Machine learning based dynamic monitoring and optimization system for agricultural water efficiency

Abstract

The application relates to the technical field of data processing, in particular to an agricultural water efficiency dynamic monitoring and optimization system based on machine learning, which comprises a processor and a memory, and the processor executes the computer program of the memory to realize the following steps: acquiring historical soil moisture content data and multi-dimensional historical environment data of any to-be-monitored position at each moment in a preset historical period; acquiring the correlation degree of each environment and soil moisture content according to the data change characteristics of the historical soil moisture content data and the multi-dimensional historical environment data; when prediction is performed by using an LSTM prediction algorithm, a prediction time period is acquired, the weight of each historical soil moisture content data is acquired according to the correlation degree of each environment and soil moisture content and the difference between each multi-dimensional historical environment data in the prediction time period and multi-dimensional environment data at the current moment, and the soil moisture content data at the next moment of the current moment is predicted, so that the accuracy of soil moisture content prediction is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a dynamic monitoring and optimization system for agricultural water use efficiency based on machine learning. Background Technology

[0002] Agricultural water use efficiency refers to the efficiency of water resource utilization in agricultural production, that is, how much crop yield can be obtained per unit of water consumption. Dynamic monitoring of agricultural water use efficiency allows for timely adjustments to irrigation strategies, ensuring normal crop growth and agricultural output while achieving water conservation and sustainable development in agriculture, and reducing resource waste. Therefore, dynamic monitoring of agricultural water use efficiency is receiving increasing attention in agricultural production and has become a key evaluation standard for sustainable agricultural development.

[0003] Since crop yields can only be accurately measured during the harvest season, current technologies primarily aim to improve irrigation efficiency by maintaining soil moisture content within a suitable range during the crop's growth period. This involves monitoring soil moisture content to determine whether irrigation intensity needs to be increased or decreased. However, adjusting irrigation strategies only when soil moisture content exceeds the normal range often lacks timeliness.

[0004] Therefore, existing technologies typically use prediction algorithms (such as LSTM prediction algorithms) to predict soil moisture content based on historical data in order to adjust irrigation strategies in a timely manner. However, when LSTM prediction algorithms predict soil moisture content based on historical data, the weights of the input data at different time steps are the same, meaning that each historical data point has the same weight. Environmental changes are random; when the weather changes abruptly, the soil moisture content also changes. For example, when there is a sudden gust of wind, the soil moisture content drops rapidly. In this case, using the same weight for each historical data point fails to reflect the impact of environmental changes on soil moisture content, leading to inaccurate predictions and affecting the effectiveness of irrigation strategies.

[0005] Therefore, how to adjust the weight of historical data in the LSTM prediction algorithm and improve the accuracy of the LSTM prediction algorithm in predicting soil moisture content has become an urgent problem to be solved. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a dynamic monitoring and optimization system for agricultural water use efficiency based on machine learning, in order to solve the problem of how to adjust the weight of historical data in the LSTM prediction algorithm and improve the accuracy of the LSTM prediction algorithm in predicting soil moisture content.

[0007] This invention provides a machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps:

[0008] For any location in the field to be monitored, the soil moisture content data and multidimensional environmental data of the location at the current moment are obtained, as well as the historical soil moisture content data and multidimensional historical environmental data for each moment of a preset historical period.

[0009] Within a preset historical period, at least one time period of water content change is obtained. Based on the data change characteristics of historical soil water content data and multidimensional historical environmental data within each time period of water content change, the correlation between each environment and soil water content is obtained.

[0010] When using the LSTM prediction algorithm to predict the soil moisture content data of the next moment from the current moment, the prediction time period before the next moment from the current moment is obtained. For any moment within the prediction time period, the weight of the historical soil moisture content data at any moment is obtained based on the correlation between each environment and soil moisture content, as well as the difference between the multidimensional historical environmental data at any moment and the multidimensional environmental data at the current moment. Using the historical soil moisture content data and its weight at each moment within the prediction time period, the soil moisture content data of the next moment from the current moment is predicted, and the prediction result of the soil moisture content data of any monitoring location at the next moment from the current moment is obtained.

