CN114365614A

CN114365614A - Water and fertilizer accurate regulation and control method, intelligent equipment and system based on Internet of things

Info

Publication number: CN114365614A
Application number: CN202111382762.9A
Authority: CN
Inventors: 王永涛; 刘坚; 黄维; 李蓉; 杨文峰; 索鑫宇; 胡芮家
Original assignee: Guizhou Water Conservancy Research Institute Guizhou Irrigation Test Center Station; Hunan University
Current assignee: Guizhou Water Conservancy Research Institute Guizhou Irrigation Test Center Station; Hunan University
Priority date: 2021-11-22
Filing date: 2021-11-22
Publication date: 2022-04-19

Abstract

The invention is suitable for the technical field of agricultural irrigation and fertilization management, and provides a water and fertilizer accurate regulation and control method, intelligent equipment and a system based on the Internet of things, wherein a four-layer architecture mode of a sensing layer, a transmission layer (an access layer and a network layer), a supporting layer and an application layer of the Internet of things is adopted; independently developing a multifunctional acquisition control device based on STM32F103RET6 and optimizing the structures of a jet and a Venturi mixer; the intelligent water and fertilizer integrated networking is realized by combining a short-distance communication protocol with various protocols such as Lora, TCP/IP and the like; the transfer function with a buffer fertilizer mixing barrel is designed based on the BPNN data fusion of a plurality of sensors such as nutrient solution EC, nutrient solution PH, soil temperature T, soil humidity H and the like, and is integrated with a two-dimensional incremental fuzzy PID control method, so that the response speed and the accuracy of the soil moisture and nutrient control process are realized; and predicting the future 7-15 day rainfall by using an LSTM long-short term memory neural network, and evaluating the drought conditions in continuous rainy-free days and main dry land crops.

Description

Water and fertilizer accurate regulation and control method, intelligent equipment and system based on Internet of things

Technical Field

The invention belongs to the technical field of agricultural irrigation and fertilization management, and particularly relates to a water and fertilizer accurate regulation and control method, intelligent equipment and system based on the Internet of things.

Background

With the rapid development of network technology and information technology, the agricultural informatization construction is given a rich connotation, the optimal conditions for crop growth are measured and obtained by adopting a wireless network on the basis of the Internet of things, and accurate regulation and control are performed, so that the aims of increasing yield, improving quality, adjusting growth period, saving water and fertilizer, reducing environmental pollution and improving economic benefit are fulfilled. The wireless sensor network and some automatic and intelligent devices are arranged in the farmland, so that the farmland information can be remotely monitored and controlled in real time, the scientific and reasonable growth of crops is guaranteed, the consumption of various human resources is greatly reduced, the traditional agricultural production mode is changed, and the agricultural development speed is increased. The application of the agricultural internet of things technology in the agricultural field gradually becomes a mark of the modernization level of the agriculture in a new era.

Finding the data to know that the range of the conductivity r (t) of the crops is generally between 0.8 and 2.5mS/cm, and the range of the deviation e of the conductivity r (t) should not exceed +/-0.8 mS/cm; most crops are suitable for growing in neutral or weakly acidic and weakly alkaline soil, the pH value of the growth of the crops is 6.5-7.5, and the excessive acidity or alkalinity of the soil is not beneficial to the growth of the crops. High efficiency and immediacy of nutrient solution preparation are required.

In the aspect of a water and fertilizer precise regulation method, a PID classical control and fuzzy control are combined to realize a dynamic regulation system for the EC value and the PH value of soil, the performances in the aspects of response speed and control precision are poor, and the requirement of rapid and accurate realization of water and fertilizer integration in agricultural production is restricted to a certain extent.

The rainfall is a weakly-related and highly-complex nonlinear power system due to the comprehensive action of various physical factors such as atmospheric circulation, hydrological meteorological factors, natural geography and the like, changes of the system do not move in a fixed period, but include changes and local fluctuations of various time scales, and the characteristics cause great difficulty and low precision in forecasting the rainfall in medium and long periods.

The integration of water and fertilizer is a high and new technology for tightly combining irrigation and fertilization, and irrigation equipment, water-soluble fertilizer and an irrigation and fertilization system are required to be organically combined. However, the prior art and products are not closely combined, only irrigation and fertilization equipment is required in some places, and the optimization of an irrigation and fertilization system and the application of water-soluble fertilizers are ignored, so that the economic and social benefits of the technology are not fully exerted.

Meanwhile, the water and fertilizer integrated equipment is complex to operate, low in intelligent degree, less in technical research and development aiming at specific conditions and crops, and not strong in applicability of the equipment and products, particularly, the fertilization set of products are few in variety and specification, single in form, and insufficient in development of large-scale, precise and intelligent equipment, so that popularization and application of water and fertilizer integration are restricted.

The water and fertilizer integrated system is not suitable for local conditions. The water and fertilizer integrated fertilizer applicator is mostly imported integrally, a system communication protocol is not open, compatibility is poor, remote control is difficult to realize, personalized customization and secondary development cannot be carried out according to user requirements, and adaptability is poor. The hydrology and geography conditions, planting conditions, irrigation habits and the like all over the country are different, and a water and fertilizer integrated system which is successfully popularized in one place can be out of order in the other place.

Disclosure of Invention

The invention provides a water and fertilizer precise regulation and control method, intelligent equipment and a system based on the Internet of things, and the integration of the Internet of things technology, automatic fertigation equipment and an intelligent control method is realized. Aiming at the moisture control threshold value of the dragon fruit in the growth period of the dragon fruit, the kiwi fruit and the tea in Guizhou, the soil environment and the nutrient content, a two-dimensional incremental fuzzy PID control method is adopted, the irrigation quantity and the fertilization quantity are accurately controlled through the accurate control of the irrigation and the fertilization, the system realizes the intellectualization of the detection, allocation and supply of water and fertilizer, and the system has the characteristics of strong practicability and easy management.

The invention is realized in this way, and the accurate regulation and control system of liquid manure based on thing networking, including information perception layer, information transmission layer, supporting layer and information application layer, wherein:

the information perception layer comprises a bar code, a two-dimensional code, an RFID, an STM32F103RET 6-based independent research and development multifunctional acquisition control device, an intelligent sensor, an intelligent camera, hypertext information and a self-organizing network;

the information transmission layer comprises an information access layer and a network layer, wherein the access layer is used for field sensor wired routing and wireless networking, RS232/485, wifi, Zigbee and Lora are mainly adopted, and the network layer is mainly used for accessing and transmitting heterogeneous networks such as GSM, 4G, 5G, the Internet of things, the communication network and the like;

the support layer design mainly comprises cloud computing, data mining, data visualization, multi-source heterogeneous data fusion, database-level GIS/GPS/RS technology and the like;

the information application layer mainly comprises monitoring, alarm query, intelligent control, an industry interface, an information system and a decision support system (a growth period and a soil moisture control threshold value).

The acquisition control device takes an embedded single chip microcomputer STM32F103RET6 controller as a core, realizes acquisition of water level, flow, gate level, pump station, video and gate (valve) opening and closing information, and realizes information transmission through Lora and RS232/485 modules.

The data mining mainly comprises the steps of adopting an LSTM long-short term memory neural network to predict the rainfall amount of future 5-15 days according to the effective rainfall amount P in a historical time period, and determining the irrigation time, plan and irrigation amount by combining a crop irrigation system through continuous rainless day drought evaluation and main dry land crop drought evaluation to improve the irrigation and fertilization accuracy.

