Search

CN-121146182-B - Agricultural water efficiency dynamic monitoring and optimizing system based on machine learning

CN121146182BCN 121146182 BCN121146182 BCN 121146182BCN-121146182-B

Abstract

The invention relates to the technical field of data processing, in particular to a machine learning-based agricultural water efficiency dynamic monitoring and optimizing system which comprises a processor and a memory, wherein the processor executes a computer program of the memory to realize the steps of acquiring historical soil water content data and multidimensional historical environment data of any position to be monitored at each moment in a preset historical period, acquiring the correlation degree of each environment and soil water content according to the data change characteristics of the historical soil water content data and the multidimensional historical environment data, acquiring a prediction time period when predicting by utilizing an LSTM prediction algorithm, acquiring the weight of each historical soil water content data according to the correlation degree of each environment and soil water content and the difference of each multidimensional historical environment data and the multidimensional environment data at the current moment in the prediction time period, predicting the soil water content data at the next moment in the current moment, and improving the accuracy of predicting the soil water content.

Inventors

  • ZHANG YUANZHEN

Assignees

  • 山东欧标信息科技有限公司

Dates

Publication Date
20260512
Application Date
20250912

Claims (5)

  1. 1. The agricultural water efficiency dynamic monitoring and optimizing system based on machine learning comprises a memory, a processor and a computer program stored in the memory and running on the processor, and is characterized in that the processor realizes the following steps when executing the computer program: aiming at any position to be monitored in a field to be monitored, acquiring soil water content data and multidimensional environment data of the position to be monitored at the current moment, and presetting historical soil water content data and multidimensional historical environment data of each moment in a historical period; Acquiring at least one water content change time period in a preset historical time period, and acquiring the correlation degree of each environment and the water content of the soil according to the data change characteristics of the historical soil water content data and the multidimensional historical environment data in each water content change time period; When predicting soil moisture content data at the next time of the current time by utilizing an LSTM prediction algorithm, acquiring a prediction time period before the next time of the current time, and aiming at any time in the prediction time period, acquiring the weight of the historical soil moisture content data at any time according to the correlation degree of each environment and the soil moisture content and the difference between the multi-dimensional historical environment data at any time and the multi-dimensional environment data at the current time, and predicting the soil moisture content data at the next time of the current time by utilizing the historical soil moisture content data and the weight of the historical soil moisture content data at each time in the prediction time period to acquire a prediction result of the soil moisture content data at the next time of the current time at any position to be monitored; acquiring a prediction result of each position to be monitored in the field to be monitored at the next moment of the current moment, and adjusting an irrigation strategy; According to the data change characteristics of the historical soil water content data and the multidimensional historical environment data in each water content change time period, the method for obtaining the correlation degree between each environment and the soil water content comprises the following steps: For any moment in any water content change time period, acquiring the absolute value of the difference value between the historical soil water content data of any moment and the historical soil water content data of the next moment to obtain a water content difference value, acquiring the duration of any moment and the time of the next moment, and calculating the ratio of the water content difference value to the duration to obtain the water content change rate of any moment; the multi-dimensional historical environment data comprises historical temperature data, historical humidity data and historical wind speed data, the multi-dimensional historical environment data at all moments in all water content change time periods are formed into a multi-dimensional historical environment data sequence, the multi-dimensional historical environment data sequence is divided into three subsequences according to dimensions, for any subsequence, historical environment data in two subsequences except for the subsequence are respectively taken as an abscissa and an ordinate, a two-dimensional environment coordinate graph is constructed, and other environmental standard degrees of each historical environment data in the subsequence are obtained according to the distance difference of data points in the two-dimensional environment coordinate graph; Forming a historical soil water content data sequence by using historical soil water content data at all moments in all water content change time periods, taking historical environment data in any subsequence as an abscissa, and taking historical soil water content data in the historical soil water content data sequence as an ordinate, and constructing an environment water content coordinate graph, wherein in the environment water content coordinate graph, the abscissa of the environment water content coordinate graph is divided into at least two data segments according to preset environment data intervals; Acquiring a representative data point of each data segment and a water content variation characteristic value thereof, and acquiring the water content variation influence degree of the representative data point of each data segment according to other environmental standard degrees of historical environmental data corresponding to the data points in each data segment; Taking the influence degree of the water content change of the representative data point of each data segment as the weight of the water content change characteristic value of the representative data point of each data segment, and acquiring the weighted pearson correlation coefficient of any subsequence and the historical soil water content data sequence as the correlation degree of the environment to which the any subsequence belongs and the soil water content; according to the correlation degree of each environment and the soil moisture