Intelligent irrigation and fertilization method and decision system for traditional Chinese medicinal material standard base based on internet of things

Abstract

The application discloses a traditional Chinese medicinal material standard base intelligent irrigation and fertilization method and decision system based on Internet of Things, and relates to the technical field of agricultural informatization.The application solves the limitations of the existing irrigation and fertilization method through multi-source data acquisition and intelligent analysis, realizes accurate decision and resource optimization, obtains comprehensive data by using a soil moisture sensor, a weather station and a plant image acquisition device, and combines principal component analysis, a long short-term memory neural network and a blockchain technology to accurately predict water and fertilizer demand, optimize an irrigation and fertilization scheme, ensure data credibility and traceability, simultaneously realize real-time control and collaborative operation of equipment through edge computing and a wireless sensor network, improve resource utilization efficiency, reduce production cost, and significantly improve the scientific nature and sustainability of traditional Chinese medicinal material planting.

Description

Technical Field

[0001] This invention relates to the field of agricultural information technology, specifically to an intelligent irrigation and fertilization method and decision-making system for standard bases of Chinese medicinal herbs based on the Internet of Things. Background Technology

[0002] The cultivation of Chinese medicinal herbs is an important area of ​​agricultural modernization, directly related to the quality, yield, and ecological sustainability of medicinal herbs, and has significant economic and social value. Precision irrigation and fertilization, as key links to improve the production efficiency of Chinese medicinal herbs, have received widespread attention in recent years. However, existing methods mostly rely on single sensor data or traditional experience models, which are difficult to fully capture the dynamic changes in soil, weather, and plant growth. This results in a lack of targeted water and fertilizer application, low resource utilization efficiency, and insufficient transparency in data management, making it difficult to meet the environmental traceability requirements of high-quality medicinal herb production. In addition, existing systems often face the contradiction between limited computing resources and the need for real-time decision-making in complex farmland environments, making it difficult to achieve efficient collaborative control. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent irrigation and fertilization method and decision-making system for standard Chinese medicinal herb bases based on the Internet of Things. Through multi-source data collection, intelligent analysis and blockchain technology, it realizes precise irrigation and fertilization of Chinese medicinal herb cultivation.

[0004] The objective of this invention can be achieved through the following technical solutions:

[0005] This application provides an intelligent irrigation and fertilization method for standard Chinese medicinal herb bases based on the Internet of Things, including the following steps:

[0006] Multi-source heterogeneous data, including soil moisture, nutrient content, temperature, rainfall and plant growth status, were obtained from soil moisture sensors, weather stations and plant image acquisition equipment. Principal component analysis was then used to extract key features of soil moisture, weather conditions and plant growth status to obtain a dimensionality-reduced feature set.

[0007] A growth requirement model is established based on the feature set, and then a long short-term memory neural network is used to train the growth requirement model. The inputs are soil moisture, nutrient content, meteorological conditions and plant growth images, and the outputs are the water and fertilizer requirements for each growth stage, thus obtaining the growth requirement prediction results.

[0008] When the water and fertilizer requirements in the growth demand forecast exceed the preset threshold, a linear programming algorithm is used to optimize irrigation and fertilization by inputting water and fertilizer requirements, soil moisture and weather conditions, and obtaining a precise irrigation and fertilization plan.

[0009] Based on the precision irrigation and fertilization plan, blockchain technology is used to store the plan data and sensor data. The irrigation and fertilization amount, soil moisture and weather records are input to generate encrypted data blocks and obtain trusted data storage results.

[0010] Encrypted data blocks are obtained from trusted data storage results, decrypted and analyzed using edge computing devices, and then real-time control commands are generated by inputting irrigation and fertilization plans and real-time sensor data to obtain a collaborative control command set.

[0011] Based on the collaborative control instruction set, the instructions are transmitted to the irrigation and fertilization equipment via a wireless sensor network. The control instructions and equipment status are input to execute precise water and fertilizer application, obtain water and fertilizer application records, and then the water and fertilizer use efficiency is analyzed using a random forest algorithm. The application records, soil moisture and plant growth status are input to judge the effect of water and fertilizer application and obtain optimization adjustment parameters.

[0012] Based on the optimized adjustment parameters, the gradient descent algorithm is used to update the growth demand model parameters. The adjustment parameters and real-time feature set are input to optimize the model prediction accuracy and obtain the updated growth demand model.

[0013] Furthermore, the reduced feature set is obtained, specifically including:

[0014] Standardized data on soil moisture, meteorological conditions, and plant growth were preprocessed using standardized methods to obtain a standardized dataset.

[0015] Principal component analysis is used to extract features from a standardized dataset to obtain a key feature set. When the principal component contribution rate of the key feature set is lower than a preset threshold, the parameters of principal component analysis are adjusted, and the key features are re-extracted to obtain an optimized feature set.

[0016] Based on the optimized feature set, the K-means clustering algorithm was used to group soil moisture, meteorological conditions and plant growth status to obtain feature grouping results. Then, the correlation coefficient between soil moisture and plant growth status in each group was calculated to obtain the correlation analysis results.

[0017] If the correlation coefficient in the correlation analysis results is lower than the preset threshold, the grouping results are re-clustered to obtain updated grouping results, and a dimensionality-reduced feature set is generated to determine the final feature set.

[0018] Furthermore, the growth requirement prediction results are obtained, specifically including:

[0019] Soil moisture, nutrient content, meteorological conditions and plant image data are acquired. Soil and meteorological information are collected through sensors, and plant images are captured using a high-resolution camera to obtain the input dataset. Then, image preprocessing techniques are used to denoise and segment the plant images to obtain the second image dataset.

[0020] By standardizing soil and meteorological data, a normalized dataset is obtained. Based on the normalized dataset and the second image dataset, texture and color features of plant images are extracted. Combined with soil moisture and meteorological conditions, a multi-dimensional feature vector is constructed to obtain a comprehensive feature set.

[0021] The growth requirement model is trained using a long short-term memory neural network. The comprehensive feature set is input, and if the training set loss is lower than a preset threshold, the training is stopped, and the trained model is obtained.

[0022] By using the trained model and inputting a comprehensive feature set collected in real time, the water and fertilizer requirements for each growth stage are predicted, and the predicted requirements are obtained.

[0023] Furthermore, a precise irrigation and fertilization plan is obtained, specifically including:

[0024] When the water and fertilizer demand exceeds the preset threshold, the water and fertilizer demand, soil moisture and meteorological conditions are obtained from the prediction results to determine the input dataset. Then, the input dataset is processed by the linear programming algorithm to optimize the irrigation and fertilizer application and obtain a preliminary irrigation and fertilization plan.