The irrigation system mainly comprises three crops of dragon fruit, kiwi fruit and tea in the growth period and the soil moisture control threshold value which are shown in tables 6, 7 and 8.

TABLE 6 Dragon fruit growth period and soil moisture control threshold

TABLE 7 Kiwi fruit growth period and soil moisture control threshold

TABLE 8 growth period of tea and soil moisture control threshold

The intelligent control system comprises a BPNN-based multi-sensor data fusion such as an EC sensor, a PH sensor, a soil temperature and humidity sensor (T, H) and the like, and the structure of the intelligent control system is shown in FIG. 8. Adopting a three-layer (4-8-1) structure, namely 4 input (EC, PH, T and H), the hidden layer number is 8, the output layer number is 1 (target nutrient solution conductivity y), and establishing a BPNN multi-sensor data fusion nonlinear function as follows:

y＝f(ec,ph,t,h) (1)

wherein ec, ph, t, h are inputs of the neural network.

The intelligent control is mainly a two-dimensional incremental fuzzy PID control method, namely a two-dimensional incremental fuzzy PID control model which takes the deviation e and the deviation change rate ec between the actual soil conductivity r and the soil target conductivity y.

Through the fuzzification processing, the fuzzy control rule, the fuzzy decision and the defuzzification processing links, the output quantity is 3 input correction parameters delta Kp, delta Ki and delta Kd of the PID controller, and a two-dimensional incremental fuzzy PID control model is formed.

When e (t) and ec (t) change according to different rules, a reasonable fuzzy rule is worked out and made according to manual experience, PID parameters are modified on line according to the fuzzy rule, so that the system performance is optimal, then fuzzy reasoning is carried out according to the fuzzy rule, the reasoning result is subjected to defuzzification processing, the clear quantities delta Kp, delta Ki and delta Kd are calculated through quantization factors (or proportional factors) to serve as output quantities, finally actual control quantities of the sampling moment are obtained through Kp, Ki and Kd, and the opening time t of the electromagnetic valve is finally determined;

transfer function for fertilization bucket with buffer

The two-dimensional incremental fuzzy PID control model considers that the process of mixing fertilizer by the Venturi ejector while absorbing fertilizer is increased by virtual fertilizer mixing volume V in the buffer fertilizer mixing barrel_FThe mixing process is the combination of plug flow and ideal stirring and mixing, the whole system is a second-order lag system, the transfer function of the system is (8),

in the formula (8), K₂Representing the second-order system gain developed after adding the venturi ejector,

T_Fshowing the time constant of the nutrient solution preparation process in the venturi ejector,

V_Findicating an increased premixing volume after the addition of the venturi ejector; t is_p＝γT_rAnd gamma represents a mixing coefficient,for advection mode, γ is 0, for ideal stirring mode, γ is 1,

V_Tfor effective mixing volume of fertilizer mixing barrel, Q_FRepresenting the liquid flow into the venturi ejector; qw represents the water flow rate injected into the fertilizer mixing barrel; tau represents a fertilizer mother liquor Q_NSMixing delay and measuring delay time of the buffer fertilizer mixing barrel, wherein the delay time comprises flowing time and mixing time of liquid in a pipeline, and tau is (1-gamma) Tr is 1; τ' is the new lag time, Q_NS≤Q_F＜Qw，Q_NSShowing the flow of fertilizer mother liquor injected into the fertilizer barrel.

The method comprises the steps of acquiring images of dragon fruits, kiwi fruits and tea leaves under the daylight condition, judging the water shortage state of crops by using a binocular machine vision technology, then extracting red, green and blue (RGB) three-color components, relative coefficients grb and chromaticity H of the components, adopting red components R, green components G, blue components B, R components and B components, independently learning and extracting each local feature of data through multi-layer convolution and pooling operations by using CNN, obtaining more effective abstract feature mapping relative to an explicit feature extraction method, and quickly judging indexes of the water shortage state of the dragon fruits, the kiwi fruits and the tea leaves.

CNN sets convolution kernel size of 1 st convolution layer of network as 9 x 9, number as 12, activation function as RELU; the convolution kernel size of the 2 nd convolution layer is 3 multiplied by 3, the number is 18, each output characteristic graph is obtained by convolution on all characteristic graphs of the previous layer by different convolution kernels, corresponding elements are added and biased after being accumulated, and then the convolution kernel is activated by a RELU function; the pooling layer adopts a mean value pooling method, the down-sampling scales are all 2 multiplied by 2, the number of neurons in an MLP hidden layer is set to be half of that of a rasterization layer, an output layer is a single neuron and is used for real value regression, the network is trained by utilizing a BP algorithm, the random initialization of convolution kernel weight values is set, the bias is all 0 initialization, the optimal parameter is selected by adopting a leave-one cross verification method, a loss function is defined as the Euclidean distance, and the calculation is carried out by the formula (12)

Wherein yp is a network predicted value, yt is an experimental measured value, the network learning rate is set to be 0.6 through experiments, and the maximum iteration number is 1000.

The intelligent water and fertilizer integrated fertilizer applicator is used for flexible secondary development according to the scale of an irrigation area, the condition of a water source, the condition of crops, irrigation habits and application scenes, and can be set into a 1-5 channel series according to requirements, and comprises a jet device, a multi-channel Venturi mixer, a general MONBUS communication protocol and a Tiny6410 embedded single board computer,

the system specifically comprises a control box system, a filtering system, a pipe network system, a metering system, a fertilizer sucking system, a fertilizer mixing system and a fertilizer injecting system.

Wherein with respect to protocol description: the self-organizing network protocol relates to a front-end data acquisition controller, a mobile phone end and a background, and comprises a collector and a background which are communicated, wherein the collector is communicated with the mobile phone (through Bluetooth), the acquired environmental parameters comprise humidity and flow (the sensors except the two sensors can be automatically increased in the following protocol), the newly added control operation comprises mode selection, manual switch control, soil moisture control threshold setting and time period setting, and the newly added reading operation comprises current working mode reading, set soil moisture control threshold reading, set time period reading and current brake (valve) actual state (switch).

The frame format and the control code of the ad hoc network protocol are shown in table 1 and table 2.

TABLE 1 frame Format

The control code represents the operation that is required to be performed, 2 bytes, and the format is shown in table 2:

TABLE 2 definition of control codes

The control box system is connected with an EC sensor, a PH sensor, a soil temperature and humidity sensor, a meteorological station and a video monitoring camera;

the fertilizer absorbing system is connected with the fertilizer storage tank, the fertilizing valve, the butterfly filter, the fertilizer absorbing channel, the ejector and the Venturi mixer;

the water source pipe network is mainly used for pumping water to a high-level water pool by a pump station for self-flow or pressurized irrigation, the pump station control adopts a PID control method, and the high-level water pool and the pump station are subjected to operation energy efficiency monitoring closed-loop control;

the reducing angle alpha of the ejector is 6 degrees, the reducing angle beta of the ejector is 4 degrees, the throat diameter d0 of the ejector is 10 degrees, and the jet ejector is connected with the suction chamber, the nozzle, the throat and the diffusion pipe;

the Venturi mixer is connected with a lower main pipeline pressure gauge, an ejector, an upper main pipeline pressure gauge, a fertilizer absorbing channel and a manual valve;

the fertilizer absorbing system is based on a four-way ejector and is used for absorbing and mixing different types of unit element liquid fertilizers;

the fertilizer mixing system is connected with a water source pipe network, a control box system, a water source main pipeline water pump, a main pipe filter, a pressure reducing valve, a ball float valve, a buffering fertilizer mixing barrel, an EC sensor, a PH sensor, a pressure sensitive switch and an upper branch pipe filter;

the fertilizer injection system is connected with a fertilizer pump, a check valve, an irrigation electromagnetic valve and a field pipe network.

