CN-121979024-A - Intelligent control method of wheat combine harvester

CN121979024ACN 121979024 ACN121979024 ACN 121979024ACN-121979024-A

Abstract

According to the intelligent control method of the wheat combine harvester, sensor detection data are used as input, and an LSTM neural network prediction model is used for obtaining operation quality prediction values of a period of time in the future, so that harvester operation performance parameters such as grain impurity rate, loss rate, breakage rate and yield are used as evaluation indexes, simulation and prediction results are made through historical time sequence data, a control instruction is generated according to the predicted operation performance parameters and parameter thresholds of the operation performance parameters, an optimal control decision result aiming at an expected state of an actuator is output, accuracy of the parameter prediction of the combine harvester is remarkably improved, more accurate control parameters of the combine harvester can be obtained, and the requirement of the combine harvester for realizing automatic intelligent control is met.

Inventors

WU YUANZE
LI KUN
Du Shucan
JI JIANGTAO
SUN JINGWEI
Zhao kaixuan
LI QIANWEN
LIU DUO
WANG HAIYUAN
WANG CHAOYANG
WEN WENLIANG
BAI JIAYI

Assignees

河南科技大学

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (3)

1. An intelligent control method of a wheat combine harvester is characterized by comprising the following steps: s1, arranging a plurality of sensors on a wheat combine harvester, wherein the sensors comprise a header rotating speed sensor, a roller rotating speed sensor, a fan rotating speed sensor, a grain yield sensor, a grain loss sensor and a grain impurity sensor, and detecting the header rotating speed, the roller rotating speed, the fan rotating speed, the grain yield, the grain loss and grain impurity data in the operation process of the wheat combine harvester; S2, filtering the raw data acquired by the sensor by adopting a Gaussian weighted moving average filtering algorithm, and firstly calculating a one-dimensional form of a Gaussian function as a weight value of a signal: ; And then carrying out normalization treatment: ; noise is eliminated by weighted averaging the signals, and a filtering result y is output: ; In the above formula, x is the distance from the position in the window to the center of the window, sigma is the standard deviation of the function, G (x) is the Gaussian weight value corresponding to the signal at the position x relative to the center of the window in the filter window, w i is the normalized weight value corresponding to the signal at the ith position in the filter window, N is the window length; S3, sending the filtered data into an LSTM neural network prediction model, preprocessing the data obtained in a period of time by the prediction model, and preprocessing historical time series data As input to the input layer, the hidden state of each time step is output And outputs the upper layer As input to the next layer, the output of the last layer is Will be Mapping to predicted output dimensions Calculating a predicted value of future job quality : ; In the above formula, W out is a weight matrix of an LSTM neural network output layer, h T is a hidden state of the LSTM neural network in the last time step T, and b out is a bias vector of the LSTM neural network output layer for adjusting a base line of a prediction result; The LSTM neural network comprises an input layer, a plurality of hidden layers and an output layer, wherein a basic unit is a memory cell, and the memory cell is updated by controlling forgetting, inputting and outputting of memory information through three gating; the forgetting gate integrates the current input value and the output value at the last moment, then compresses the input to the interval of (0, 1) through a sigmoid function, and forgets the information of the component if the vector becomes 0 after passing through the sigmoid layer; the input gate extracts effective information in the vector through a tanh function, and then controls whether the memory cell enters a unit state or not through a sigmoid function; The output gate integrates the current input value with the output value at the previous moment, extracts information through a sigmoid function, and then compresses and maps the current unit state into a section (-1, 1) through a tanh function to obtain the output value at the current moment; s4, inputting a predicted value of future operation quality into a control decision algorithm to obtain an initial control strategy; the control decision algorithm is a particle swarm optimization algorithm, and the objective function is expressed as: ; In the above-mentioned method, the step of, Is the actual working performance of the wheat combine at the ith moment, comprises grain loss and grain impurity, N is the number of data points for the target operation performance after correction according to the predicted value of the future operation quality; The particle swarm optimization algorithm adjusts the position of particles through the solving process of the objective function in the solution space, so as to obtain optimal control parameters; And S5, the wheat combine harvester continues to operate according to the optimal control parameters.
2. The intelligent control method of the wheat combine harvester according to claim 1, wherein the original data of the S1 are transmitted to the edge computing device through the CAN bus, the edge computing device executes the S2, the filtered data are transmitted to the cloud server, the cloud server executes the S3 and the S4, the optimal control parameters are sent to the edge computing device through the network communication unit, the edge computing device receives and analyzes the decision instruction, the control instruction is sent to the corresponding actuator through the CAN bus, and the actuator receives and executes the control instruction from the CAN bus.
3. The intelligent control method of a wheat combine according to claim 1, wherein the input gate is defined as i t , the forgetting gate is defined as f t , the output gate is defined as o t , the memory cell is updated as C t , and the three gating and memory cell updating formulas are as follows: ; ; ; ; ; In the above, h t-1 is the hidden state of the LSTM neural network in the previous time step, x t is the original feature vector of the LSTM neural network in the current time step, [ h t-1 , x t ] represents vector splicing operation as the input of gating calculation, tanh represents hyperbolic tangent activation function, W i 、W c 、W f 、W o is a learnable weight matrix corresponding to an input gate, a candidate cell state, a forgetting gate and an output gate respectively, and b i 、b c 、b f 、b o is a learnable bias vector corresponding to the input gate, the candidate cell state, the forgetting gate and the output gate respectively.