[0025] Based on the preliminary irrigation and fertilization plan, combined with soil moisture and meteorological conditions, it is determined whether the plan meets the constraints of precision irrigation and fertilization, and an adjusted irrigation and fertilization plan is obtained. If the adjusted irrigation and fertilization plan meets the constraints, a time series analysis algorithm is used to predict the future trend of water and fertilizer demand and obtain demand forecast data.

[0026] Based on demand forecast data, combined with current soil moisture and weather conditions, the input parameters of the linear programming algorithm are updated to obtain an optimized precision irrigation and fertilization plan.

[0027] Based on the optimized precision irrigation and fertilization plan, real-time control commands are generated and output to the irrigation and fertilization equipment to determine the execution plan. After the execution plan is running, real-time data on soil moisture and weather conditions are collected to determine whether the water and fertilizer requirements need to be adjusted, and an updated input dataset is obtained.

[0028] Furthermore, the trusted data storage results are obtained, specifically including:

[0029] By combining solution data records with sensor data through blockchain technology, encrypted data blocks are generated and stored in a distributed ledger to obtain trusted data storage results. When the storage timestamp of the encrypted data block is later than a preset threshold, data verification is triggered through a smart contract to determine the data integrity and obtain the verification result.

[0030] Based on the verification results, the correlation between soil moisture and meteorological records is analyzed using a decision tree algorithm to determine the adjustment parameters for irrigation and fertilization. The adjustment parameters are then broadcast through the blockchain network to obtain consensus confirmation from distributed nodes, thus obtaining the final irrigation and fertilization plan.

[0031] Key parameters are extracted from the final irrigation and fertilization plan, and control commands are generated using data mapping technology and output to the irrigation equipment to obtain the execution results.

[0032] Furthermore, a coordinated control instruction set is obtained, specifically including:

[0033] The encrypted data block is obtained from trusted storage. The integrity of the data is verified using a preset key. If the integrity is verified, the encrypted data block is determined to be valid. The valid encrypted data block is decrypted through an edge device. The data structure is parsed using the AES algorithm to obtain the decrypted data.

[0034] Environmental and crop parameters are extracted from the decrypted data, sensor data is fused to generate a fused dataset, and a decision tree algorithm is used to determine irrigation and fertilization needs and real-time control parameters based on the fused dataset and the preset irrigation and fertilization plan.

[0035] Based on real-time control parameters, generate equipment control commands. If the parameters exceed the preset threshold, adjust the command priority to obtain a set of control commands.

[0036] The control instruction set is integrated by a collaborative scheduling algorithm to generate a collaborative control instruction set. Then, the execution sequence is extracted from the collaborative control instruction set and output to the agricultural equipment for execution.

[0037] Furthermore, water and fertilizer application records were obtained, specifically including:

[0038] The system collects environmental parameters and device status through a wireless sensor network, generates sensor data, and generates control commands based on collaborative control logic when the sensor data exceeds a preset threshold.

[0039] Control commands are sent to irrigation and fertilization equipment via network transmission. The command reception status is determined, and the operating data of the irrigation and fertilization equipment is obtained based on the command execution status to determine the degree of accuracy in application.

[0040] By comparing operational data with sensor data, the random forest algorithm is used to optimize the water and fertilizer management strategy and obtain the adjusted control commands.

[0041] Based on the adjusted control instructions, a new round of precise water and fertilizer application is executed, application records are generated, key parameters are extracted from the application records, the effectiveness of water and fertilizer management is judged, and the collaborative control logic is updated.

[0042] Furthermore, the optimized adjustment parameters are obtained, specifically including:

[0043] We acquire water and fertilizer application records, soil moisture data, and plant growth status, and integrate them into a structured dataset, which includes water application data, fertilizer application amount, soil moisture changes, and plant growth indicators, to obtain a preprocessed dataset.

[0044] Then, the random forest algorithm is used to train the preprocessed dataset to build a water and fertilizer use efficiency prediction model. The input data are water application amount, fertilizer application amount, soil moisture change and plant growth index, and the output is the predicted value of water and fertilizer use efficiency.

[0045] The predicted value of water and fertilizer use efficiency is compared with the preset water and fertilizer use efficiency threshold. The deviation between the water and fertilizer use efficiency and the preset threshold is calculated. When the deviation exceeds the preset threshold, it is marked as inefficient application, and the application effect judgment result is obtained.

[0046] Based on the results of the application effect judgment, records of inefficient application are extracted, and the distribution characteristics of water application data and fertilizer application amount are analyzed to obtain the abnormal application amount pattern. Then, the linear regression algorithm is used to fit the abnormal application amount pattern. Input water application data, fertilizer application amount and soil moisture changes, output the adjusted application amount suggestion, and obtain the preliminary optimization parameters.

[0047] Based on the preliminary optimization parameters, combined with plant growth indicators and soil moisture changes, a preset rule base is used for correction. When the matching degree between the optimized parameters and the growth indicators in the rule base is lower than the preset threshold, the parameter values ​​are adjusted to obtain the final optimized adjustment parameters.

[0048] By finally optimizing and adjusting the parameters, updating the water and fertilizer application records, generating a new application plan, and outputting it to the database, the optimized application scheme is obtained.

[0049] Furthermore, the updated growth requirement model is obtained, specifically including:

[0050] The real-time feature set and adjustment parameters are acquired through the real-time acquisition system, stored as a standardized dataset, and the processed input data is obtained. Then, the data preprocessing module is used to normalize and denoise the input data to generate a feature set suitable for model training.

[0051] When the dimension of the feature set exceeds the preset threshold, the dimensionality is reduced by principal component analysis to obtain a compressed feature set. Based on the compressed feature set, the parameters of the growth requirement model are iteratively updated using the gradient descent algorithm, the loss function value is calculated, and a preliminary optimized model is obtained.

[0052] The prediction accuracy of the initial optimized model is evaluated by cross-validation, and the accuracy evaluation index is calculated to obtain model performance data. If the model performance data does not reach the preset threshold, the learning rate is adjusted and the gradient descent algorithm is re-executed to obtain the final optimized model and output the updated growth requirement model parameters.

[0053] This application provides an IoT-based intelligent irrigation and fertilization decision-making system for standard medicinal herb production bases, which includes the following methods for implementing IoT-based intelligent irrigation and fertilization in standard medicinal herb production bases:

[0054] The data acquisition module obtains multi-source heterogeneous data from soil moisture sensors, weather stations, and plant image acquisition devices, including soil moisture, nutrient content, temperature, rainfall, and plant growth status.

[0055] The preprocessing and feature extraction module standardizes the collected data, extracts key features using principal component analysis, reduces data dimensionality, and optimizes the feature set.

[0056] The growth demand prediction module constructs and trains a growth demand model based on the optimized feature set, and uses a long short-term memory neural network to predict the water and fertilizer demand at each growth stage, generating an optimized water and fertilizer scheduling scheme.