The method for accurately regulating and controlling the water and the fertilizer based on the Internet of things comprises the following steps:

the method comprises the following steps: initializing a system;

step two: awakening the wireless sensor node;

step three: the initial setting is time T0, humidity R0, nutrient EC0 and PH0, the initial values are communicated with the cloud server, data fusion processing is synchronously completed, and the cloud server is extracted to complete diagnosis, decision and prediction according to a crop irrigation system;

step four: selecting a working mode, namely selecting three working modes of irrigation, fertilization and water and fertilizer integration;

step five: when irrigation is selected, the system finishes comparison between the current humidity and the set humidity, the current humidity R1 is lower than the set humidity R0, irrigation is started when no rainfall P is predicted in the next 5 days, and irrigation is stopped when the current humidity R1 is not lower than the set humidity R0; when fertilization is selected, the system finishes the comparison between the current EC1 and the set EC0, if the current EC1 is lower than the set EC0 and no rainfall P is predicted in the future 5 days, fertilization is started, and if the current EC1 is not lower than the set EC0, fertilization is stopped; stopping irrigation; and when water and fertilizer integration is selected, the system finishes the comparison of the current humidity R1 or the current EC1 with the set humidity R0 or the set EC0, starts the irrigation and fertilization if the current humidity R1 or the current EC1 is not equal and the rainfall P is predicted to be absent in the future 5 days, and stops the irrigation and fertilization if the current humidity R1 or the current EC1 is equal to the set humidity R0 or the set EC 0.

Wherein, the system also comprises information system software which is built by adopting a JavaEE platform and a Spring MVC + Mybaties framework basically,

the information system mainly comprises a water conservancy Internet of things soil moisture content monitoring and water and fertilizer integrated irrigation control system, an intelligent video monitoring system, a pump room pool self-adaptive control system and a remote network meteorological monitoring system, is used as a comprehensive control system integrating remote data acquisition, remote video transmission, remote monitoring control, pressure regulation and flow regulation,

the information system can comprehensively check the flow, pressure, PH value, wind speed, soil humidity, air humidity, temperature, radiation and high-level pool liquid level real-time data of each substation, automatically calculate the average irrigation water utilization rate of the region,

the area average irrigation water utilization rate is calculated according to the irrigation water consumption of different types and scales of wool in the area and the irrigation water utilization rate of the corresponding spot irrigation areas in a weighted average manner:

in formula (13): eta_Region(s)-regional average irrigation water utilization; eta_{Big (a)}、η_In、η_SmallRespectively are regionally largeThe utilization rate of irrigation water of medium and small irrigation areas; w_{Big (a)}、W_In、W_SmallThe water consumption for the wool in the large, medium and small irrigation areas of the area, ten thousand m3,

for the average irrigation water utilization rate of the bundled large irrigation areas, the formula is calculated as shown in formula (14):

in the formula: eta_{Big (a)}-bundling the irrigation water utilization rate of the large irrigation areas; eta_In、η_Small-the irrigation water utilization ratio of the medium and small irrigation areas which form the bundled large irrigation area respectively; w_In、W_SmallThe water consumption for the wool forming the medium and small irrigation areas of the large-sized irrigation area to be bundled, m3, respectively.

Compared with the prior art, the invention has the beneficial effects that: the invention adopts a four-layer architecture mode of a sensing layer, a transmission layer (an access layer and a network layer), a supporting layer and an application layer of the Internet of things; independently developing a multifunctional acquisition control device based on STM32F103RET6 and optimizing the structures of a jet and a Venturi mixer; the intelligent water and fertilizer integrated networking is realized by combining a short-distance communication protocol with various protocols such as Lora, TCP/IP and the like; a transfer function with a buffer fertilizer mixing barrel is designed and integrated with a two-dimensional incremental fuzzy PID control method, so that the response speed and the accuracy of the soil moisture and nutrient control process are realized; predicting the precipitation of future 7-15 days by adopting an LSTM long-short term memory neural network, and evaluating the drought conditions of continuous rainy-free days and main dry land crops; the irrigation time, schedule and irrigation amount are determined by combining the irrigation system of the Guizhou pitaya, the kiwi fruit and the tea, and the accuracy of irrigation and fertilization is improved; a calculation formula for the average irrigation water utilization rate of a large-bundle large irrigation area is designed, technical support is provided for evaluating the water saving rate of a system, and the requirement for quickly and accurately realizing water and fertilizer integration in agricultural production can be met.

The invention solves the problems that the prior water and fertilizer integrated fertilizer applicator mostly adopts an integral import, the communication protocol of the system is not open, the compatibility is poor, the remote control is difficult to realize, a user can not individually customize and develop for the second time according to the irrigation system and the fertilizer application formula requirements of different areas and different crops, the adaptability is poor, and the like. The method can meet the requirements of water and fertilizer of different growth periods of the dragon fruit, the kiwi fruit and the tea in Guizhou, automatically detect, allocate and supply the water and the fertilizer according to the soil environment and the nutrient content conditions, achieve accurate control of irrigation amount, fertilization amount and agricultural management informatization, and realize water conservation of 41.3% and fertilizer conservation of 32.1% compared with the traditional water-saving irrigation technology; the irrigation management mode is changed from decentralized management to centralized management, 1-2 people can complete farmland irrigation management work of more than ten thousand mu, labor force is saved by 1023 working days/hundred mu, and the novel modernization requirement of the characteristic agriculture of Guizhou mountainous regions is met.

Drawings

FIG. 1 shows an Internet of things accurate water and fertilizer regulation and control system architecture model of an intelligent water and fertilizer integrated fertilizer applicator;

FIG. 2 is a diagram of an acquisition control device;

in fig. 2: 1-switching value output (solenoid valve); a 2-4G communication module; 3-a rainfall sensor; 4-RS232/485 serial port; 5-solar power supply interface (power interface); 6-Lora communication module; 7-analog input (humidity sensor); 8-analog input (humidity sensor); 9-analog output (electric valve); 10-analog output (electrically operated valve); 11-RJ-45 input (camera interface); 12-pulse signal (flow); 13-analog input (temperature sensor); 14-JATG interface; 15-analog input (flow meter); 17-an intelligent water and fertilizer integrated machine interface; 18-on light output (solenoid valve); 19-embedded singlechip STM32F103RET 6;

fig. 3 is an internal structural view of the ejector of the present invention;

in fig. 3: 26-a suction chamber; 27-a nozzle; 28-throat pipe; 29-diffuser (venturi effect);

FIG. 4 is a schematic diagram of the venturi mixer of the present invention;

in fig. 4: 30-lower main pipe pressure gauge; 31-an ejector; 32-upper main pipe pressure gauge; 33-a fertilizer suction channel; 34-a manual valve;

FIG. 5 is an assembly drawing of the intelligent water and fertilizer integrated fertilizer applicator of the present invention;

in fig. 5: 4-control box system; 19-a filtration system; 20-a pipe network system; 21-a metering system; 22-a fertilizer suction system; 23-a fertilizer mixing system; 24-a fertilizer injection system;

FIG. 6 is a system installation diagram of the parallel type water and fertilizer integrated machine of the invention;

in fig. 6: 1-a fertilizer-assisting pump; 2-buffering the fertilizer mixing barrel; 3-an ejector; 4-control box system; 5-an EC sensor; 6-a pH sensor; 7-a ball float valve; 8-a pressure gauge; 9-upper branch filter; 10-a pressure sensitive switch; 11-a check valve; 12-water source main pipeline water pump; 13-a main tube filter; 14-a pressure relief valve; 15-fertilizer storage barrel; 16 a fertilizing valve; 17-a butterfly filter; 18-a flow meter;