[0011] The prediction results for each monitoring location in the field to be monitored at the next time step are obtained from the current time step, and the irrigation strategy is adjusted accordingly.

[0012] Preferably, obtaining at least one time period of water content change within a preset historical time period includes:

[0013] A historical soil moisture content curve is constructed, where the horizontal axis of the historical soil moisture content curve represents time, and the vertical axis represents the historical soil moisture content data at each time. In the historical soil moisture content curve, the slope of the tangent line corresponding to the data point at each time is obtained. Data points with negative slopes are recorded as data points to be analyzed, and the time period corresponding to consecutive data points to be analyzed is recorded as the time period of moisture content change.

[0014] Preferably, the step of obtaining the correlation between each environment and soil moisture content based on the data change characteristics of historical soil moisture content data and multidimensional historical environmental data within each moisture content change period includes:

[0015] For any moment within any time period of moisture content change, obtain the absolute value of the difference between the historical soil moisture content data of the time and the next moment to obtain the moisture content difference; obtain the duration between the time and the next moment; calculate the ratio of the moisture content difference to the duration to obtain the moisture content change rate at any moment.

[0016] Multidimensional historical environmental data includes historical temperature data, historical humidity data, and historical wind speed data. Multidimensional historical environmental data from all moments within all water content change periods are combined into a multidimensional historical environmental data sequence. This multidimensional historical environmental data sequence is divided into three subsequences according to dimensions. For any subsequence, a two-dimensional environmental coordinate graph is constructed using the historical environmental data from the two subsequences other than the first subsequence as the x-axis and y-axis, respectively. Based on the distance differences between data points in the two-dimensional environmental coordinate graph, the degree of other environmental standards for each historical environmental data in any subsequence is obtained.

[0017] Historical soil moisture data at all times within all time periods of moisture content variation are compiled into a historical soil moisture data sequence. An environmental moisture coordinate map is constructed using historical environmental data in any subsequence as the x-axis and historical soil moisture data in the historical soil moisture data sequence as the y-axis. In the environmental moisture coordinate map, the x-axis is divided into at least two data segments according to a preset environmental data interval.

[0018] Obtain representative data points and their water content change characteristic values ​​for each data segment. Based on the degree of other environmental standards of the historical environmental data corresponding to the data points in each data segment, obtain the degree of influence of water content change on the representative data points of each data segment.

[0019] The influence of water content change on representative data points of each data segment is used as the weight of the water content change feature value of representative data points of each data segment. The weighted Pearson correlation coefficient between any subsequence and the historical soil water content data sequence is obtained as the degree of correlation between the environment and soil water content of any subsequence.

[0020] Preferably, obtaining the degree of other environmental standards for each historical environmental data in any subsequence based on the distance differences of data points in the two-dimensional environmental coordinate graph includes:

[0021] Cluster the data points in the two-dimensional environment coordinate map to obtain at least one cluster, and the cluster center of the cluster containing the most data points is recorded as the standard data point.

[0022] For any historical environmental data in any subsequence, the distance between the data point corresponding to any historical environmental data and the standard data point is obtained to obtain the distance difference. The reciprocal of the sum of the distance difference and a preset constant is obtained to obtain the other environmental standard degree of any historical environmental data.

[0023] Preferably, obtaining representative data points and their water content change characteristic values ​​for each data segment includes:

[0024] For any data segment, obtain the median of the horizontal coordinate of the data segment, and denote it as the representative value of the horizontal coordinate;

[0025] For any data point in any data segment, obtain the absolute value of the difference between the horizontal coordinate of the data point and the representative value of the horizontal coordinate to obtain the horizontal coordinate difference. Obtain the reciprocal of the sum of the horizontal coordinate difference and a preset constant to obtain the horizontal coordinate proximity. Obtain the product between the other environmental standard degree of the historical environmental data corresponding to the data point and the horizontal coordinate proximity to obtain the representativeness of the data point.

[0026] The representativeness of each data point is obtained, and the representativeness of each data point is used as the weighting coefficient of the rate of change of water content at the time corresponding to the historical environmental data of each data point. The weighted average of the rate of change of water content at the time corresponding to the historical environmental data of each data point is obtained and recorded as the characteristic value of water content change.