[0057] The irrigation and fertilization plan generation module uses a linear programming algorithm to optimize irrigation and fertilization amounts when the predicted water and fertilizer demand exceeds a preset threshold. It combines soil moisture and weather conditions to generate a precise irrigation and fertilization plan.

[0058] The data storage and verification module uses blockchain technology to store irrigation and fertilization plans and sensor data, generates encrypted data blocks and stores them in a distributed ledger, and verifies data integrity through smart contracts.

[0059] The real-time control and collaborative scheduling module retrieves encrypted data blocks from trusted data storage, decrypts and analyzes the data through edge computing devices, generates a collaborative control instruction set, and performs collaborative operations of irrigation and fertilization equipment.

[0060] The efficiency assessment and optimization module transmits control commands to the irrigation and fertilization equipment through a wireless sensor network to execute precise water and fertilizer application. It uses a random forest algorithm to analyze water and fertilizer use efficiency and optimize the water and fertilizer application plan.

[0061] The model update module updates the growth requirement model parameters based on the optimized adjustment parameters and uses the gradient descent algorithm to optimize the model prediction accuracy.

[0062] The beneficial effects of this invention are as follows:

[0063] This invention overcomes the limitations of existing methods that rely on single sensor data or traditional empirical models by comprehensively collecting and extracting features from multi-source heterogeneous data. It acquires multi-dimensional data through soil moisture sensors, weather stations, and plant image acquisition equipment, and combines principal component analysis and long short-term memory neural networks to accurately predict the water and fertilizer requirements of Chinese medicinal herbs at each growth stage. This process not only improves the interpretability of the data and the accuracy of the model, but also provides a scientific basis for precision irrigation and fertilization, effectively enhancing the scientific nature and precision of Chinese medicinal herb cultivation.

[0064] By employing a linear programming algorithm to optimize irrigation and fertilization amounts, and combining this with blockchain technology to store scheme data and sensor data, the problem of insufficient targeting of water and fertilizer application and lack of transparency in data management in existing methods is solved. Blockchain technology ensures the credibility and traceability of data, meeting the environmental traceability requirements of high-quality medicinal herb production. At the same time, the optimized irrigation and fertilization scheme significantly improves resource utilization efficiency, reduces water and fertilizer waste, lowers production costs, and achieves sustainable development of Chinese medicinal herb cultivation.

[0065] By utilizing edge computing devices and wireless sensor networks, real-time control and collaborative operation of irrigation and fertilization equipment are achieved, resolving the contradiction between limited computing resources and real-time decision-making requirements in complex farmland environments. Through the rapid response of edge computing and the data security guarantee of blockchain technology, precise water and fertilizer application tasks can be executed efficiently. At the same time, the random forest algorithm is used to analyze water and fertilizer utilization efficiency, further optimizing the application plan, ensuring the high efficiency and flexibility of Chinese medicinal herb cultivation, and improving overall production efficiency. Attached Figure Description

[0066] To better understand and implement this application, the technical solution is described in detail below with reference to the accompanying drawings.

[0067] Figure 1 A schematic diagram of the process for an intelligent irrigation and fertilization method for a standard Chinese medicinal herb base based on the Internet of Things, provided in Embodiment 1 of this application;

[0068] Figure 2 A schematic diagram of the process for obtaining growth demand prediction results using the IoT-based intelligent irrigation and fertilization method for standard Chinese medicinal herb bases provided in Embodiment 1 of this application;

[0069] Figure 3 A flowchart illustrating the process of obtaining a precision irrigation and fertilization scheme from the IoT-based intelligent irrigation and fertilization method for standard Chinese medicinal herb bases provided in Embodiment 1 of this application;

[0070] Figure 4 This is a schematic diagram of the structure of the Internet of Things-based intelligent irrigation and fertilization decision-making system for standard bases of Chinese medicinal materials provided in Embodiment 2 of this application. Detailed Implementation

[0071] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, exemplary embodiments will be described in detail below, examples of which are illustrated in the accompanying drawings. In the following description relating to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and systems consistent with some aspects of this application as detailed in the appended claims.

[0072] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0073] The following detailed description of the specific implementation methods, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided in detail.

[0074] Example 1

[0075] Please see Figures 1-3 This embodiment provides an intelligent irrigation and fertilization method for a standard Chinese medicinal herb base based on the Internet of Things, including the following steps:

[0076] S1. Obtain multi-source heterogeneous data from soil moisture sensors, weather stations, and plant image acquisition equipment, including soil moisture, nutrient content, temperature, rainfall, and plant growth status. Then, use principal component analysis to extract key features of soil moisture, meteorological conditions, and plant growth status to obtain a dimension-reduced feature set.

[0077] Further, in step S1, the dimensionality-reduced feature set is obtained, specifically including:

[0078] Standardized methods were used to preprocess the collected soil moisture, meteorological conditions, and plant growth data to obtain a standardized dataset.

[0079] Principal component analysis is used to extract features from a standardized dataset to obtain a key feature set. When the principal component contribution rate of the key feature set is lower than a preset threshold, the parameters of principal component analysis are adjusted, and the key features are re-extracted to obtain an optimized feature set.

[0080] Based on the optimized feature set, the K-means clustering algorithm was used to group soil moisture, meteorological conditions and plant growth status to obtain feature grouping results. Then, the correlation coefficient between soil moisture and plant growth status in each group was calculated to obtain the correlation analysis results.

[0081] If the correlation coefficient in the correlation analysis results is lower than the preset threshold, the grouping results are re-clustered to obtain updated grouping results, and a dimensionality-reduced feature set is generated to determine the final feature set.

[0082] Specifically, key features are accurately extracted from multi-source heterogeneous data and grouped reasonably according to feature correlation. This process not only effectively reduces data dimensionality and computational complexity, but also improves data interpretability and model accuracy. The resulting optimized feature set can more accurately reflect the intrinsic relationship between soil moisture, meteorological conditions and plant growth status, providing a scientific basis for subsequent irrigation and fertilization decisions, thereby achieving precision irrigation and fertilization and improving the yield and quality of Chinese medicinal herbs.

[0083] S2. Establish a growth demand model based on the feature set, and then train the growth demand model using a long short-term memory neural network. Input soil moisture, nutrient content, meteorological conditions and plant growth images, and output the water and fertilizer requirements for each growth stage to obtain the growth demand prediction results.

[0084] Further, in step S2, the growth demand prediction results are obtained, specifically including:

[0085] S21. Acquire soil moisture, nutrient content, meteorological conditions and plant image data. Collect soil and meteorological information through sensors, take plant images with a high-resolution camera to obtain the input dataset, and then use image preprocessing technology to denoise and segment the plant images to obtain the second image dataset.

