and after illumination, the initialized illumination intensity is left to stand for a set time. Within the set time, the sum of the temperature and humidity detected in real time is a constant A, the product of the temperature and humidity is a constant B, and the light intensity detected by the photosensor within the set time after illumination is a constant C.

[0103] Among them, the value ranges of constants A, B, and C are dynamically adjusted according to different types of medicinal materials and growth stages.

[0104] In a specific embodiment of the present invention, the loss function of the neural network model is as follows:

[0105]

[0106] Among them, L is the loss function value, α is the first hyperparameter, and t i is the illumination duration in the empirical data, is the illumination duration predicted by the neural network model of illumination duration, β is the second hyperparameter, and p i is the light intensity detected by the photosensor in the empirical data, is the light intensity detected by the photosensor predicted by the neural network model of illumination duration, and T is the total number of empirical data; the values ​​of α and β are determined by the cross-validation method to balance the accuracy of illumination duration and light intensity prediction.

[0107] This loss function contains two parts, corresponding to the prediction errors of light duration and light intensity. α and β are two hyperparameters used to balance the importance of these two objectives. In practical applications, the present invention will find that light duration has a greater impact on the growth of medicinal materials, so α>β will be set. For example, the present invention sets α=0.7 and β=0.3. In this way, the model will pay more attention to improving the prediction accuracy of light duration during training.

[0108] Preferably, the present invention also adds an L2 regularization term in the loss function to prevent the model from overfitting. The modified loss function is as follows:

[0109]

[0110] Where λ is the regularization coefficient, ∑w 2 is the sum of the squares of all the weights of the model. The value of λ is usually very small, such as 0.0001, which can prevent overfitting without excessively affecting the model's fitting ability.

[0111] In a specific embodiment of the present invention, the specific structure of the neural network model is as follows:

[0112] The neural network model of illumination duration includes: an input layer, a hidden layer, a fully connected layer and an output layer connected in sequence; the input layer is used to sequentially receive the temperature, humidity and light intensity detected by the photosensor; the hidden layer receives the temperature, humidity and light intensity detected by the photosensor, and outputs an illumination duration and a corresponding illumination intensity; the hidden layer includes a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and a fifth hidden layer connected in sequence; the illumination duration output by the first hidden layer has a value range of 0-5 hours, and the illumination intensity output by the second hidden layer has a value range of 0-200; the illumination duration output by the first hidden layer has a value range of 6-8 hours, and the illumination intensity output by the third hidden layer has a value range of 210-240; the illumination duration output by the fourth hidden layer has a value range of 9-12 hours, and the illumination intensity output by the fifth hidden layer has a value range of 241-300; the fully connected layer receives the illumination duration and illumination intensity, and outputs them; the output layer receives the illumination duration and illumination intensity, and outputs them.

[0113] In a preferred embodiment of the present invention, the input layer receives normalized temperature, humidity and light intensity data. The hidden layer is divided into five sublayers, each of which is responsible for predicting light duration and light intensity in a different range. The advantage of this hierarchical design is that it can more accurately capture the characteristics of medicinal material growth under different light conditions.

[0114] For example, for shade-loving ginseng, the first and second hidden layers play a greater role because they are responsible for handling shorter light durations and lower illumination intensities, while for light-loving angelica, the fourth and fifth hidden layers are more important.

[0115] Each hidden layer uses a different activation function to enhance the nonlinear fitting ability of the model. For example, the present invention uses the ReLU function in the first hidden layer, the Sigmoid function in the second hidden layer, and the Tanh function in the third hidden layer. This diverse combination of activation functions can make the model better adapt to complex nonlinear relationships.

[0116] The function of the fully connected layer is to integrate the outputs of the previous hidden layers to generate the final prediction result. In this embodiment, the present invention uses a fully connected layer of 1024 neurons to fully capture the complex interactive relationship between various features.

[0117] Finally, the output layer will give the final illumination duration and illumination intensity prediction values. Here, the present invention will use a linear activation function to obtain continuous prediction values.

[0118] In a specific embodiment of the present invention, the retraining and optimization process of the system is as follows:

[0119] The process also includes the following steps: regularly retraining the neural network model of light duration to adapt to seasonal changes in the medicinal material growth environment; conducting in-depth mining of long-term accumulated planting data through cloud servers to optimize the structure and parameters of the neural network model; and providing a human-computer interaction interface that allows growers to manually adjust system parameters and view historical data.

