Intelligent biogas slurry application method, system and storage medium based on machine learning
Through the intelligent worm fluid application method based on machine learning, the state classification model and application measure prediction model are used to adjust the worm fluid application strategy in real time, solving the problem of inaccurate worm fluid application in the existing technology, and improving irrigation efficiency and soil health.
Patent Information
- Application Number
- CN202110986629.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-08-26
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2041-08-26
Smart Images

Figure CN113642729B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an intelligent biogas slurry application method, system and storage medium based on machine learning, and belongs to the technical field of agricultural fertilization and machine learning. Background Art
[0002] Biogas slurry contains a variety of nutrients for plant growth, rich amino acids, various growth hormones and vitamins, etc. It is a fast-acting organic fertilizer. Returning biogas slurry to fields has the advantages of enhancing soil fertilizer efficiency, promoting crop yields and income, and inhibiting plant diseases.
[0003] The agricultural industry needs to accelerate the treatment and resource utilization of livestock and poultry breeding waste; comprehensively promote the resource utilization of livestock and poultry breeding waste, build a new pattern of sustainable development that combines farming and breeding and agricultural and animal husbandry cycles, and encourage the scientific return of biogas slurry and harmlessly treated livestock and poultry breeding wastewater as fertilizers. The return of biogas slurry to the fields meets the needs of national industrial development and is conducive to promoting green and sustainable agricultural development.
[0004] At present, the nutrient content of biogas slurry varies significantly, and the nutrient content varies greatly between different biogas projects. The nutrient content of the same biogas project is also affected by many factors such as fermentation raw materials, biogas project type and operation time. At present, it is difficult to achieve real-time detection in farmland utilization, and there is a lack of application amount basis for returning to the field. There are often problems such as insufficient or excessive application of biogas slurry, which affects crop yields, destroys the acid-base balance of the soil, increases the content of heavy metals in the soil, and even causes surface pollution problems;
[0005] At present, there is very little research and development content on biogas slurry application equipment, and there is even less research on methods for scientific biogas slurry application; the patent number "CN108156886A" discloses a biogas slurry water and fertilizer supply device, which belongs to the field of agricultural organic waste resource and energy utilization. It solves the problems of high cost and difficulty in control in the process of mixing and irrigating biogas slurry with water in the prior art. The device mainly includes: a liquid storage tank located underground and a closed mixing and stirring tank on the ground; the liquid storage tank and the mixing and stirring tank are connected by a biogas slurry pump; the mixing and stirring tank is also connected to a water pump and a liquid outlet, and the liquid outlet is connected to the first fertilizer outlet pipe and the second fertilizer outlet pipe through a three-way joint, and each pipeline is provided with a solenoid valve; the mixing and stirring tank is also provided with a display screen, an operation button, a battery, a processor, a solar panel and a signal receiver, and a stirring motor is arranged in the middle of the top of the mixing and stirring tank, and the stirring motor is connected to a stirring shaft, and a cross stirring paddle and a stirring blade are respectively connected to the stirring shaft, and a liquid level sensor is arranged in the mixing and stirring tank. The device realizes water-fertilizer mixing and timed and quantitative irrigation, with strong controllability and easy use. It can be seen that the traditional water-fertilizer irrigation method is an intelligent irrigation method with timed and fixed parameters, which cannot be dynamically adjusted according to the actual application scenario to achieve precise fertilization, and some system equipment is complicated to operate, which is difficult for farmers to accept. Therefore, it is necessary to combine multidisciplinary mature technologies to realize dynamic adjustment of biogas slurry application, improve the quality and efficiency of biogas slurry irrigation, and meet market demand. Summary of the invention
[0006] In order to solve the problems existing in the above-mentioned prior art, the present invention proposes an intelligent biogas slurry application method, system and storage medium based on machine learning. By training the biogas slurry application model through machine learning, the biogas slurry application strategy can be dynamically adjusted according to the actual changes of farmland, making the biogas slurry application intelligent.