[0086] S22. By standardizing soil and meteorological data, a normalized dataset is obtained. Based on the normalized dataset and the second image dataset, the texture and color features of the plant images are extracted. Combined with soil moisture and meteorological conditions, a multi-dimensional feature vector is constructed to obtain a comprehensive feature set.

[0087] S23. Use a long short-term memory neural network to train the growth requirement model. Input a comprehensive feature set. If the training set loss is lower than a preset threshold, stop training and obtain the trained model.

[0088] S24. Using the trained model and inputting the comprehensive feature set collected in real time, predict the water and fertilizer requirements for each growth stage and obtain the predicted requirements results.

[0089] S25. Based on the predicted demand results, when the predicted water and fertilizer demand exceeds the range supported by soil and meteorological conditions, the irrigation and fertilization parameters are adjusted to obtain an optimized water and fertilizer scheduling plan. For the optimized water and fertilizer scheduling plan, the irrigation and fertilization equipment are automatically adjusted by the control system to obtain the actual water and fertilizer supply results.

[0090] Specifically, sensor deployment is crucial for collecting data on soil moisture, nutrient content, meteorological conditions, and plant images. For example, in one possible implementation, soil moisture sensors can be buried at different depths in the farmland, such as 10 cm and 20 cm, to capture moisture changes in the root-active zone. Nutrient sensors detect nitrogen, phosphorus, and potassium content. Preferably, 2 to 3 sensors are deployed per acre to ensure data representativeness. A meteorological station is installed in an open area of ​​the field, avoiding obstructions from buildings and trees, to record real-time meteorological data such as temperature, rainfall, and light intensity. This data is essential for the base to understand the impact of environmental changes on the growth of medicinal herbs. High-resolution cameras are used for plant image acquisition, fixed to supports in the field, and taken at regular intervals daily to ensure consistent lighting conditions for each shot, obtaining clear and stable plant images and comprehensively recording the plant's growth status. This multi-source data acquisition method obtains information from multiple dimensions, including soil, meteorology, and plants, allowing the base to fully understand the environmental and plant conditions.

[0091] For plant image preprocessing, image denoising and segmentation are key. Specifically, median filtering can be used to remove noise and preserve edge details, while segmentation can be achieved by separating plants from the background using color thresholding. For example, in one embodiment, a green channel threshold is set to extract the plant area as a binary image. This method can effectively isolate plants, reduce background interference, and provide clear input for feature extraction. Standardization of soil and meteorological data requires unified units.

[0092] Understandably, humidity data is expressed as a percentage, nutrient content as milligrams per kilogram, and meteorological data such as temperature as degrees Celsius. Through min-max normalization, the data is mapped to the range of 0 to 1. For example, soil moisture ranges from 20% to 80%, and after normalization, it is 0.25 to 1. This processing ensures that data of different dimensions can be directly compared. The construction of the comprehensive feature set involves the fusion of image texture, color features, and environmental data. For example, the gray-level co-occurrence matrix of plant images is extracted to quantify texture roughness; color features are characterized by the RGB mean to represent the health status of leaves; and soil moisture, nutrient, and meteorological data are directly added as numerical features.

[0093] For example, the feature vector may include texture contrast of 0.6, green mean of 120, soil moisture of 0.5, etc. This multi-dimensional feature set can comprehensively reflect the plant and environmental status. Long short-term memory neural network is used to train the growth demand model. Preferably, the network contains two layers of LSTM units, each with 128 nodes. The input is a time series feature set, such as the comprehensive features of 7 consecutive days as input, to predict the water and fertilizer demand for the next day. During training, the loss function adopts the mean squared error. If the loss is lower than 0.01, the training stops. This model can capture time dependence and is suitable for dynamic prediction.

[0094] Water and fertilizer demand prediction is based on a real-time feature input model. For example, given the current feature set, the model outputs an irrigation volume of 200 liters per acre and a nitrogen fertilizer application rate of 5 kg per acre. If the predicted demand exceeds the soil moisture support range, such as a moisture content of only 30%, the irrigation volume is adjusted to 250 liters. This dynamic adjustment ensures that water and fertilizer supply matches actual demand. The optimized water and fertilizer scheduling plan is executed by the control system. Specifically, the irrigation equipment is connected to the drip irrigation pipeline, and the fertilizer applicator injects fertilizer according to the set amount. For example, the control system receives the scheduling plan, automatically starts drip irrigation for 2 hours, and the fertilizer applicator releases 5 kg of nitrogen fertilizer. This automated execution improves efficiency and ensures accurate supply.

[0095] S3. When the water and fertilizer demand in the growth demand prediction results exceeds the preset threshold, the linear programming algorithm is used to optimize the irrigation and fertilization amount by inputting the water and fertilizer demand, soil moisture and meteorological conditions, and obtaining a precise irrigation and fertilization plan.

[0096] Furthermore, in step S3, a precise irrigation and fertilization plan is obtained, specifically including:

[0097] S31. When the water and fertilizer demand exceeds the preset threshold, the water and fertilizer demand, soil moisture and meteorological conditions are obtained from the prediction results to determine the input dataset. Then, the input dataset is processed by the linear programming algorithm to optimize the irrigation and fertilization amount and obtain a preliminary irrigation and fertilization plan.

[0098] S32. Based on the preliminary irrigation and fertilization plan, combined with soil moisture and meteorological conditions, determine whether the plan meets the constraints of precision irrigation and fertilization, and obtain the adjusted irrigation and fertilization plan. If the adjusted irrigation and fertilization plan meets the constraints, then use the time series analysis algorithm to predict the future trend of water and fertilizer demand and obtain demand forecast data.

[0099] S33. Based on the demand forecast data, combined with the current soil moisture and weather conditions, update the input parameters of the linear programming algorithm to obtain the optimized precision irrigation and fertilization plan.

[0100] S34. Generate real-time control commands based on the optimized precision irrigation and fertilization plan, output them to the irrigation and fertilization equipment, determine the execution plan, and collect real-time data on soil moisture and weather conditions after the execution plan is running to determine whether the water and fertilizer requirements need to be adjusted, and obtain the updated input dataset.