[0120] In another embodiment of the present invention, the system retrains the neural network model once a month. This is because the growth characteristics of medicinal materials change with the seasons. For example, in summer, ginseng requires a shorter light duration (such as 7 hours) and a lower irradiation intensity (such as 40%), while in winter, ginseng requires a relatively longer light duration (such as 9 hours) and a higher irradiation intensity (such as 60%). Through regular retraining, the model of the present invention can adapt to these seasonal changes and always maintain a high degree of prediction accuracy.

[0121] In another embodiment of the present invention, the following steps are also included:

[0122] Outputting the predicted illumination duration and the corresponding predicted illumination intensity to the light sensor component;

[0123] The light sensor assembly receives the predicted illumination duration and the corresponding predicted illumination intensity, and determines the illumination intensity according to the predicted illumination intensity and determines the illumination duration according to the predicted illumination duration;

[0124] The medicinal material growth lighting process regulated by red light is executed according to the illumination intensity and illumination duration.

[0125] In addition, the system also uses cloud servers to deeply mine the long-term accumulated planting data. For example, by analyzing many years of planting data, the present invention will find that certain specific combinations of environmental conditions are particularly beneficial to the growth of ginseng. For example, the present invention found that under the conditions of temperature 20-22°C, humidity 65-70%, and light intensity 50-55%, ginseng grows fastest and has the best quality. These findings can be used to optimize the structure and parameters of the neural network model so that the model can better capture these complex relationships.

[0126] The human-computer interaction interface provides an intuitive operating platform for growers. Through this interface, growers can manually adjust system parameters and view historical data and forecast results. For example, if the grower believes that the current light intensity is too high based on his own experience, he can manually adjust the intensity from 60% to 55% through the interface. The system will record this adjustment and take this human factor into account in future forecasts. This human-computer collaboration method fully utilizes the advantages of artificial intelligence while retaining the experience and judgment of human experts, thereby achieving more intelligent and flexible planting management.

[0127] In a specific embodiment of the present invention, the specific implementation of the system is as follows:

[0128] The method is applied to a medicinal material growth lighting optimization system based on red light regulation, the system comprising: a sensor component 1, used to obtain environmental parameters; a red light lamp 2, used to provide adjustable lighting; a neural network model 3 of lighting duration, installed in a planting space, used to receive environmental parameters detected by the sensor component 1, and generate the neural network model 3 of lighting duration; a communication module 4, used for data transmission; a cloud server 5, used to receive the predicted lighting duration and the corresponding predicted irradiation intensity, and send the irradiation intensity and lighting duration to the red light lamp 2 through the communication module 4.

[0129] In a specific embodiment of the present invention, the sensor assembly 1 includes a temperature sensor, a humidity sensor, and a light intensity sensor. These sensors are arranged in different locations of the planting space to obtain comprehensive environmental data. For example, the present invention places a set of sensors at the upper, middle, and lower heights of the planting area to capture environmental differences at different heights.

[0130] The red light fixture 2 uses an LED light source, and can accurately adjust the light intensity from 0% to 100%. Preferably, the wavelength range of the red light fixture 2 is 630-660nm, which is the wavelength range with the highest photosynthesis efficiency for most medicinal materials.

[0131] The neural network model 3 is deployed on a low-power embedded system, which can process sensor data in real time and give prediction results. The communication module 4 uses 5G technology to ensure the real-time and stability of data transmission.

[0132] The cloud server 5 is not only responsible for storing and analyzing a large amount of historical data, but also undertakes the task of model retraining. Through this distributed architecture design, the present invention not only ensures the real-time response capability of the system, but also provides powerful background support.

[0133] In a specific embodiment of the present invention, the system further includes a data storage and analysis module 6, which is used to: store long-term accumulated planting data; perform in-depth mining on the planting data; and optimize the structure and parameters of the neural network model 3 according to the analysis results.

[0134] In another embodiment of the present invention, the data storage and analysis module 6 adopts a distributed storage system, which can efficiently store and process PB-level planting data. This module not only simply stores data, but also can perform in-depth data mining.

[0135] For example, through the analysis of historical data, the present invention will find that certain meteorological factors (such as air pressure changes) also have a significant impact on the growth of medicinal materials. Based on this discovery, the present invention will add an air pressure sensor to the next version of the system and adjust the structure of the neural network model 3 accordingly, adding an input node for receiving air pressure data.

[0136] In addition, data analysis also reveals some nonlinear relationships. For example, the present invention finds that there is an interaction between temperature and humidity: at low temperatures, higher humidity is conducive to the growth of medicinal materials; while at high temperatures, lower humidity is more beneficial. Based on this discovery, the present invention adds some feature cross layers to the neural network model 3 to better capture this complex nonlinear relationship.