content and the difference between the multidimensional historical environment data at any moment and the multidimensional environment data at the current moment, the weight of the historical soil moisture content data at any moment is obtained, and the weight comprises the following steps: acquiring a water content change difference characteristic value of any moment and the current moment according to the correlation degree of each environment and the water content of the soil and the difference between the multi-dimensional historical environment data of any moment and the multi-dimensional environment data of the current moment; Obtaining a difference value between a constant 1 and the water content change difference characteristic value to obtain the weight of the historical soil water content data at any moment; According to the correlation degree of each environment and the soil water content and the difference between the multidimensional historical environment data at any moment and the multidimensional environment data at the current moment, the method for obtaining the water content change difference characteristic value between any moment and the current moment comprises the following steps: acquiring a difference value of the arbitrary dimension historical environmental data and the environmental data of the same dimension at the current moment aiming at the arbitrary dimension historical environmental data in the multidimensional historical environmental data at the arbitrary moment to obtain an environmental data difference value; Recording the correlation degree of the environment to which the history environmental data of any dimension belongs and the water content of the soil as the correlation degree corresponding to the history environmental data of any dimension; obtaining a product of the environmental data difference value and the correlation degree corresponding to the history environmental data of any dimension to obtain an environmental factor influence value of the history environmental data of any dimension; And obtaining an average value of environmental factor influence values of each dimension of the multi-dimension historical environmental data at any moment to obtain a water content change difference characteristic value at any moment and the current moment.
  2. 2. The machine learning based agricultural water efficiency dynamic monitoring and optimization system of claim 1, wherein the acquiring at least one water content variation time period within a preset history period comprises: And constructing a historical soil moisture content curve graph, wherein the abscissa of the historical soil moisture content curve graph is moment, the ordinate is historical soil moisture content data of each moment, in the historical soil moisture content curve graph, the slope of a tangent line of a data point corresponding to each moment is obtained, the data point with the negative slope is recorded as a data point to be analyzed, and the time period corresponding to the continuous data point to be analyzed is recorded as a moisture content change time period.
  3. 3. The machine learning based agricultural water efficiency dynamic monitoring and optimization system according to claim 1, wherein the obtaining the other environmental standard degree of each historical environmental data in any one of the subsequences according to the difference of the distances between the data points in the two-dimensional environmental coordinate graph comprises: clustering the data points in the two-dimensional environmental coordinate graph to obtain at least one class cluster, and marking the class cluster center of the class cluster containing the most data points as a standard data point; And aiming at any historical environmental data in any subsequence, acquiring the distance between a data point corresponding to any historical environmental data and the standard data point to obtain a distance difference value, and acquiring the reciprocal of the addition result of the distance difference value and a preset constant to obtain other environmental standard degrees of any historical environmental data.
  4. 4. The machine learning based agricultural water efficiency dynamic monitoring and optimization system of claim 1, wherein the obtaining representative data points for each data segment and their water content variation characteristic values comprises: for any data segment, acquiring a median value of an abscissa where the any data segment is located, and marking the median value as an abscissa representative value; For any data point in any data segment, acquiring the absolute value of the difference between the abscissa of the any data point and the abscissa representative value to obtain an abscissa difference value, acquiring the reciprocal of the addition result of the abscissa difference value and a preset constant to obtain an abscissa approach degree, and acquiring the product between the other environmental standard degrees of the historical environmental data corresponding to the any data point and the abscissa approach degree to obtain the representative degree of the any data point; Obtaining the representative degree of each data point, taking the representative degree of each data point as a weight coefficient of the water content change rate of the historical environment data corresponding to each data point at the moment, obtaining a weighted average value of the water content change rate of the historical environment data corresponding to each data point at the moment, and recording the weighted average value as a water content change characteristic value; And forming a representative data point of any data segment by taking the abscissa representing value as an abscissa and the water content change characteristic value as an ordinate, and marking the water content change characteristic value as the water content change characteristic value of the representative data point.
  5. 5. The machine learning based agricultural water efficiency dynamic monitoring and optimization system according to claim 1, wherein the obtaining the water content variation influence degree of the representative data point of each data segment according to the other environmental standard degree of the historical environmental data corresponding to the data point of each data segment comprises: And aiming at any data segment, acquiring the average value of other environmental standard degrees of the historical environmental data corresponding to the data points in any data segment, recording the average value as the average value of other environmental standard degrees, and carrying out normalization processing on the average value of other environmental standard degrees to obtain the influence degree of the water content change of the representative data point of any data segment.