[0101] Specifically, in the IoT-based standard base for Chinese medicinal herbs, when the demand for water and fertilizer exceeds a preset threshold, the system will initiate an optimization process. First, it obtains the water and fertilizer demand, soil moisture, and meteorological conditions from the prediction results to form an input dataset. For example, during the critical growth period of Chinese medicinal herbs in a certain plot, the soil moisture sensor measures a moisture content of 15% (lower than the suitable value of 18%–22%), and the weather station shows no rainfall and a temperature of 25°C. These data directly reflect the current growth environment. A linear programming algorithm is used to optimize irrigation and fertilization, with minimizing water and fertilizer input as the objective function. Combined with crop demand constraints (such as 400 cubic meters of water and 20 kg of nitrogen fertilizer per acre), a preliminary plan is calculated based on real-time data. For example, it is recommended to irrigate 300 cubic meters and apply 15 kg of fertilizer, which ensures growth needs while avoiding resource waste. When judging the preliminary plan, it is necessary to verify whether the soil moisture has recovered to the suitable range. If the soil moisture content rises to 20% after irrigation, and the meteorological conditions indicate high temperature and little rain in the future, the irrigation amount is adjusted to 350 cubic meters. The scheme that passes the constraint verification proceeds to the next step.

[0102] Using time series analysis algorithms, the water and fertilizer demand trends of Chinese medicinal herbs are predicted based on historical data during their growth stages. For example, analyzing the water requirement patterns from the early to mid-growth stages, the algorithm predicts a 10% increase in water demand over the next week, providing a basis for dynamic adjustments. When updating the plan, the algorithm recalculates by combining the predicted data with real-time soil moisture. If the predicted water demand is 440 cubic meters and the current soil moisture content is 16%, the algorithm regenerates a plan: irrigate 380 cubic meters and apply 18 kilograms of fertilizer. The control system translates the optimized plan into equipment instructions, precisely controlling drip irrigation and fertilization. After the plan is executed, real-time data is collected for feedback and adjustments. For example, if the soil moisture content meets the standard after irrigation, but rainfall exceeds expectations, the irrigation amount for the next day is reduced, forming a closed-loop mechanism to ensure that water and fertilizer supply always meets actual needs. This multi-stage collaborative approach achieves the dual goals of efficient resource utilization and increased yield.

[0103] S4. Based on the precision irrigation and fertilization plan, use blockchain technology to store the plan data and sensor data, input irrigation and fertilization amount, soil moisture and weather records, generate encrypted data blocks, and obtain trusted data storage results;

[0104] Further, in step S4, the trusted data storage result is obtained, specifically including:

[0105] By combining solution data records with sensor data through blockchain technology, encrypted data blocks are generated and stored in a distributed ledger to obtain trusted data storage results. When the storage timestamp of the encrypted data block is later than a preset threshold, data verification is triggered through a smart contract to determine the data integrity and obtain the verification result.

[0106] Based on the verification results, the correlation between soil moisture and meteorological records is analyzed using a decision tree algorithm to determine the adjustment parameters for irrigation and fertilization. The adjustment parameters are then broadcast through the blockchain network to obtain consensus confirmation from distributed nodes, thus obtaining the final irrigation and fertilization plan.

[0107] Key parameters are extracted from the final irrigation and fertilization plan, and control commands are generated using data mapping technology and output to the irrigation equipment to obtain the execution results.

[0108] Specifically, this method uses blockchain to store irrigation and fertilization plans and sensor data, generating encrypted data blocks and verifying data integrity using smart contracts. Based on the verification results, a decision tree algorithm is used to analyze the correlation between soil moisture and meteorological records, determining irrigation and fertilization adjustment parameters. These parameters are then confirmed through consensus obtained via the blockchain network, forming the final plan. Finally, the key parameters are converted into control commands and output to the irrigation equipment for execution. This method achieves precise control and intelligent management of irrigation and fertilization, ensuring data reliability and traceability, and improving the scientific nature and resource utilization efficiency of medicinal herb cultivation.

[0109] S5. Obtain encrypted data blocks from trusted data storage results, decrypt and analyze the data using edge computing devices, input irrigation and fertilization schemes and real-time sensor data, generate real-time control instructions, and obtain a collaborative control instruction set;

[0110] Further, in step S5, a coordinated control instruction set is obtained, specifically including:

[0111] The encrypted data block is obtained from trusted storage. The integrity of the data is verified using a preset key. If the integrity is verified, the encrypted data block is determined to be valid. The valid encrypted data block is decrypted through an edge device. The data structure is parsed using the AES algorithm to obtain the decrypted data.

[0112] Environmental and crop parameters are extracted from the decrypted data, sensor data is fused to generate a fused dataset, and a decision tree algorithm is used to determine irrigation and fertilization needs and real-time control parameters based on the fused dataset and the preset irrigation and fertilization plan.

[0113] Based on real-time control parameters, generate equipment control commands. If the parameters exceed the preset threshold, adjust the command priority to obtain a control command set.

[0114] The control instruction set is integrated by a collaborative scheduling algorithm to generate a collaborative control instruction set. Then, the execution sequence is extracted from the collaborative control instruction set and output to the agricultural equipment for execution.

[0115] Specifically, encrypted data blocks are obtained from trusted data storage, decrypted and analyzed by edge computing devices, and combined with real-time sensor data to generate a collaborative control instruction set. This process uses blockchain to ensure data security, leverages edge computing to achieve rapid response, and combines decision tree algorithms to accurately determine irrigation and fertilization needs, optimize control instruction generation, and ultimately achieve collaborative operation of irrigation and fertilization equipment, improve resource utilization efficiency, enhance system flexibility, and provide an efficient and intelligent control solution for the cultivation of Chinese medicinal herbs.

[0116] S6. Based on the collaborative control instruction set, the instructions are transmitted to the irrigation and fertilization equipment via a wireless sensor network. The control instructions and equipment status are input, and precise water and fertilizer application is executed to obtain water and fertilizer application records. Then, the water and fertilizer utilization efficiency is analyzed using a random forest algorithm. The application records, soil moisture and plant growth status are input to judge the water and fertilizer application effect and obtain the optimization adjustment parameters.

[0117] Further, in step S6, a water and fertilizer application record is obtained, specifically including:

[0118] The system collects environmental parameters and device status through a wireless sensor network, generates sensor data, and generates control commands based on collaborative control logic when the sensor data exceeds a preset threshold.

[0119] Control commands are sent to irrigation and fertilization equipment via network transmission. The command reception status is determined, and the operating data of the irrigation and fertilization equipment is obtained based on the command execution status to determine the degree of accuracy in application.

[0120] By comparing operational data with sensor data, the random forest algorithm is used to optimize the water and fertilizer management strategy and obtain the adjusted control commands.

[0121] Based on the adjusted control instructions, a new round of precise water and fertilizer application is executed, application records are generated, key parameters are extracted from the application records, the effectiveness of water and fertilizer management is judged, and the collaborative control logic is updated.

[0122] Furthermore, the optimized adjustment parameters are obtained, specifically including:

[0123] We acquire water and fertilizer application records, soil moisture data, and plant growth status, and integrate them into a structured dataset, which includes water application data, fertilizer application amount, soil moisture changes, and plant growth indicators, to obtain a preprocessed dataset.