[0007] The technical solution of the present invention is as follows:
[0008] An intelligent biogas slurry application method based on machine learning comprises the following steps:
[0009] Deploy a decision factor collection unit and a biogas slurry application unit on the target farmland, wherein the decision factor collection unit is used to collect decision factors, wherein the decision factors include at least one data related to crop growth, and the biogas slurry application unit is used to irrigate the target farmland according to a specified biogas slurry application strategy;
[0010] Collect historical data, and obtain data related to crop growth and irrigation data in different unit time periods based on the historical data; establish a biogas slurry application model, which includes a state classification model and an application measure prediction model based on machine learning; obtain characteristic indicators and corresponding characteristic values or target characteristic indicators and corresponding characteristic values calculated by characteristic indicators from historical data as training samples, and train the state classification model and application measure prediction model respectively; the state classification model is used to predict the farmland state type according to the decision factors, and the application measure prediction model is used to predict the biogas slurry application measures and application characteristic values to be adopted according to the decision factors and the farmland state type;
[0011] The decision factors are collected by the decision factor collection unit, the farmland state type is determined by the state classification model, the decision factors and the farmland state type are input into the application measure prediction model, and a biogas slurry application strategy consisting of the biogas slurry application measures and application characteristic values to be adopted is output; the biogas slurry application unit is controlled according to the biogas slurry application strategy to irrigate the target farmland.
[0012] Preferably, the decision factors include at least soil data and environmental data; the soil data include soil moisture, soil pH value, and soil nutrient content, and the environmental data include meteorological data and environmental temperature and humidity.
[0013] Preferably, the training method of the state classification model is specifically as follows:
[0014] Acquire soil data and environmental data in different unit time periods through historical data, classify farmland status types in different unit time periods according to the soil data and environmental data in the unit time periods, and obtain farmland status classification groups composed of several farmland status types;
[0015] The characteristic index is calculated by soil data and environmental data in a unit time period, and the characteristic index is used as the input of the neural network. A certain farmland state type in the farmland state classification group is used as the output to train the neural network and obtain a trained state classification model.
[0016] Preferably, the characteristic index includes an absorbable moisture characteristic index calculated through soil moisture, meteorological data and ambient temperature and humidity, and an absorbable nutrient characteristic index calculated through soil pH value and soil nutrient content.
[0017] Preferably, the application measure prediction model includes a measure selection model and a measure assignment model, and the specific method for training the application measure prediction model is:
[0018] Building an application measure library, wherein the application measure library contains all biogas slurry application measures extracted from historical irrigation data;
[0019] Taking the characteristic index and the farmland state type in a unit time period as the input of the neural network, selecting at least one biogas slurry application measure from the application measure library as the output of the neural network;
[0020] Constructing a calculation formula for the selection hit rate, wherein the selection hit rate refers to the probability that the application measures output by the neural network completely match the application measures adopted within the unit time period;
[0021] A loss function is constructed based on the selected hit rate, and the parameters of the neural network are iteratively updated by the gradient descent method to obtain an application measure prediction model;
[0022] The method for training the measure assignment model is specifically as follows:
[0023] The characteristic indicators, farmland status types, biogas slurry application measures and corresponding application characteristic values within a unit time period are combined into a training sample and put into the experience playback library;
[0024] The characteristic indicators, farmland status type and sludge application measures of each training sample are taken as the input of the neural network, and the predicted value corresponding to the sludge application measures is estimated. The loss function is constructed based on the difference between the corresponding predicted value and the application characteristic value. The gradient descent method is used to iteratively update the parameters of the neural network to obtain the measure assignment model.
[0025] Preferably, the biogas slurry application measures include biogas slurry application time nodes, adjustment of the biogas slurry and clean water ratio, adjustment of the biogas slurry flow rate and biogas slurry application duration, and the application characteristic value is the numerical value of the corresponding biogas slurry application measure.
[0026] Preferably, the target farmland is further provided with an intelligent terminal, the biogas slurry application model is established on a cloud server, and the intelligent terminal is communicatively connected with the decision factor collection unit, the biogas slurry application unit and the cloud server, for receiving the decision factors collected by the decision factor collection unit, and sending them to the cloud server, and receiving the biogas slurry application strategy output by the cloud server through the biogas slurry application model and the decision factors, and controlling the biogas slurry application unit to irrigate the target farmland according to the biogas slurry application strategy.
[0027] Preferably, the decision factor acquisition unit includes a field monitoring module and a meteorological data acquisition module, and the biogas slurry application unit includes a water-fertilizer adjustment module and a water-fertilizer transportation module.
[0028] The present invention also provides an intelligent biogas slurry application system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the biogas slurry application method as described in any embodiment of the present invention is implemented.
[0029] The present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the biogas slurry application method as described in any embodiment of the present invention.
[0030] The present invention has the following beneficial effects:
[0031] The present invention provides an intelligent biogas slurry application method based on machine learning, establishes a biogas slurry application model based on machine learning, can adjust the biogas slurry application strategy in real time according to decision factors, and automatically applies the biogas slurry according to the adjusted biogas slurry application strategy, so as to make the biogas slurry application intelligent. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION
[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0034] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.