[0137] In a specific embodiment of the present invention, the system also includes a model retraining module 7 and a human-computer interaction interface 8, wherein: the model retraining module 7 is used to regularly retrain the light duration neural network model 3 to adapt the model to the seasonal changes in the medicinal material growth environment, and update the model parameters to improve the prediction accuracy; the human-computer interaction interface 8 is used to allow growers to manually adjust system parameters, display historical data and prediction results, and provide real-time feedback on the system operation status.

[0138] In a preferred embodiment of the present invention, the model retraining module 7 adopts an incremental learning method. Whenever new planting data is generated, the system does not need to retrain the entire model, but can fine-tune the existing model based on the new data. This method greatly improves the efficiency of retraining and enables the system to quickly adapt to changes in the environment.

[0139] For example, if the planting data from the most recent week shows that slightly extending the light duration (such as from 8 hours to 8.5 hours) can significantly increase the growth rate of ginseng in the current season, the model retraining module 7 will quickly adjust the model parameters so that it tends to give slightly longer light duration recommendations in subsequent predictions.

[0140] The human-computer interaction interface 8 adopts an intuitive graphical design. Growers can easily view the real-time data and historical trends of various environmental parameters by touching the screen. There is also an "expert mode" button on the interface, which can display more detailed system operating parameters when activated, such as the current prediction accuracy of the neural network model 3 and the working status of each sensor.

[0141] In addition, the human-computer interaction interface 8 also provides a "manual adjustment" function. If growers observe any abnormal conditions in the medicinal materials, they can temporarily override the automatic control of the system through this function. For example, if the ginseng leaves are found to be slightly yellow, the grower will determine that this is caused by insufficient light, so he will increase the light intensity from the current 50% to 55% through the interface. The system will record this manual adjustment and use it as important feedback information for subsequent model optimization.

[0142] In general, the present invention realizes intelligent and precise control of the medicinal material growth environment by combining advanced neural network technology with traditional planting experience. It can not only improve the yield and quality of medicinal materials, but also significantly reduce labor costs and energy consumption, providing an innovative solution for modern agricultural production. Through continuous data accumulation and model optimization, this system is expected to develop the best growth strategies for different types of medicinal materials in the future, and promote the entire Chinese medicinal material planting industry to develop in a more scientific and efficient direction.

[0143] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing illumination for medicinal material growth based on red light regulation, characterized in that: The steps include: Acquiring environmental parameters, and processing the environmental parameters to generate environmental characteristic data; Set the illumination duration for the growth of medicinal materials, and at the same time set the illumination intensity of the red light fixture within the illumination duration; During the illumination period, real-time detection of environmental parameters for the growth of medicinal materials is performed, and the real-time detected environmental parameters are processed to obtain environmental characteristic data; After inputting the environmental characteristic data into the neural network model of the illumination duration, the predicted illumination duration is output, and it is determined whether the output predicted illumination duration is consistent with the illumination duration for the growth of the medicinal material. If so, the illumination intensity of the red light fixture is kept unchanged; if not, the illumination intensity of the red light fixture is updated according to the predicted illumination duration; According to the updated red light intensity, adjust the red light in the medicinal material growth environment to achieve real-time optimization control of the medicinal material growth environment; Among them, the training method of the neural network model of the lighting duration includes: initializing the parameters of the neural network; inputting the empirical data into the neural network model; outputting the predicted lighting duration and the illumination intensity of the red light fixture within the corresponding predicted lighting duration; calculating the empirical error; judging whether the empirical error is greater than a first preset threshold, and if so, updating the parameters of the neural network by back-propagating the empirical error through the gradient descent method; if not, returning to the step of inputting the empirical data into the neural network model.

2. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: Before the step of inputting the empirical data into the neural network model, it also includes: Reading empirical data, wherein the empirical data is composed of a number of environmental characteristic data and corresponding neural network prediction data of illumination duration; The empirical data is obtained through historical planting records, expert knowledge base and experimental data, and is processed through data cleaning and standardization; After the step of calculating the empirical error, the method further includes: determining whether the empirical error is greater than a second preset threshold value, and if so, determining a neural network model and parameters of the illumination duration, and outputting the neural network model of the illumination duration; if not, returning to the step of inputting the empirical data into the neural network model; Among them, the first preset threshold is smaller than the second preset threshold to achieve fine control and rapid convergence of model training.

3. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: The environmental parameters include temperature, humidity and light intensity detected by the photosensor; The steps of processing the real-time detected environmental parameters to obtain environmental characteristic data are specifically: normalizing the environmental parameters to obtain environmental characteristic data; Among them, the normalization process uses the minimum-maximum scaling method to ensure that all feature data are within the interval [0, 1]; The illumination intensity and illumination duration of the red light fixture meet the following conditions: within the illumination duration, the light is irradiated at an initialized illumination intensity, and after illumination, the initialized illumination intensity is left to stand for a set time. Within the set time, the sum of the temperature and humidity detected in real time is a constant A, the product of the temperature and humidity is a constant B, and the light intensity detected by the photosensor within the set time after illumination is a constant C. Among them, the value ranges of constants A, B, and C are dynamically adjusted according to different types of medicinal materials and growth stages.

4. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: The loss function of the neural network model training for the illumination duration is: Among them, L is the loss function value, α is the first hyperparameter, t i is the illumination duration in the empirical data, is the illumination duration predicted by the neural network model of illumination duration, β is the second hyperparameter, and p i is the light intensity detected by the photosensor in the empirical data, is the light intensity detected by the photosensor predicted by the neural network model of the illumination duration, and T is the total number of empirical data; the values ​​of α and β are determined by the cross-validation method to balance the accuracy of the illumination duration and light intensity prediction; The loss function also contains a regularization term to prevent the model from overfitting.

5. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: The neural network model of the illumination duration includes: The input layer, hidden layer, fully connected layer, and output layer are connected in sequence; The input layer is used to sequentially receive the temperature, humidity and light intensity detected by the photosensor; The hidden layer receives the temperature, humidity and light intensity detected by the photosensor, and outputs a lighting duration and a corresponding illumination intensity; The hidden layer includes a first hidden layer, a second hidden layer, a third hidden layer, a fourth hidden layer and a fifth hidden layer connected in sequence; The illumination duration output by the first hidden layer has a value range of 0-5 hours, and the illumination intensity output by the second hidden layer has a value range of 0-200; The illumination duration output by the first hidden layer is in the range of 6-8 hours, and the illumination intensity output by the third hidden layer is in the range of 210-240; The illumination duration output by the fourth hidden layer has a value range of 9-12 hours, and the illumination intensity output by the fifth hidden layer has a value range of 241-300; The fully connected layer receives the illumination duration and illumination intensity, and outputs them; The output layer receives the illumination duration and illumination intensity, and outputs them; Among them, each hidden layer uses a different activation function to enhance the nonlinear fitting ability of the model.

6. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: The following steps are also included: Regularly retrain the neural network model of light duration to adapt to seasonal changes in the medicinal material growing environment; Through cloud servers, we conduct in-depth mining of long-term accumulated planting data to optimize the structure and parameters of the neural network model; Provides a human-computer interaction interface that allows growers to manually adjust system parameters and view historical data.

7. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 1, characterized in that: The following steps are also included: Outputting the predicted illumination duration and the corresponding predicted illumination intensity to the light sensor component; The light sensor assembly receives the predicted illumination duration and the corresponding predicted illumination intensity, and determines the illumination intensity according to the predicted illumination intensity and determines the illumination duration according to the predicted illumination duration; The medicinal material growth lighting process regulated by red light is executed according to the illumination intensity and illumination duration.

8. A medicinal material growth illumination optimization system based on red light regulation, which implements the method according to any one of claims 1 to 7, characterized in that: The system includes: A sensor component for acquiring environmental parameters; Red light fixtures for providing adjustable lighting; A neural network model of light duration, installed in the planting space, for receiving environmental parameters detected by the sensor assembly and generating a neural network model of light duration; Communication module, used for data transmission; The cloud server is used to receive the predicted illumination duration and the corresponding predicted illumination intensity, and send the illumination intensity and illumination duration to the red light fixture through the communication module.

9. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 8, characterized in that: The system also includes a data storage and analysis module, which is used to: store long-term accumulated planting data; conduct in-depth mining on the planting data; and optimize the structure and parameters of the neural network model according to the analysis results.

10. The method for optimizing illumination of medicinal material growth based on red light regulation according to claim 8, characterized in that: The system also includes a model retraining module and a human-computer interaction interface, wherein: The model retraining module is used to regularly retrain the light duration neural network model to adapt the model to seasonal changes in the medicinal material growth environment and update the model parameters to improve the prediction accuracy; The human-computer interaction interface is used to allow growers to manually adjust system parameters, display historical data and forecast results, and provide real-time feedback on the system operating status.

Images