[0124] Then, the random forest algorithm is used to train the preprocessed dataset to build a water and fertilizer use efficiency prediction model. The input data are water application amount, fertilizer application amount, soil moisture change and plant growth index, and the output is the predicted value of water and fertilizer use efficiency.

[0125] The predicted value of water and fertilizer use efficiency is compared with the preset water and fertilizer use efficiency threshold. The deviation between the water and fertilizer use efficiency and the preset threshold is calculated. When the deviation exceeds the preset threshold, it is marked as inefficient application, and the application effect judgment result is obtained.

[0126] Based on the results of the application effect judgment, records of inefficient application are extracted, and the distribution characteristics of water application data and fertilizer application amount are analyzed to obtain the abnormal application amount pattern. Then, the linear regression algorithm is used to fit the abnormal application amount pattern. Input water application data, fertilizer application amount and soil moisture changes, output the adjusted application amount suggestion, and obtain the preliminary optimization parameters.

[0127] Based on the preliminary optimization parameters, combined with plant growth indicators and soil moisture changes, a preset rule base is used for correction. When the matching degree between the optimized parameters and the growth indicators in the rule base is lower than the preset threshold, the parameter values ​​are adjusted to obtain the final optimized adjustment parameters.

[0128] By finally optimizing and adjusting the parameters, updating the water and fertilizer application records, generating a new application plan, and outputting it to the database, the optimized application scheme is obtained.

[0129] Specifically, in IoT-based standardized medicinal herb bases, acquiring water and fertilizer application records, soil moisture data, and plant growth status are key to achieving intelligent management. Water and fertilizer application records detail the amount of water used for each irrigation and the nitrogen, phosphorus, and potassium content of the fertilizer. For example, a record might show that a certain area in the base received 500 liters of water per mu (approximately 0.067 hectares) and 2 kg of nitrogen fertilizer per mu. Soil moisture data is precisely collected using embedded sensors, such as monitoring an increase in soil moisture from 30% to 45%. Plant growth status is obtained through drone image analysis, acquiring indicators such as chlorophyll content and plant height. For example, a batch of medicinal herbs might have a chlorophyll index of 0.7 and a plant height of 50 cm. This data is integrated into a package... The structured dataset, containing timestamps, water application volume, fertilizer type and dosage, humidity changes, and growth indicators, provides detailed evidence for subsequent analysis. The base uses the random forest algorithm to train a water and fertilizer use efficiency prediction model. This algorithm integrates multiple decision trees and uses water application volume, fertilizer application volume, soil moisture changes, and plant growth indicators as inputs to accurately predict water and fertilizer use efficiency. For example, with inputs of 500 liters of water, 2 kilograms of nitrogen fertilizer, a 15% humidity change, and a chlorophyll index of 0.7, the model outputs an efficiency value of 85%. This approach can fully integrate multidimensional features, effectively adapt to the complex nonlinear data characteristics in the cultivation of Chinese medicinal herbs, and significantly improve the stability and reliability of predictions.

[0130] In practical applications, the system calculates the deviation between the predicted efficiency and a preset threshold (e.g., 90%). If the efficiency is 85% and the deviation reaches 5%, exceeding the preset 3% threshold, the application record will be marked as inefficient. Database queries can quickly filter out inefficient records with water usage exceeding 400 liters / acre and efficiency below 90%. Analyzing the application distribution of these records reveals that water usage is concentrated between 450 and 550 liters / acre, and nitrogen fertilizer usage is excessive. This anomaly reflects over-application, which easily leads to resource waste. In this case, a linear regression algorithm is used to fit the abnormal pattern. Inputting water usage, fertilizer usage, and humidity changes, the system outputs adjustment suggestions. For example, for a situation with 550 liters / acre of water and 2.5 kg / acre of nitrogen fertilizer, it is recommended to adjust to 450 liters / acre and 1.8 kg / acre. By leveraging historical data trend analysis, quantitative parameters are provided for optimization, simplifying the adjustment process.

[0131] By combining plant growth indicators and soil moisture changes, parameters are further corrected and optimized. The base's rule library stipulates that when the chlorophyll index is below 0.6, the water volume needs to be increased by 10%. If the initial water volume adjustment is 450 liters / mu, but the actual chlorophyll index is 0.5, the water volume is adjusted to 495 liters / mu after correction. This ensures that the optimized parameters match the actual growth needs of the medicinal herbs, improving the accuracy of water and fertilizer application. Finally, the optimized parameters are updated to the application records, generating a new water and fertilizer application plan and storing it in the database. For example, the new plan specifies that the water volume for the next day is 480 liters / mu, and the nitrogen fertilizer is 1.7 kg / mu. The structured plan stored in the database facilitates automatic execution by irrigation and fertilization equipment and also facilitates subsequent traceability and management. Through data-driven continuous optimization of water and fertilizer application schemes, water and fertilizer utilization efficiency is effectively improved, resource waste is reduced, and strong support is provided for the precision planting of medicinal herbs in standardized bases.

[0132] S7. Based on the optimized adjustment parameters, the growth demand model parameters are updated using the gradient descent algorithm. The adjustment parameters and real-time feature set are input to optimize the model prediction accuracy and obtain the updated growth demand model.

[0133] Further, in step S7, the updated growth requirement model is obtained, specifically including:

[0134] The real-time feature set and adjustment parameters are acquired through the real-time acquisition system, stored as a standardized dataset, and the processed input data is obtained. Then, the data preprocessing module is used to normalize and denoise the input data to generate a feature set suitable for model training.

[0135] When the dimension of the feature set exceeds the preset threshold, the dimensionality is reduced by principal component analysis to obtain a compressed feature set. Based on the compressed feature set, the parameters of the growth requirement model are iteratively updated using the gradient descent algorithm, the loss function value is calculated, and a preliminary optimized model is obtained.

[0136] The prediction accuracy of the preliminary optimized model is evaluated by cross-validation, the accuracy evaluation index is calculated, and the model performance data is obtained. If the model performance data does not reach the preset threshold, the learning rate is adjusted and the gradient descent algorithm is re-executed to obtain the final optimized model. The updated growth requirement model parameters are output and stored as a model file to obtain the deployment preparation model.

[0137] Example 2

[0138] Please see Figure 4 This embodiment provides an IoT-based intelligent irrigation and fertilization decision-making system for standard medicinal herb production bases. The system implements IoT-based intelligent irrigation and fertilization methods for standard medicinal herb production bases, including:

[0139] The data acquisition module obtains multi-source heterogeneous data from soil moisture sensors, weather stations, and plant image acquisition devices, including soil moisture, nutrient content, temperature, rainfall, and plant growth status.

[0140] The data preprocessing and feature extraction module standardizes the collected data, extracts key features using principal component analysis, reduces data dimensionality, and optimizes the feature set.