[0035] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0036] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.
[0037] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.
[0038] See also Figure 1 , an intelligent biogas slurry application method based on machine learning, comprising the following steps:
[0039] Deploy a decision factor collection unit and a biogas slurry application unit on the target farmland, wherein the decision factor collection unit is used to collect decision factors, wherein the decision factors include at least one data related to crop growth, and the biogas slurry application unit is used to irrigate the target farmland according to a specified biogas slurry application strategy;
[0040] Collect historical data from planting to harvesting of multiple crops as samples, which can be historical data of the target farmland or historical data of other farmlands. For other farmlands, at least the geographical environment and the crops planted are basically consistent with those of the target farmland; obtain data related to crop growth and irrigation data in different unit time periods according to the historical data. In this embodiment, the unit time period refers to one day, that is, in this embodiment, the data related to crop growth and irrigation data for each day from planting to harvesting of crops are obtained according to the historical data; in the historical data, the data related to crop growth in different unit time periods are objective data, and the irrigation data are empirical data, which are different biogas slurry application measures and corresponding characteristic values of farmers for biogas slurry application on farmland according to specific conditions. For example, the biogas slurry application measures adjust the mixing ratio of biogas slurry and clean water, and the characteristic value is the specific value of the mixing ratio;
[0041] A biogas slurry application model is established, the biogas slurry application model includes a state classification model and an application measure prediction model based on machine learning; characteristic indicators and corresponding characteristic values are obtained from historical data or target characteristic indicators and corresponding characteristic values calculated by characteristic indicators are used as training samples, and the state classification model and the application measure prediction model are trained respectively; the state classification model is used to predict the farmland state type according to the decision factors, and the farmland state type is classified by experienced farmers and can be divided into N types, such as farmland drought, farmland with sufficient water, farmland with sufficient nutrition, etc., and the farmland state type label is artificially added to each training sample, and then the state classification model is trained; the application measure prediction model is used to predict the biogas slurry application measures and application characteristic values to be adopted according to the decision factors and the farmland state type;
[0042] The decision factors of the target farmland in the current time period are collected in real time through the decision factor collection unit, the farmland state type of the current target farmland is determined through the state classification model, the decision factors in the current time period and the farmland state type of the current target farmland are input into the application measure prediction model, and a biogas slurry application strategy consisting of the biogas slurry application measures and application characteristic values to be adopted is output; the biogas slurry application unit is controlled according to the biogas slurry application strategy to irrigate the target farmland.
[0043] In another embodiment of the present invention, the collected decision factors include at least soil data and environmental data, the soil data includes soil moisture, soil pH value, soil nutrient content, the soil nutrient content is specifically the content of nitrogen, phosphorus and potassium, and the environmental data includes meteorological data and ambient temperature and humidity.
[0044] In another embodiment of the present invention, the training method of the state classification model is specifically:
[0045] Soil data and environmental data in different time periods are obtained through historical data. The farmland status types in different time periods are classified according to the soil data and environmental data in different time periods. The classification is done manually. Farmers judge the farmland status type of the day based on the soil data and environmental data of each day according to their experience and label them. The farmland status types summarized manually form farmland status classification groups. ;
[0046] The characteristic index is calculated by soil data and environmental data in a unit time period, and the characteristic index is used as the input of the neural network. As the prediction type output, a loss function is constructed based on the output farmland state type and the actual farmland state type (that is, label) within the unit time period: , train the neural network, iteratively update the parameters of the neural network through the gradient descent method, and obtain the trained state classification model.
[0047] In another embodiment of the present invention, the characteristic index includes an absorbable moisture characteristic index α calculated by soil moisture, meteorological data and ambient temperature and humidity, and an absorbable nutrient characteristic index β calculated by soil pH value and soil nutrient content; in this embodiment, the absorbable moisture characteristic index α=k1*a1+k2*a2+k3*a3, wherein a1, a2, a3 are soil moisture, expected precipitation obtained by meteorological data, and air humidity obtained by ambient temperature and humidity, respectively, k1, k2 and k3 are coefficients, k1 and k3 are assigned by artificial experience, k2 is determined by the expected weather in the meteorological data, and k2 is assigned different values according to different weather; in this embodiment, the absorbable nutrient characteristic index β=h1*b1+h2*b2, wherein b1 and b2 are soil pH value and soil nutrient content, respectively, and h1 and h2 are coefficients assigned by artificial experience.