FIG. 7 is a flow chart of the automatic control system;

FIG. 8 BPNN-based multi-sensor data fusion model structure

FIG. 9 is a block diagram of a BPNN fuzzy adaptive PID structure

FIG. 10 is a simulation model of an incremental fuzzy PID control system;

FIG. 11 illustrates the FUZZY-PID subsystem connection model;

FIG. 12 is a fuzzy PIDMATLAB simulation diagram;

FIG. 13 is an overall architecture diagram of a fertigation decision support system;

FIG. 14 is a flow chart of a fertigation decision support system modeling;

FIG. 15 evaluation one of drought in continuous no-rain days;

FIG. 16 evaluation II of drought in continuous rainy days;

FIG. 17 is a management system E-R diagram.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

As shown in fig. 1, the accurate regulation and control system for water and fertilizer of the internet of things mainly comprises an information sensing layer, an information transmission layer, a supporting layer and an information application layer.

Information perception layer

The information perception layer comprises bar codes, two-dimensional codes, RFID, an STM32F103RET 6-based independent research and development multifunctional acquisition control device, an intelligent sensor, an intelligent camera, hypertext information, a self-organizing network and the like.

The acquisition control device is shown in figure 2, and is connected with an on-off output (electromagnetic valve) 1, a 4G communication module 2, a rainfall sensor 3, an RS232/485 serial port 4, a solar power supply interface 5 (power interface), a Lora communication module 6, an analog input 7 (humidity sensor), an analog input 8 (humidity sensor), an analog output 9 (electrically operated valve), an analog output 10 (electrically operated valve), an RJ-45 input 11 (camera interface), a pulse signal 12 (flow), an analog input 13 (temperature sensor), a JATG interface 14, an analog input (flow meter) 15, an intelligent water and fertilizer all-in-one machine interface 17, an on-off light output 18 (electromagnetic valve) and an embedded single chip microcomputer 19STM32F103RET6, wherein the device takes an embedded single chip microcomputer STM32F103RET6 controller as a core to realize acquisition of water level, flow, gate position, pump station, video and gate (valve) opening and closing information, and realize the transmission of the information through modules such as Lora, RS232/485, etc.

The intelligent equipment for accurately regulating and controlling the water and the fertilizer is mainly an intelligent water and fertilizer integrated fertilizer applicator. Aiming at the moisture control threshold value of the growth period of the dragon fruit, the kiwi fruit and the tea in Guizhou, the optimized ejector shown in fig. 3 and the multi-channel Venturi mixer shown in fig. 4 are taken as cores, the universal MONBUS communication protocol and the Tiny6410 embedded single board computer are adopted, online networking and remote control are realized, a 1-5-channel series intelligent water and fertilizer integrated fertilizer applicator can be customized according to the requirements of users, the secondary development is flexible according to the scale, the water source condition, the crop condition, the irrigation habit and the application scene, the applicability and the practicability are strong, and the water control method can be popularized and applied in large areas in agricultural parks, irrigation areas and the like.

The intelligent water and fertilizer integrated fertilizer applicator is characterized in that a water source is automatically drained or irrigated under pressure to a high-level water tank by adopting a pump station to lift water according to the features of mountain terrain fluctuation, strip and block division and the feature of long-distance water delivery irrigation.

An intelligent water and fertilizer integrated fertilizer applicator control box (comprising devices such as a controller) is connected with a soil humidity sensor PH-TS, a soil EC, a PH value sensor, a meteorological station and a general MONBUS communication module. The environmental parameters such as soil moisture, soil temperature, soil EC, PH value, air humiture and the like can be automatically collected.

An intelligent water and fertilizer integrated fertilizer applicator is shown in figure 5. The support system is connected with a control box (comprising a controller, an information acquisition transmission layer and the like) system 4, a filtering system 19, a pipe network system 20 (a water source pipe network and an irrigation pipe network), a metering system 21, a fertilizer suction system 22, a fertilizer mixing system 23 and a fertilizer injection system 24.

The control box (comprising a controller, an information acquisition transmission layer and the like) system 4 is connected with an EC sensor 5, a PH sensor 6, a soil temperature and humidity sensor 35, a meteorological station 36 and a video monitoring camera 37.

The fertilizer absorbing system 22 is connected with the fertilizer storage tank 15, the fertilizing valve 16, the butterfly filter 17, the fertilizer absorbing channel, the ejector 20, the venturi mixer and the like.

The optimized ejector has a taper angle α of 6 °, a taper angle β of 4 °, and a throat diameter d0 of 10, and is connected to the suction chamber 26, the nozzle 27, the throat 28, and the diffuser 29 (venturi effect).

The optimized Venturi mixer is connected with a lower main pipeline pressure gauge 30; an ejector 31; an upper main pipe pressure gauge 32; a fertilizer suction passage 33; a manual valve 34.

The fertilizer absorbing system 22 is based on a four-way ejector and realizes the absorption and mixing of liquid fertilizers of different types of single elements. In the operation of the water and fertilizer integrated machine, the fertilizer absorbing system 22 enters the nozzle reducing section of the jet device through the upper support pipe filter 9 under the action of the fertilizer assisting pump 1 of the main irrigation pipeline, the water flow pressure is increased along with the reduction of the cross section area, and the water flow speed is increased along with the increase of the water flow pressure. According to the working principle of the ejector, the vacuum negative pressure generated by the suction chamber and the external air pressure form a pressure difference, and the fertilizer mother liquor is sucked from the fertilizer storage tank 15 connected with the fertilizer suction port of the ejector by utilizing the pressure difference, so that the fertilizer suction process is completed.

The installation of the parallel type water and fertilizer integrated machine system is shown in figure 6. The water source pipe network is connected with a control box (comprising a controller and other devices) system, a water source main pipeline water pump 12, a main pipeline filter 13, a pressure reducing valve 14, a ball float valve 7, a buffer fertilizer mixing barrel 2, an EC sensor, a PH sensor, a pressure sensitive switch 10 and an upper branch pipeline filter 9 at the position of 0.15 MPa-0.25 MPa.

The fertilizer injection system 24 is connected with a fertilizer pump 1, a check valve 11, an irrigation electromagnetic valve 38 and a field pipe network.

The model of the controller is S7-200SMRT, the model of the EC sensor is a PH/ORP controller + 5m conventional electrode, the model of the PH sensor is 0-4000 mu S, the model of the fertilizer pump 1 is DL8-120, and the model of the irrigation electromagnetic valve 38 is ZS1DF13V4D 16.

The self-organizing network protocol relates to a front-end data acquisition controller, a mobile phone end and a background, and comprises a collector and a background which are communicated, wherein the collector is communicated with the mobile phone (through Bluetooth). The collected environmental parameters include humidity and flow (the sensors except the two sensors can be automatically added in the following protocol), the newly added control operation includes mode selection, manual switch control, soil moisture control threshold value setting and time period setting, and the newly added reading operation includes current working mode reading, set soil moisture control threshold value reading, set time period reading, current brake (valve) actual state (switch) and the like.

Description of protocol

When a request is sent to a monitor at a background or a mobile phone end, FF FF FF FF FF FF FF FF FF FF (10 0 xFFs) of wake-up codes are added, the wake-up codes are sent first, and a request frame is sent after an interval of 0.2-0.5 seconds.