[0141] The growth demand prediction module constructs and trains a growth demand model based on the optimized feature set, and uses a long short-term memory neural network to predict the water and fertilizer demand at each growth stage, generating an optimized water and fertilizer scheduling scheme.

[0142] The irrigation and fertilization plan generation module uses a linear programming algorithm to optimize irrigation and fertilization amounts when the predicted water and fertilizer demand exceeds a preset threshold. It combines soil moisture and weather conditions to generate a precise irrigation and fertilization plan.

[0143] The data storage and verification module uses blockchain technology to store irrigation and fertilization plans and sensor data, generates encrypted data blocks and stores them in a distributed ledger, and verifies data integrity through smart contracts to ensure data trustworthiness.

[0144] The real-time control and collaborative scheduling module retrieves encrypted data blocks from trusted data storage, decrypts and analyzes the data through edge computing devices, generates a collaborative control instruction set, and realizes collaborative operation of irrigation and fertilization equipment.

[0145] The efficiency assessment and optimization module transmits control commands to the irrigation and fertilization equipment through a wireless sensor network to execute precise water and fertilizer application. It uses a random forest algorithm to analyze water and fertilizer use efficiency and optimize the water and fertilizer application plan.

[0146] The model update module updates the growth requirement model parameters based on the optimized adjustment parameters and uses the gradient descent algorithm to optimize the model prediction accuracy and ensure the accuracy and reliability of the model.

[0147] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart irrigation and fertilization method for standardized medicinal herb production bases based on the Internet of Things, characterized by: Includes the following steps: Multi-source heterogeneous data, including soil moisture, nutrient content, temperature, rainfall and plant growth status, were obtained from soil moisture sensors, weather stations and plant image acquisition equipment. Principal component analysis was then used to extract key features of soil moisture, weather conditions and plant growth status to obtain a dimensionality-reduced feature set. The resulting dimensionality-reduced feature set includes: Standardized data on soil moisture, meteorological conditions, and plant growth were preprocessed using standardized methods to obtain a standardized dataset. The principal component analysis method is used to extract features from the standardized dataset to obtain a key feature set. When the principal component contribution rate of the key feature set is lower than the preset threshold, the parameters of the principal component analysis are adjusted, the key features are re-extracted, and an optimized feature set is obtained. Based on the optimized feature set, the K-means clustering algorithm was used to group soil moisture, meteorological conditions and plant growth status to obtain feature grouping results. Then, the correlation coefficient between soil moisture and plant growth status in each group was calculated to obtain the correlation analysis results. If the correlation coefficient in the correlation analysis results is lower than the preset threshold, the grouping results are re-clustered to obtain updated grouping results, and a dimensionality-reduced feature set is generated to determine the final feature set. A growth requirement model is established based on the feature set, and then a long short-term memory neural network is used to train the growth requirement model. The inputs are soil moisture, nutrient content, meteorological conditions and plant growth images, and the outputs are the water and fertilizer requirements for each growth stage, thus obtaining the growth requirement prediction results. The growth requirement prediction results include: Soil moisture, nutrient content, meteorological conditions and plant image data are acquired. Soil and meteorological information are collected through sensors, and plant images are captured using a high-resolution camera to obtain the input dataset. Then, image preprocessing techniques are used to denoise and segment the plant images to obtain the second image dataset. By standardizing soil and meteorological data, a normalized dataset is obtained. Based on the normalized dataset and the second image dataset, texture and color features of plant images are extracted. Combined with soil moisture and meteorological conditions, a multi-dimensional feature vector is constructed to obtain a comprehensive feature set. The growth requirement model is trained using a long short-term memory neural network. The comprehensive feature set is input, and if the training set loss is lower than a preset threshold, the training is stopped, and the trained model is obtained. By using the trained model and inputting the comprehensive feature set collected in real time, the water and fertilizer requirements at each growth stage are predicted, and the predicted requirements are obtained. When the water and fertilizer requirements in the growth demand forecast exceed the preset threshold, a linear programming algorithm is used to optimize irrigation and fertilization by inputting water and fertilizer requirements, soil moisture and weather conditions, and obtaining a precise irrigation and fertilization plan. Based on the precision irrigation and fertilization plan, blockchain technology is used to store the plan data and sensor data. The irrigation and fertilization amount, soil moisture and weather records are input to generate encrypted data blocks and obtain trusted data storage results. Specifically, obtaining trusted data storage results includes: By combining solution data records with sensor data through blockchain technology, encrypted data blocks are generated and stored in a distributed ledger to obtain trusted data storage results. When the storage timestamp of the encrypted data block is later than a preset threshold, data verification is triggered through a smart contract to determine the data integrity and obtain the verification result. Based on the verification results, the correlation between soil moisture and meteorological records is analyzed using a decision tree algorithm to determine the adjustment parameters for irrigation and fertilization. The adjustment parameters are then broadcast through the blockchain network to obtain consensus confirmation from distributed nodes, thus obtaining the final irrigation and fertilization plan. Key parameters are extracted from the final irrigation and fertilization plan, and control commands are generated using data mapping technology and output to the irrigation equipment to obtain the execution results. Encrypted data blocks are obtained from trusted data storage results, decrypted and analyzed using edge computing devices, and then real-time control commands are generated by inputting irrigation and fertilization plans and real-time sensor data to obtain a collaborative control command set. The obtained collaborative control instruction set specifically includes: The encrypted data block is obtained from trusted storage. The integrity of the data is verified using a preset key. If the integrity is verified, the encrypted data block is determined to be valid. The valid encrypted data block is decrypted through an edge device. The data structure is parsed using the AES algorithm to obtain the decrypted data. Environmental and crop parameters are extracted from the decrypted data, sensor data is fused to generate a fused dataset, and a decision tree algorithm is used to determine irrigation and fertilization needs and real-time control parameters based on the fused dataset and the preset irrigation and fertilization plan. Based on real-time control parameters, generate equipment control commands. If the parameters exceed the preset threshold, adjust the command priority to obtain a control command set. The control instruction set is integrated by a collaborative scheduling algorithm to generate a collaborative control instruction set. Then, the execution sequence is extracted from the collaborative control instruction set and output to the agricultural equipment for execution. Based on the collaborative control instruction set, the instructions are transmitted to the irrigation and fertilization equipment via a wireless sensor network. The control instructions and equipment status are input to execute precise water and fertilizer application, obtain water and fertilizer application records, and then the water and fertilizer use efficiency is analyzed using a random forest algorithm. The application records, soil moisture and plant growth status are input to judge the effect of water and fertilizer application and obtain optimization adjustment parameters. Based on the optimized adjustment parameters, the gradient descent algorithm is used to update the growth demand model parameters. The adjustment parameters and real-time feature set are input to optimize the model prediction accuracy and obtain the updated growth demand model.