[0048] In another embodiment of the present invention, the application measure prediction model includes a measure selection model and a measure assignment model, and the specific method of training the application measure prediction model is:
[0049] Building an application measure library, wherein the application measure library contains all biogas slurry application measures extracted from historical irrigation data;
[0050] Taking the characteristic index and the farmland state type in a unit time period as the input of the neural network, selecting at least one biogas slurry application measure from the application measure library as the output of the neural network;
[0051] Constructing the selection hit rate calculation formula , which means that when the neural network outputs the set of biogas slurry application measures When the biogas slurry application measures adopted in the unit time period can be completely matched, the value of the indicator function is 1, otherwise it is 0; the selection hit rate refers to the probability that the application measures output by the neural network completely match the application measures adopted in the unit time period;
[0052] A loss function is constructed based on the selected hit rate, and the parameters of the neural network are iteratively updated by the gradient descent method to obtain an application measure prediction model;
[0053] The method for training the measure assignment model is specifically as follows:
[0054] The characteristic indicators α and β within a unit time period, the farmland status type , take measures to apply biogas slurry and the corresponding application characteristic value Merge into one training sample Put it into the experience replay library;
[0055] The characteristic index, farmland status type and biogas slurry application measures of each training sample are used as the input of the neural network to estimate the prediction value corresponding to the biogas slurry application measures. , the loss function is constructed based on the difference between the corresponding predicted value and the applied feature value , the gradient descent method is used to iteratively update the parameters of the neural network to obtain the measure assignment model.
[0056] In another embodiment of the present invention, the biogas slurry application measures include biogas slurry application time nodes, adjustment of the biogas slurry and clean water ratio, adjustment of the biogas slurry flow rate and biogas slurry application duration, and the application characteristic value is the numerical value of the corresponding biogas slurry application measure. For the biogas slurry application time node, the application characteristic is a time node such as 12 o'clock and 17:30 o'clock. For ease of use, the application characteristic value converts the time node into a rational number, such as 12 and 17.5; the application characteristic value for adjusting the biogas slurry and clean water ratio is specifically the EC value after the biogas slurry and clean water are mixed; the application characteristic value for adjusting the biogas slurry flow rate is the adjusted biogas slurry flow value; the application characteristic value for the biogas slurry application duration is the interval from the start of biogas slurry application to the end of biogas slurry application, which also needs to be converted into a rational number.
[0057] In another embodiment of the present invention, a smart terminal is further provided in the target farmland, and the smart terminal may be a PC, a workstation or other terminals capable of information transmission and processing. The biogas slurry application model is established on a cloud server, and the smart terminal is communicatively connected with a decision factor acquisition unit, a biogas slurry application unit and a cloud server, and is configured to receive decision factors collected by the decision factor acquisition unit, and upload the decision factors to the cloud server, and receive a biogas slurry application strategy output by the cloud server through the biogas slurry application model and the decision factors, and control the biogas slurry application unit to irrigate the target farmland according to the biogas slurry application strategy.
[0058] In another embodiment of the present invention, the decision factor acquisition unit includes a field monitoring module and a meteorological data acquisition module, and the biogas slurry application unit includes a water-fertilizer allocation module and a water-fertilizer delivery module. The field monitoring module includes a number of sensors, including at least a temperature and humidity sensor, a soil moisture sensor, a pH sensor, and an electrochemical sensor, which are used to collect soil moisture, soil pH value, soil nutrient content, and ambient temperature and humidity respectively; the meteorological data acquisition module is used to receive weather forecasts through the Internet to obtain meteorological data, and the meteorological data includes expected weather, expected precipitation, etc.; the water-fertilizer allocation includes a biogas slurry tank, a clean water tank, a biogas slurry electric regulating valve, a clean water electric regulating valve, a centrifugal pump, etc., which can adjust the fertilization time and application flow rate of the water and biogas slurry mixture according to the set biogas slurry EC value; the water-fertilizer delivery module includes a solenoid valve and a field pipeline, which is used to transport the biogas slurry to various parts of the farmland.
[0059] The present invention also provides an intelligent biogas slurry application system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the biogas slurry application method as described in any embodiment of the present invention is implemented.
[0060] The present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the biogas slurry application method as described in any embodiment of the present invention.