Protocol frame format

Byte format

Serial transfer format of bytes: 1 bit start bit; 8 bit data bits; 1 stop bit, no parity.

Frame format

Description of frame format

Each frame is composed of 7 fields, such as a frame start symbol, an address field, a control field, a data length field, a data field, a frame information check field, a frame end symbol and the like. The format is shown in table 1:

TABLE 1 frame Format

Frame start Symbol (STA)

Identifying the beginning of a frame of information, 1 byte, whose value is fixed to E8H ═ 11101000B

Address field (AD)

Address identifying current receiving (transmitting) device, 2 bytes, 0xFFFF being broadcast address

Control area (C)

The control code represents the operation required to be executed, 2 bytes, and the format is shown in table 2:

TABLE 2 definition of control codes

D7: direction of conveyance

D7 ═ 0: instruction frames sent by the background/mobile phone;

d7 ═ 1: a response frame sent by the collector.

D6: abnormal mark

D6 ═ 0: a table correct answer;

d6 ═ 1: the response to the exception information is tabulated.

D5-D0: function code

(01H) The method comprises the following steps Reading SD card recent data (days) (Bluetooth)

(02H) The method comprises the following steps Production setup (set address number) (bluetooth);

(03H) the method comprises the following steps Reading collector fetch address (bluetooth);

(04H) the method comprises the following steps Clear rainfall information (this system is not used);

(05H) the method comprises the following steps Setting the working time(s) after the CPU of the collector is awakened;

(06H) the method comprises the following steps Reading the working time(s) after the CPU is awakened;

(07H) the method comprises the following steps Setting sleep time(s) of collector CPU

(08H) The method comprises the following steps Reading collector CPU dormancy time(s)

(09H) The method comprises the following steps Timing

(0 AH): configuring sensor information

(0 BH): reading out already deployed sensors

(0 CH): equipment production initialization, SD card data retention

(0 DH): altering flow parameters

(0 EH): reading all working sensor information and data collected (passive mode, i.e. responding to commands issued by the background)

(0 FH): the collector actively sends the data collected by all the working sensors to the background

(10H) The method comprises the following steps Temporary reservation

(11H) The method comprises the following steps Working mode selection (0-manual + time 1-automatic 2-manual)

(12H) The method comprises the following steps Current operating mode reading

(13H) The method comprises the following steps Manually controlled brake (valve) switch operation

(14H) The method comprises the following steps The setting range of the soil moisture control threshold value setting unit Lx is 0-65535 Lx

(15H) The method comprises the following steps Reading out the set soil moisture control threshold value unit Lx

(16H) The method comprises the following steps Time period settings including evening hours, minutes and early morning hours, minutes

(17H) The method comprises the following steps Read out the set time period including the evening hour, minute and early morning hour, minute

(18H) The method comprises the following steps Reading the switch state of the current brake

(19H) The method comprises the following steps Used for testing and reading the flag bit of the collector software gate (valve)

(1 AH): setting PWM threshold value and changing opening of electric valve

Length (LEN)

Representing the total number of bytes in the data field, 2 bytes

DATA field (DATA)

Data of

Check code (CRC)

CRC-16 cyclic redundancy check is adopted, the check contents are AD, C, LEN and DATA, and 2 bytes are adopted

END of frame (END)

End of frame information is identified, 1 byte, fixed to E6H-11100100B

Byte deposit mode

Using little-endian mode, i.e. with the low byte stored at the low address, the low byte at the front and the high byte at the back

Information transmission layer

The information transmission layer comprises an information access layer and a network layer. And the access layer field sensor is subjected to limited routing wireless networking and access. RS232/485, wifi, Zigbee and Lora are mainly adopted, and the network layer mainly has access and transmission of heterogeneous networks such as GSM, 4G, 5G, the Internet of things and a communication network.

Supporting layer

The support layer design mainly comprises cloud computing, data mining, data visualization, multi-source heterogeneous data fusion, database-level GIS/GPS/RS technology and the like.

The data mining mainly comprises the steps of adopting an LSTM long-short term memory neural network to predict the rainfall amount of future 7-15 days according to the effective rainfall amount P in the historical time period, and determining the irrigation time, plan and irrigation amount by continuous rainless day drought evaluation and main dry land crop drought evaluation in combination with a crop irrigation system. Improve the accuracy of fertigation.

Continuous no-rain-day drought assessment

The continuous rainy-day drought evaluation mainly carries out self-timing query time evaluation on drought in a certain region, can help query and can provide quick evaluation by knowing the drought condition in a certain time period at any time. The self-defined time evaluation of a single area can be carried out on the page, and the self-defined time evaluation of all sites can also be carried out.

Evaluation of drought in major dry land crops

Drought assessment of dry land crops is similar to drought assessment on rainy days, except that crop options are added at the time of assessment. Crop evaluation involves different crops having different crop coefficients, and the updating and modification of the crop coefficients is performed in an "administrator tool".

The irrigation database mainly comprises an irrigation object information table, wherein the included data items comprise crop names (char), lower humidity limits (float), upper humidity limits (float), environmental temperatures (float) and the like; an irrigation area information table, wherein the data items comprise irrigation area ID (int), irrigation area (float), crop name (char) and irrigation mode (char); and the irrigation plan information table comprises data items including irrigation area ID (int), irrigation time (datetime), irrigation quantity (float) and the like. An irrigation state information table, wherein the data items comprise irrigation area ID (int), equipment communication state (pool), temperature acquisition state (pool), humidity acquisition state (pool), temperature value (float) and humidity value (float); irrigation history information table. The data items include irrigation area ID (int), irrigation time (datetime), irrigation water amount (float), pre-irrigation humidity (float), post-irrigation humidity (float), and the like.

The fertilization database mainly comprises a fertigation object information table, and the contained data items comprise crop names (char), EC lower limits (float), EC upper limits (float), environment temperatures (float) and the like; a fertilization area information table, wherein the contained data items comprise fertilization area ID (int), fertilization area (float), crop name (char) and fertilization mode (char); and thirdly, a fertilization plan information table, wherein the contained data items comprise fertilization area ID (int), fertilization time (datetime), fertilization amount (float) and the like. A fertilization state information table, wherein the data items comprise fertilization area ID (int), equipment communication state (pool), soil nutrient acquisition state (pool), EC value (float) and PH value (float); fertilizing history information table. The data items include fertilization area id (int), fertilization time (datetime), fertilization amount (float), pre-fertilization ec (float), post-fertilization ec (float), and the like.

Information application layer

The intelligent control comprises a two-dimensional incremental fuzzy PID control method and automatic irrigation and fertilization control.

The intelligent control comprises data fusion of multiple sensors such as an EC sensor, a PH sensor and a soil temperature and humidity sensor (T, H) based on BPNN, and the structure of the intelligent control is shown in FIG. 8. Adopting a three-layer (4-8-1) structure, namely 4 input (EC, PH, T and H), the hidden layer number is 8, the output layer number is 1 (target nutrient solution conductivity y), and establishing a BPNN multi-sensor data fusion nonlinear function as follows:

y＝f(ec,ph,t,h) (1)

and ec, ph, t and h are input of the neural network, wherein the learning rate is 0.2, the inertia coefficient is 0.05, the positive and negative sigmoid functions of the activation function of the hidden layer are shown in (2), and the positive and negative sigmoid functions of the activation function of the output layer are shown in (3).

On the basis of the conventional fuzzy PID, the output quantity is 3 input correction parameter increments delta Kp, delta Ki and delta Kd of the PID controller through links such as fuzzification processing, fuzzy control rule, fuzzy decision, defuzzification processing and the like.