Description

Intelligent control method of wheat combine harvester Technical Field The invention relates to the technical field of intelligent control of combine harvesters, in particular to an intelligent control method of a wheat combine harvester. Background The loss rate and the impurity rate are key indexes for evaluating the working quality of the wheat harvester. The harvester operator adjusts the harvesting parameters by experience, has the defects of delayed decision, inaccurate control and the like, and is difficult to meet the production requirement of real-time decision control. The new generation electronic information technology provides a technical basis for intelligent operation of the wheat harvesting process, such as the internet of things technology, information interaction among users (people), harvesters (machines), sensors (objects) and systems (clouds) is realized, and real-time interoperation among harvester operation data, models and harvesters can be further realized. However, the prior art still has the defects of the method for predicting and automatically controlling the parameters of the combine harvester, the prediction accuracy is lower, and the control parameter setting of the combine harvester is also inaccurate, so that the requirement of the combine harvester for realizing automatic intelligent control cannot be met. Disclosure of Invention The invention aims to provide an intelligent control method of a wheat combine harvester, which improves the accuracy of the parameter prediction of the combine harvester and obtains more accurate control parameters of the combine harvester. The invention adopts the technical scheme that the intelligent control method of the wheat combine harvester comprises the following steps: s1, arranging a plurality of sensors on a wheat combine harvester, wherein the sensors comprise a header rotating speed sensor, a roller rotating speed sensor, a fan rotating speed sensor, a grain yield sensor, a grain loss sensor and a grain impurity sensor, and detecting the header rotating speed, the roller rotating speed, the fan rotating speed, the grain yield, the grain loss and grain impurity data in the operation process of the wheat combine harvester; S2, filtering the raw data acquired by the sensor by adopting a Gaussian weighted moving average filtering algorithm, and firstly calculating a one-dimensional form of a Gaussian function as a weight value of a signal: ; And then carrying out normalization treatment: ; noise is eliminated by weighted averaging the signals, and a filtering result y is output: ; In the above formula, x is the distance from the position in the window to the center of the window, sigma is the standard deviation of the function, G (x) is the Gaussian weight value corresponding to the signal at the position x relative to the center of the window in the filter window, w i is the normalized weight value corresponding to the signal at the ith position in the filter window, N is the window length; S3, sending the filtered data into an LSTM neural network prediction model, preprocessing the data obtained in a period of time by the prediction model, and preprocessing historical time series data As input to the input layer, the hidden state of each time step is outputAnd outputs the upper layerAs input to the next layer, the output of the last layer isWill beMapping to predicted output dimensionsCalculating a predicted value of future job quality: ; In the above formula, W out is a weight matrix of an LSTM neural network output layer, h T is a hidden state of the LSTM neural network in the last time step T, and b out is a bias vector of the LSTM neural network output layer for adjusting a base line of a prediction result; The LSTM neural network comprises an input layer, a plurality of hidden layers and an output layer, wherein a basic unit is a memory cell, and the memory cell is updated by controlling forgetting, inputting and outputting of memory information through three gating; the forgetting gate integrates the current input value and the output value at the last moment, then compresses the input to the interval of (0, 1) through a sigmoid function, and forgets the information of the component if the vector becomes 0 after passing through the sigmoid layer; the input gate extracts effective information in the vector through a tanh function, and then controls whether the memory cell enters a unit state or not through a sigmoid function; The output gate integrates the current input value with the output value at the previous moment, extracts information through a sigmoid function, and then compresses and maps the current unit state into a section (-1, 1) through a tanh function to obtain the output value at the current moment; s4, inputting a predicted value of future operation quality into a control decision algorithm to obtain an initial control strategy; the control decision algorithm is a particle swarm optimization algorithm, and the obje