2. The intelligent irrigation and fertilization method for standard medicinal herb bases based on the Internet of Things as described in claim 1, characterized in that: Obtain a precise irrigation and fertilization plan, specifically including: When the water and fertilizer demand exceeds the preset threshold, the water and fertilizer demand, soil moisture and meteorological conditions are obtained from the prediction results to determine the input dataset. Then, the input dataset is processed by the linear programming algorithm to optimize the irrigation and fertilizer application and obtain a preliminary irrigation and fertilization plan. Based on the preliminary irrigation and fertilization plan, combined with soil moisture and meteorological conditions, it is determined whether the plan meets the constraints of precision irrigation and fertilization, and an adjusted irrigation and fertilization plan is obtained. If the adjusted irrigation and fertilization plan meets the constraints, a time series analysis algorithm is used to predict the future trend of water and fertilizer demand and obtain demand forecast data. Based on demand forecast data, combined with current soil moisture and weather conditions, the input parameters of the linear programming algorithm are updated to obtain an optimized precision irrigation and fertilization plan. Based on the optimized precision irrigation and fertilization plan, real-time control commands are generated and output to the irrigation and fertilization equipment to determine the execution plan. After the execution plan is running, real-time data on soil moisture and weather conditions are collected to determine whether the water and fertilizer requirements need to be adjusted, and an updated input dataset is obtained.

3. The intelligent irrigation and fertilization method for standard medicinal herb bases based on the Internet of Things as described in claim 1, characterized in that: Obtain water and fertilizer application records, specifically including: The system collects environmental parameters and device status through a wireless sensor network, generates sensor data, and generates control commands based on collaborative control logic when the sensor data exceeds a preset threshold. Control commands are sent to irrigation and fertilization equipment via network transmission. The command reception status is determined, and the operating data of the irrigation and fertilization equipment is obtained based on the command execution status to determine the degree of accuracy in application. By comparing operational data with sensor data, the random forest algorithm is used to optimize the water and fertilizer management strategy and obtain the adjusted control commands. Based on the adjusted control instructions, a new round of precise water and fertilizer application is executed, application records are generated, key parameters are extracted from the application records, the effectiveness of water and fertilizer management is judged, and the collaborative control logic is updated.

4. The intelligent irrigation and fertilization method for standard medicinal herb bases based on the Internet of Things as described in claim 1, characterized in that: The optimized adjustment parameters include: We acquire water and fertilizer application records, soil moisture data, and plant growth status, and integrate them into a structured dataset, which includes water application data, fertilizer application amount, soil moisture changes, and plant growth indicators, to obtain a preprocessed dataset. Then, the random forest algorithm is used to train the preprocessed dataset to build a water and fertilizer use efficiency prediction model. The input data are water application amount, fertilizer application amount, soil moisture change and plant growth index, and the output is the predicted value of water and fertilizer use efficiency. The predicted value of water and fertilizer use efficiency is compared with the preset water and fertilizer use efficiency threshold. The deviation between the water and fertilizer use efficiency and the preset threshold is calculated. When the deviation exceeds the preset threshold, it is marked as inefficient application, and the application effect judgment result is obtained. Based on the results of the application effect judgment, records of inefficient application are extracted, and the distribution characteristics of water application data and fertilizer application amount are analyzed to obtain the abnormal application amount pattern. Then, the linear regression algorithm is used to fit the abnormal application amount pattern. Input water application data, fertilizer application amount and soil moisture changes, output the adjusted application amount suggestion, and obtain the preliminary optimization parameters. Based on the preliminary optimization parameters, combined with plant growth indicators and soil moisture changes, a preset rule base is used for correction. When the matching degree between the optimized parameters and the growth indicators in the rule base is lower than the preset threshold, the parameter values ​​are adjusted to obtain the final optimized adjustment parameters. By finally optimizing and adjusting the parameters, updating the water and fertilizer application records, generating a new application plan, and outputting it to the database, the optimized application scheme is obtained.

5. The intelligent irrigation and fertilization method for standard medicinal herb bases based on the Internet of Things as described in claim 1, characterized in that: The updated growth requirement model is obtained, specifically including: The real-time feature set and adjustment parameters are acquired through the real-time acquisition system, stored as a standardized dataset, and the processed input data is obtained. Then, the data preprocessing module is used to normalize and denoise the input data to generate a feature set suitable for model training. When the dimension of the feature set exceeds the preset threshold, the dimensionality is reduced by principal component analysis to obtain a compressed feature set. Based on the compressed feature set, the parameters of the growth requirement model are iteratively updated using the gradient descent algorithm, the loss function value is calculated, and a preliminary optimized model is obtained. The prediction accuracy of the initial optimized model is evaluated by cross-validation, and the accuracy evaluation index is calculated to obtain model performance data. If the model performance data does not reach the preset threshold, the learning rate is adjusted and the gradient descent algorithm is re-executed to obtain the final optimized model and output the updated growth requirement model parameters.

6. An IoT-based intelligent irrigation and fertilization decision-making system for standard medicinal herb production bases, used to implement the IoT-based intelligent irrigation and fertilization method for standard medicinal herb production bases as described in any one of claims 1-5, characterized in that: include: The data acquisition module obtains multi-source heterogeneous data from soil moisture sensors, weather stations, and plant image acquisition devices, including soil moisture, nutrient content, temperature, rainfall, and plant growth status. The preprocessing and feature extraction module standardizes the collected data, extracts key features using principal component analysis, reduces data dimensionality, and optimizes the feature set. The growth demand prediction module constructs and trains a growth demand model based on the optimized feature set, and uses a long short-term memory neural network to predict the water and fertilizer demand at each growth stage, generating an optimized water and fertilizer scheduling scheme. The irrigation and fertilization plan generation module uses a linear programming algorithm to optimize irrigation and fertilization amounts when the predicted water and fertilizer demand exceeds a preset threshold. It combines soil moisture and weather conditions to generate a precise irrigation and fertilization plan. The data storage and verification module uses blockchain technology to store irrigation and fertilization plans and sensor data, generates encrypted data blocks and stores them in a distributed ledger, and verifies data integrity through smart contracts. The real-time control and collaborative scheduling module retrieves encrypted data blocks from trusted data storage, decrypts and analyzes the data through edge computing devices, generates a collaborative control instruction set, and performs collaborative operations of irrigation and fertilization equipment. The efficiency assessment and optimization module transmits control commands to the irrigation and fertilization equipment through a wireless sensor network to execute precise water and fertilizer application. It uses a random forest algorithm to analyze water and fertilizer use efficiency and optimize the water and fertilizer application plan. The model update module updates the growth requirement model parameters based on the optimized adjustment parameters and uses the gradient descent algorithm to optimize the model prediction accuracy.

Images