When e (t) and ec (t) change according to different rules, a reasonable fuzzy rule is worked out and made according to manual experience. And on-line modifying each PID parameter according to fuzzy rule to make system performance reach optimum. Then, fuzzy reasoning is carried out according to a fuzzy rule, the reasoning result is subjected to defuzzification processing, the clear quantities delta Kp, delta Ki and delta Kd are calculated through a quantization factor (or a proportional factor) to be used as output quantities, finally actual control quantities of the sampling moment are obtained through Kp, Ki and Kd, and finally the opening time t of the electromagnetic valve is determined.

Fuzzy PID control strategy

The system inputs the error e (t) of the conductivity and the change rate ec (t) of the conductivity error, outputs delta Kp, delta Ki and delta Kd to further adaptively adjust three parameters of the PID controller, improves the dynamic response speed of the system, and enhances the adaptability of the system to the change of the water and fertilizer of the soil.

In the formula: r (t) is actually measured soil conductivity; y (t) is target soil conductivity; e (t) is the error in conductivity at time t; e (t-1) is the error in conductivity at time t-1.

Fuzzy incremental PID controller

On the basis of the conventional PID, the output quantity is 3 input correction parameters delta Kp, delta Ki and delta Kd of the PID controller through links of fuzzification processing, fuzzy control rule, fuzzy decision, defuzzification processing and the like. A block diagram of the fuzzy adaptive PID structure is shown in fig. 9.

Input and output fuzzy variable and domain of discourse

The soil conductivity of the multi-channel water and fertilizer integrated device is increased from 0 to 2.5mS/cm, and the used time is about 180 s. The basic domain of deviation e is: [ -0.8,0.8 ]; the basic domain of deviation ec is: [ -0.8,0.8 ]; the set-up ambiguity domain for the bias is: [ -4,4 ]; the set ambiguity domain for the variance of the deviation is: [ -4,4 ]; the basic domain of output solenoid valve t is [0,180 ]; the fuzzy domain of the output solenoid valve y (t) is [0,3 ]. Inputting quantization factors Ke and Kec of the quantities e and ec; the scaling factor for the output is Kt.

The basic argument for the output Δ Kp is: [ -100, 100](ii) a The basic domain of argument for the output Δ Ki is: [ -50, 50](ii) a Output ofThe basic domain of Δ Kd is: [ -100, 100](ii) a The ambiguity domains for the three outputs are all set as: [ -3,3]. The initial value of PID is set to K by the analysis of the single PID controller_p0＝8，K_i0＝0.01，K_d0When the value is-27, and the variation range of Kp is [10, 20 ]]Ki varies in the range of [0,1]Kd varies in the range of [0, 0.16 ]]And the control effect of the system is better. Three parameters K of PID controller_p、K_i、K_dAnd the corresponding output scale factor formula (6) is calculated as a_p＝0.0005、a_i＝0.05、a_d＝0.008。

In formula (7): k_Pmax、K_imaxAnd K_dmaxRespectively correspond to K_p、K_i、K_dMaximum value within the range of variation; Δ K_Pmax、ΔK_imaxAnd Δ K_dmaxRespectively correspond to Δ K_p、ΔK_i、ΔK_dMaximum value within the variation range.

Control tables for fuzzy controllers

And (3) selecting the deviation e and the deviation change ec as input and outputting the increment delta Kp, delta Ki and delta Kd according to system operation parameters. The input linguistic variables and the output linguistic variables are respectively provided with seven fuzzy values of { NB, NM, NS, ZO, PS, PM and PB } (respectively representing negative large, negative middle, negative small, zero, positive small, middle and positive large). And (4) formulating a fuzzy control rule table of tables 3-5 according to the principle and the empirical knowledge of the system.

TABLE 3 fuzzy control rules for Δ Kp

TABLE 4 fuzzy control rules for Δ Ki

TABLE 5 fuzzy control rule of Δ Kd

Transfer function with buffer fertilization bucket

The two-dimensional incremental fuzzy PID control model considers that the process of mixing fertilizer by the Venturi ejector while absorbing fertilizer is increased by virtual fertilizer mixing volume V in the buffer fertilizer mixing barrel 2_FThe mixing process is a combination of plug flow and ideal stirring and mixing. The whole system is a second-order lag system, and Qw is set to be 1L/s, Q_F＝0.05L/s。Q_NS0.01L/s. Taking a 10L fertilizer mixing barrel as an example (the working effective volume is 9L), the system transfer function is (8), V_T＝5L，V_F＝4L。

In the formula (8), K₂Representing the second-order system gain developed after adding the venturi ejector 3,

T_Fshowing the time constant of the nutrient solution preparation process in the venturi ejector 3,

V_Findicating an increased premixing volume after the addition of the venturi ejector 3; t is_p＝γT_r0.8-5-4, gamma denotes the mixing coefficient, and is used herein

For advection mode, γ is 0, and for ideal stirring mode, γ is 1.

V_TThe effective mixing volume of the fertilizer mixing barrel is provided. Q_FRepresenting the liquid flow into the venturi ejector 3; qw represents the water flow rate injected into the fertilizer mixing barrel; tau represents a fertilizer mother liquor Q_NSMixing delay and measuring delay time of the buffer fertilizer mixing barrel 2, wherein the delay time comprises flowing time and mixing time of liquid in a pipeline, and tau is (1-gamma) Tr is 1; τ' is the new lag time, Q_NS≤Q_F＜Qw，Q_NSShowing the flow of fertilizer mother liquor injected into the fertilizer barrel.

Fuzzy PID controller simulation model

Simulating by adopting a Simulink environment of MATLAB, inputting a step signal with ec being 2ms/cm, controlling the transfer function of an object to be controlled to be (8), and outputting a control performance curve of the traditional PID and fuzzy self-adaptive PID controller by a synchronous oscilloscope by setting the same initial value. FIG. 10 is a simulation model of an incremental FUZZY PID control system, and FIG. 11 is a FUZZY-PID subsystem model.

The transfer function of the fuzzy PID controller simulation model EC value closed-loop regulation and control system is as follows:

setting a target EC value of 2ms/cm and an initial value K of PID_p0＝8、K_i0＝0.01、K_d0-27; the operating frequency of the fertilizer injection valve 16 is 6Hz, and the fertilizer injection valve is always in an open state (the duty ratio is 100%); the inlet pressure of the apparatus was 0.25 MPa. The traditional PID control simulation result shows that: the peak error is 29.17%, ec is stabilized at 1.79ms/cm, and the steady-state error is 10.5%.

Two-dimensional incremental fuzzy PID controllers. The initial condition settings are the same, and the simulation result shows that: the increment type fuzzy PID overshoot is about 5.88%, the EC value of the system is stably controlled at 1.97ms/cm, the steady state error is 5%, and the stable stage is started from about 50 s.

In fig. 12, the cyan line is the target step signal with the input ec being 2ms/cm, the blue dotted line is the incremental fuzzy PID control output, and the red line is the conventional PID control output. As can be seen from fig. 11: the convergence time of the incremental fuzzy PID and the traditional PID control is about 50 s. The increment type fuzzy PID overshoot is about 5.88%, the ec value of the system is stably controlled to be 1.97ms/cm, and the steady-state error is 5%; and the traditional PID control has a peak error of 29.17%, ec is stabilized at 1.79ms/cm, and a steady-state error is 10.5%. Compared with simulation results, the incremental fuzzy PID has higher response speed and higher precision, and can meet the requirement of quickly and accurately realizing water and fertilizer integration in agricultural production.