Description

Agricultural water efficiency dynamic monitoring and optimizing system based on machine learning Technical Field The invention relates to the technical field of data processing, in particular to a dynamic monitoring and optimizing system for agricultural water efficiency based on machine learning. Background The agricultural water efficiency refers to the water resource utilization efficiency in agricultural production, i.e., how much crop yield can be obtained per unit water consumption. The irrigation strategy can be timely adjusted through dynamic monitoring of the agricultural water consumption efficiency, so that normal growth of crops is guaranteed, agricultural output is guaranteed, agricultural water saving and sustainable development are realized, and resource waste is reduced. Therefore, dynamic monitoring of agricultural water efficiency is becoming more and more important in the agricultural production process, and becomes a key evaluation standard for sustainable development of agriculture. Because crop yield can only accurately measure in the crop harvesting period, in order to reasonably and efficiently irrigate during the crop growth period, the prior art mainly improves the irrigation efficiency by maintaining the water content of soil in a proper water-containing interval, namely, whether the water content of the soil is in a set range or not is monitored, and whether the irrigation intensity is required to be increased or reduced is judged. However, the irrigation strategy is adjusted when the water content of the soil is detected to be out of the normal range, and the timeliness is often insufficient. Therefore, prediction algorithms (e.g., LSTM prediction algorithms) are commonly used in the art to predict soil moisture content based on historical data, and to adjust irrigation strategies in time. However, when the LSTM prediction algorithm predicts the moisture content of the soil according to the historical data, the weights of the input data at different time steps are the same, that is, the weight of each historical data is the same, the environmental change is random, when the weather is abrupt, the moisture content of the soil also changes, for example, the moisture content of the soil rapidly drops when the weather is suddenly blown, at this time, each historical data uses the same weight, and the influence of the environmental change on the moisture content of the soil cannot be reflected, so that the prediction result of the moisture content of the soil is inaccurate, and the effectiveness of the irrigation strategy is affected. Therefore, how to adjust the weight of the historical data in the LSTM prediction algorithm, and to improve the accuracy of the LSTM prediction algorithm on the soil moisture content prediction becomes a problem to be solved. Disclosure of Invention In view of the above, the embodiment of the invention provides a machine learning-based agricultural water efficiency dynamic monitoring and optimizing system, which is used for solving the problem of how to adjust the weight of historical data in an LSTM prediction algorithm and improving the accuracy of the LSTM prediction algorithm on soil water content prediction. The embodiment of the invention provides a dynamic monitoring and optimizing system for agricultural water efficiency based on machine learning, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor realizes the following steps when executing the computer program: aiming at any position to be monitored in a field to be monitored, acquiring soil water content data and multidimensional environment data of the position to be monitored at the current moment, and presetting historical soil water content data and multidimensional historical environment data of each moment in a historical period; Acquiring at least one water content change time period in a preset historical time period, and acquiring the correlation degree of each environment and the water content of the soil according to the data change characteristics of the historical soil water content data and the multidimensional historical environment data in each water content change time period; When predicting soil moisture content data at the next time of the current time by utilizing an LSTM prediction algorithm, acquiring a prediction time period before the next time of the current time, and aiming at any time in the prediction time period, acquiring the weight of the historical soil moisture content data at any time according to the correlation degree of each environment and the soil moisture content and the difference between the multi-dimensional historical environment data at any time and the multi-dimensional environment data at the current time, and predicting the soil moisture content data at the next time of the current time by utilizing the historical soil moisture content data and the weight of the h