[0061] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. An intelligent biogas slurry application method based on machine learning, characterized in that: The following steps are involved: Deploy a decision factor collection unit and a biogas slurry application unit on the target farmland, wherein the decision factor collection unit is used to collect decision factors, wherein the decision factors include at least one data related to crop growth, and the biogas slurry application unit is used to irrigate the target farmland according to a specified biogas slurry application strategy; Collect historical data, and obtain data related to crop growth and irrigation data in different unit time periods based on the historical data; establish a biogas slurry application model, which includes a state classification model and an application measure prediction model based on machine learning; obtain characteristic indicators and corresponding characteristic values or target characteristic indicators and corresponding characteristic values calculated by characteristic indicators from historical data as training samples, and train the state classification model and application measure prediction model respectively; the state classification model is used to predict the farmland state type according to the decision factors, and the application measure prediction model is used to predict the biogas slurry application measures and application characteristic values to be adopted according to the decision factors and the farmland state type; The decision factor collection unit collects decision factors, determines the farmland state type through the state classification model, inputs the decision factors and the farmland state type into the application measure prediction model, and outputs a biogas slurry application strategy consisting of biogas slurry application measures and application characteristic values to be adopted; controls the biogas slurry application unit according to the biogas slurry application strategy to irrigate the target farmland; The decision factors include at least soil data and environmental data; the soil data include soil moisture, soil pH value, and soil nutrient content, and the environmental data include meteorological data and environmental temperature and humidity; The training method of the state classification model is specifically as follows: Acquire soil data and environmental data in different unit time periods through historical data, classify farmland status types in different unit time periods according to the soil data and environmental data in the unit time periods, and obtain farmland status classification groups composed of several farmland status types; The characteristic index is calculated through the soil data and environmental data in a unit time period, and the characteristic index is used as the input of the neural network. A certain farmland state type in the farmland state classification group is used as the output to train the neural network and obtain a trained state classification model; The characteristic index includes an absorbable moisture characteristic index calculated by soil moisture, meteorological data and ambient temperature and humidity, and an absorbable nutrient characteristic index calculated by soil pH value and soil nutrient content; The application measure prediction model includes a measure selection model and a measure assignment model. The specific method for training the application measure selection model is: Building an application measure library, wherein the application measure library contains all biogas slurry application measures extracted from historical irrigation data; Taking the characteristic index and the farmland state type in a unit time period as the input of the neural network, selecting at least one biogas slurry application measure from the application measure library as the output of the neural network; Constructing a calculation formula for the selection hit rate, wherein the selection hit rate refers to the probability that the application measures output by the neural network completely match the application measures adopted within the unit time period; A loss function is constructed based on the selected hit rate, and the parameters of the neural network are iteratively updated by the gradient descent method to obtain an application measure selection model; The method for training the measure assignment model is specifically as follows: The characteristic indicators, farmland status types, biogas slurry application measures and corresponding application characteristic values within a unit time period are combined into a training sample and put into the experience playback library; The characteristic indicators, farmland status type and sludge application measures of each training sample are taken as the input of the neural network, and the predicted value corresponding to the sludge application measures is estimated. The loss function is constructed based on the difference between the corresponding predicted value and the application characteristic value. The gradient descent method is used to iteratively update the parameters of the neural network to obtain the measure assignment model.
2. The intelligent biogas slurry application method based on machine learning according to claim 1, characterized in that: The biogas slurry application measures include biogas slurry application time nodes, adjustment of biogas slurry and clean water ratio, adjustment of biogas slurry flow rate and biogas slurry application duration, and the application characteristic value is the value of the corresponding biogas slurry application measure.
3. The intelligent biogas slurry application method based on machine learning according to claim 1, characterized in that: An intelligent terminal is also provided in the target farmland, and the biogas slurry application model is established on a cloud server. The intelligent terminal is communicatively connected with the decision factor collection unit, the biogas slurry application unit and the cloud server, and is used for receiving the decision factors collected by the decision factor collection unit, and sending them to the cloud server, and receiving the biogas slurry application strategy output by the cloud server through the biogas slurry application model and the decision factors, and controlling the biogas slurry application unit to irrigate the target farmland according to the biogas slurry application strategy.
4. The intelligent biogas slurry application method based on machine learning according to claim 3 is characterized in that: The decision factor collection unit includes a field monitoring module and a meteorological data collection module, and the biogas slurry application unit includes a water and fertilizer allocation module and a water and fertilizer transportation module.
5. An intelligent biogas slurry application system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the biogas slurry application method according to any one of claims 1 to 4 is implemented.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the biogas slurry application method according to any one of claims 1 to 4 is implemented.
Citation Information
Patent Citations
Biogas slurry water fertilizer supply device
CN108156886A
Soil moisture status prediction method and system based on DBN (Deep Belief Network) model
CN105825230A
System and method for integrally and intelligently controlling water and fertilizer in field based on multi-source information fusion
CN106707767A
Agriculture drought monitoring, prewarning and forecasting method
CN107451522A
Crop growth irrigation decision-making control system
CN107945042A