The automatic irrigation and fertilization control is provided with automatic irrigation and fertilization control according to the humidity difference, the fertilization amount and the set time, the software diagram of an automatic control system is shown in figure 7, and the specific flow is as follows:

scheme F1: and (5) initializing the system.

Scheme F2: and waking up the wireless sensor node.

Scheme F3: the initial setting is time T0, humidity R0, nutrients EC0 and PH0, the initial values are communicated with the cloud server, data fusion processing is synchronously completed, and the cloud server is extracted to complete diagnosis, decision and prediction according to a crop irrigation system.

Scheme F4: the working mode is selected, and three working modes of irrigation, fertilization and water and fertilizer integration can be selected.

Scheme F5: when irrigation is selected, the system finishes comparison between the current humidity and the set humidity, the current humidity R1 is lower than the set humidity R0, irrigation is started when no rainfall P is predicted in the next 5 days, and irrigation is stopped when the current humidity R1 is not lower than the set humidity R0; when fertilization is selected, the system finishes the comparison between the current EC1 and the set EC0, if the current EC1 is lower than the set EC0 and no rainfall P is predicted in the future 5 days, fertilization is started, and if the current EC1 is not lower than the set EC0, fertilization is stopped; stopping irrigation; and when water and fertilizer integration is selected, the system finishes the comparison of the current humidity R1 or the current EC1 with the set humidity R0 or the set EC0, starts the irrigation and fertilization if the current humidity R1 or the current EC1 is not equal and the rainfall P is predicted to be absent in the future 5 days, and stops the irrigation and fertilization if the current humidity R1 or the current EC1 is equal to the set humidity R0 or the set EC 0.

Scheme F6: the system ends.

The decision support system functional architecture can be roughly divided into three layers, namely a source data layer which defines the type of data to be accessed and the corresponding data content; the data center is mainly responsible for analyzing and summarizing the accessed source data, and finally forming an application market region to provide data support for upper-layer application; the platform layer, namely the agricultural product irrigation and fertilization decision system is mainly responsible for providing decision results of irrigation and fertilization operations for daily production operators in various scenes, and also comprises management of deployed models and execution of decision tasks in various scenes by using the deployed models. The overall architecture of the fertigation decision support system is shown in fig. 13.

The decision support system modeling process comprises the following steps: data preparation, data preprocessing, modeling, model deployment, and model application. The data preparation is mainly to investigate data characteristics which can influence the model prediction result before modeling; the data preprocessing mainly comprises the steps of carrying out quality inspection and characteristic engineering on the provided data, and changing the data of the data mart layer into data which can be used as model input; the modeling process comprises algorithm model selection, model design, model training and model verification; the model deployment is to deploy a model with good fitting effect in the verification process to an application system, and the main contents include: adding a model, adjusting model parameters, evaluating the model and the like; the model application is to use the model deployed in the application system to give a predicted value to the index quantity in various scenes so as to guide production practice activities. The irrigation fertilization decision support system modeling flow chart is shown in fig. 14.

The irrigation system mainly comprises calculation of water demand of crops, growth periods of three crops of dragon fruit, kiwi fruit and tea and a wetting layer suitable for planning. Setting a soil moisture control threshold value through crops, and controlling irrigation according to +/-soil moisture information measured by a sensor by a system; and when the soil moisture is close to and lower than the lower limit value of the soil moisture, implementing irrigation, and stopping irrigation when the requirement of the upper limit value of the soil moisture is met so as to adapt to the requirement of high-efficiency water use of crops in different hydrological years.

The water demand of crops is in direct proportion to solar radiation, sunlight, temperature and wind speed and in inverse proportion to rainfall and humidity. The dragon fruit is a dry crop, and a field test method is adopted due to the limit of test conditions. Therefore, the crop water consumption is calculated by a farmland water balance method. A farmland water quantity balance method calculation formula:

ET＝ΔW+I+G+P-D (10)

in the formula: ET-Water consumption in each growth stage of the crop, mm; delta W is the water storage capacity change of soil in a time interval, mm; i-irrigation quantity in time interval, mm; g is groundwater supply quantity within a period of time, mm; p is the effective rainfall in time period, mm; d-the deep layer leakage in the time period, mm.

Wherein: Δ W is 10p Σ r_iH_i(θ_i2-θ_i1) (11)

In the formula: p-designed soil wetting ratio by spray irrigation, (%); ri-dry volume weight of soil (g/cm3) Hi corresponding to layer i-thickness of soil (cm) corresponding to layer i; theta i1 and theta i2, i-the value (%) of the water content of the i-th layer soil at the beginning and end of the period.

And (3) predicting the rainfall amount of future 5-15 days by using an LSTM long-short term memory neural network, and determining the irrigation time, plan and irrigation amount by continuous rainless drought evaluation and main drought evaluation of dry land crops in combination with a crop irrigation system.

The method for predicting rainfall by the LSTM long-short term memory neural network is characterized by comprising the following steps:

s1, step S1, acquiring a multi-feature sensor related to rainfall through the Internet of things platform, manually inputting 50-year-long sequence rainfall historical data, and acquiring rainfall data on a weather official website;

step S2, cleaning and normalizing the acquired historical data related to the multi-feature precipitation, and dividing the data into a training set and a test set according to the proportion;

step S3, the training set data is sent to an improved LSTM long-short term memory neural network, and the LSTM network model is subjected to iterative optimization;

and step S4, predicting the test set by using the trained LSTM model, and evaluating the model error.

Continuous no-rain-day drought assessment

The continuous rainy-day drought evaluation mainly carries out self-timing query time evaluation on drought in a certain region, can help query and can provide quick evaluation by knowing the drought condition in a certain time period at any time. The self-defined time evaluation of a single area can be carried out on the page, and the self-defined time evaluation of all sites can also be carried out. The calculation of the number of continuous rainless and rainless days and the evaluation of the drought in continuous rainless days are shown in fig. 15 and fig. 16.

Evaluation of drought in major dry land crops

The growth period and the soil moisture control threshold value are mainly three crops of Guizhou pitaya, kiwi fruit and tea.

TABLE 6 Dragon fruit growth period and soil moisture control threshold

TABLE 7 Kiwi fruit growth period and soil moisture control threshold

TABLE 8 growth period of tea and soil moisture control threshold

The binocular machine vision technology judges the water shortage state of crops by collecting images of pitaya, kiwi fruit and tea leaves under the daylight condition and then respectively extracting red, green and blue (RGB) three-color components, relative coefficients grb of the components and chromaticity H of the components. By adopting the red component R, the green component G, the blue component B, the R component and the B component, each local feature of data can be independently learned and extracted through multilayer convolution and pooling operations through CNN, more effective abstract feature mapping relative to an explicit feature extraction method is obtained, and indexes of water shortage conditions of pitaya, kiwi fruit and tea are rapidly judged.

CNN sets convolution kernel size of 1 st convolution layer of network as 9 x 9, number as 12, activation function as RELU; the convolution kernel size of the 2 nd convolution layer is 3 multiplied by 3, the number is 18, each output characteristic graph is obtained by convolution on all characteristic graphs of the previous layer by different convolution kernels, corresponding elements are added and biased after being accumulated, and then the convolution kernel is activated by a RELU function; the pooling layer adopts a mean pooling method, and the down-sampling scales are all 2 multiplied by 2. The number of MLP hidden layer neurons is set to be half of that of the rasterization layer, and the output layer is a single neuron and is used for real value regression. The network utilizes a BP algorithm to train, the convolution kernel weight is set to be initialized randomly, the bias is initialized to be all 0, and the optimal parameter is selected by adopting a leave-one-out cross verification method. The loss function is defined as the Euclidean distance and is calculated by equation (12)

Where yp is the network predicted value and yt is the experimentally determined value. The network learning rate is set to be 0.6 through experiments, and the maximum iteration number is 1000.

The information system software is realized by adopting a JavaEE platform and a Spring MVC + Mybaties framework.

The information system mainly comprises a water conservancy Internet of things soil moisture monitoring and water and fertilizer integrated irrigation control system, an intelligent video monitoring system, a pump room pool self-adaptive control system and a remote network meteorological monitoring system, and is a comprehensive control system integrating remote data acquisition, remote video transmission, remote monitoring control, pressure regulation and flow regulation.

The information system can comprehensively check the flow, pressure, PH value, wind speed, soil humidity, air humidity, temperature, radiation and high-level water tank liquid level real-time data of each substation, and automatically calculate the average irrigation water utilization rate of the region.

The average irrigation water utilization rate of the area is calculated according to the irrigation water consumption of different types and scales of hairs in the area and the irrigation water utilization rate of the irrigation area of the corresponding sampling points by weighted average:

in formula (13): eta_Region(s)-regional average irrigation water utilization; eta_{Big (a)}、η_In、η_Small-irrigation water utilization ratio of large, medium and small irrigation areas of the area respectively; w_{Big (a)}、W_In、W_SmallThe water consumption for the wool in the large, medium and small irrigation areas of the area is ten thousand meters³。

in the formula: eta_{Big (a)}-bundling the irrigation water utilization rate of the large irrigation areas; eta_In、η_Small-the irrigation water utilization ratio of the medium and small irrigation areas which form the bundled large irrigation area respectively; w_In、W_SmallThe amount of water used for the hairs of the medium and small irrigation areas forming the large-sized irrigation area to be bundled, m³。

The system administrator can be divided into a general administrator and a super administrator. The system administrator mainly completes acquisition and monitoring, automatic irrigation, water and fertilizer integration, video monitoring, water metering and charging management, irrigation system setting and updating, data statistics and analysis, equipment management and maintenance and user management, and a management system E-R is shown in figure 17.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims

1. Accurate regulation and control system of liquid manure based on thing networking, its characterized in that: including information perception layer, information transmission layer, supporting layer and information application layer, wherein:

2. The accurate regulation and control system of liquid manure based on thing networking of claim 1 characterized in that: the acquisition control device takes an embedded single chip microcomputer STM32F103RET6 controller as a core, realizes acquisition of water level, flow, gate level, pump station, video and gate (valve) opening and closing information, and realizes information transmission through Lora and RS232/485 modules.

3. The accurate regulation and control system of liquid manure based on thing networking of claim 1 characterized in that: the data mining mainly comprises the steps of adopting an LSTM long-short term memory neural network to predict the rainfall amount of future 5-15 days according to the effective rainfall amount P in a historical time period, and determining the irrigation time, plan and irrigation amount by combining a crop irrigation system through continuous rainless day drought evaluation and main dry land crop drought evaluation to improve the irrigation and fertilization accuracy.

TABLE 6 Dragon fruit growth period and soil moisture control threshold

TABLE 7 Kiwi fruit growth period and soil moisture control threshold

TABLE 8 growth period of tea and soil moisture control threshold

4. The accurate regulation and control system of liquid manure based on thing networking of claim 1 characterized in that: the intelligent control system comprises a BPNN-based multi-sensor data fusion such as an EC sensor, a PH sensor, a soil temperature and humidity sensor (T, H) and the like, and the structure of the intelligent control system is shown in FIG. 8. Adopting a three-layer (4-8-1) structure, namely 4 input (EC, PH, T and H), the hidden layer number is 8, the output layer number is 1 (target nutrient solution conductivity y), and establishing a BPNN multi-sensor data fusion nonlinear function as follows:

y＝f(ec,ph,t,h) (1)

wherein ec, ph, t, h are inputs of the neural network.

transfer function for fertilization bucket with buffer

V_Findicating an increased premixing volume after the addition of the venturi ejector; t is_p＝γT_rγ denotes a mixing coefficient, γ is 0 for the advection mode, γ is 1 for the ideal stirring mode,

5. The accurate regulation and control system of liquid manure based on thing networking of claim 1 characterized in that: the method comprises the steps of acquiring images of dragon fruits, kiwi fruits and tea leaves under the daylight condition, judging the water shortage state of crops by using a binocular machine vision technology, then extracting red, green and blue (RGB) three-color components, relative coefficients grb and chromaticity H of the components, adopting red components R, green components G, blue components B, R components and B components, independently learning and extracting each local feature of data through multi-layer convolution and pooling operations by using CNN, obtaining more effective abstract feature mapping relative to an explicit feature extraction method, and quickly judging indexes of the water shortage state of the dragon fruits, the kiwi fruits and the tea leaves.

6. Accurate regulation and control intelligence of liquid manure is equipped based on thing networking, its characterized in that: the intelligent water and fertilizer integrated fertilizer applicator is used for flexible secondary development according to the scale of an irrigation area, the condition of a water source, the condition of crops, irrigation habits and application scenes, and can be set into a 1-5 channel series according to requirements, and comprises a jet device, a multi-channel Venturi mixer, a general MONBUS communication protocol and a Tiny6410 embedded single board computer,

7. The Internet of things-based intelligent equipment for accurately regulating and controlling water and fertilizer, according to claim 6, is characterized in that: wherein with respect to protocol description: the self-organizing network protocol relates to a front-end data acquisition controller, a mobile phone end and a background, and comprises a collector and a background which are communicated, wherein the collector is communicated with the mobile phone (through Bluetooth), the acquired environmental parameters comprise humidity and flow (the sensors except the two sensors can be automatically increased in the following protocol), the newly added control operation comprises mode selection, manual switch control, soil moisture control threshold setting and time period setting, and the newly added reading operation comprises current working mode reading, set soil moisture control threshold reading, set time period reading and current brake (valve) actual state (switch).

TABLE 1 frame Format

TABLE 2 definition of control codes

8. The Internet of things-based intelligent equipment for accurately regulating and controlling water and fertilizer, according to claim 6, is characterized in that: the control box system is connected with an EC sensor, a PH sensor, a soil temperature and humidity sensor, a meteorological station and a video monitoring camera;

9. An accurate water and fertilizer regulation and control method based on the Internet of things is characterized in that: the method comprises the following steps:

the method comprises the following steps: initializing a system;

step two: awakening the wireless sensor node;

10. The accurate water and fertilizer regulation and control method based on the Internet of things of claim 9, characterized in that: wherein, the system also comprises information system software which is built by adopting a JavaEE platform and a Spring MVC + Mybaties framework basically,

in formula (13): eta_Region(s)-regional average irrigation water utilization; eta_{Big (a)}、η_In、η_Small-irrigation water utilization ratio of large, medium and small irrigation areas of the area respectively; w_{Big (a)}、W_In、W_SmallThe water consumption for the wool in the large, medium and small irrigation areas of the area is ten thousand meters³，

in the formula: eta_{Big (a)}-bundling the irrigation water utilization rate of the large irrigation areas; eta_In、η_Small-the irrigation water utilization ratio of the medium and small irrigation areas which form the bundled large irrigation area respectively; w_In、W_SmallMedium and small irrigation areas, respectively, forming a bundled large irrigation areaAmount of water used for wool, m³。