[0027] Using the horizontal axis as the representative value and the water content change characteristic value as the vertical axis, representative data points are formed for any data segment, and the water content change characteristic value is recorded as the water content change characteristic value of the representative data point.

[0028] Preferably, the step of obtaining the degree of influence of water content change on representative data points of each data segment based on the degree of other environmental standards of historical environmental data corresponding to data points in each data segment includes:

[0029] For any data segment, the mean of other environmental standard degrees of historical environmental data corresponding to the data points in the data segment is obtained and denoted as the mean of other environmental standard degrees. The mean of other environmental standard degrees is normalized to obtain the degree of influence of water content change on the representative data points of the data segment.

[0030] Preferably, the step of obtaining the weight of the historical soil moisture content data at any given time based on the correlation between each environmental condition and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, includes:

[0031] Based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, the feature value of the difference in moisture content change between any given time and the current time is obtained.

[0032] The difference between constant 1 and the characteristic value of the difference in water content change is obtained to obtain the weight of the historical soil water content data at any given time.

[0033] Preferably, the step of obtaining the characteristic value of the difference in water content change between any given time and the current time based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, includes:

[0034] For any dimension of the multidimensional historical environmental data at any given time, the difference between the historical environmental data of that dimension and the environmental data of the same dimension at the current time is obtained to obtain the environmental data difference.

[0035] The degree of correlation between the environment to which the historical environmental data of any dimension belongs and the soil moisture content is recorded as the degree of correlation corresponding to the historical environmental data of any dimension.

[0036] The environmental factor impact value of the historical environmental data in any dimension is obtained by multiplying the difference in the environmental data with the degree of correlation between the historical environmental data in any dimension.

[0037] The average value of the environmental factor influence value of each dimension of the historical environmental data at any given time is obtained to obtain the characteristic value of the difference in water content change between any given time and the current time.

[0038] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0039] This invention targets any monitoring location in a field to be monitored, acquiring soil moisture content data and multidimensional environmental data at the current moment for that location, as well as historical soil moisture content data and multidimensional historical environmental data for each moment within a preset historical time period. Within the preset historical time period, at least one time period of moisture content variation is acquired. Based on the data change characteristics of historical soil moisture content data and multidimensional historical environmental data within each time period of moisture content variation, the correlation between each environmental factor and soil moisture content is obtained. When using the LSTM prediction algorithm to predict the soil moisture content data for the next moment from the current moment, the prediction time prior to the next moment from the current moment is acquired. In this method, for any given moment within the prediction time period, the weight of the historical soil moisture content data at any given moment is obtained based on the correlation between each environmental factor and soil moisture content, as well as the difference between the multidimensional historical environmental data at that moment and the multidimensional environmental data at the current moment. Using the historical soil moisture content data and its weights at each moment within the prediction time period, the soil moisture content data for the next moment is predicted, resulting in a prediction result for the soil moisture content data of any monitored location at the next moment. The prediction result for each monitored location in the monitored field at the next moment is then obtained to adjust irrigation strategies. Specifically, by combining the influence of environmental factors on soil moisture content changes, the accuracy of soil moisture content prediction is improved, allowing for timely adjustment of irrigation strategies and ensuring normal crop growth. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of a method for dynamic monitoring and optimization of agricultural water use efficiency based on machine learning, provided in Embodiment 1 of the present invention. Detailed Implementation

[0042] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0043] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0044] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0045] This invention provides a machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency, including a processor and a memory. The processor executes a computer program in the memory to implement the machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency. Figure 1 As shown, the method includes the following steps:

[0046] Step S101: For any monitoring location in the field to be monitored, acquire the soil moisture content data and multidimensional environmental data of the monitoring location at the current moment, as well as the historical soil moisture content data and multidimensional historical environmental data of each moment in a preset historical period.

[0047] Agricultural water use efficiency refers to the efficiency of water resource utilization in agricultural production, that is, how much crop yield can be obtained per unit of water consumption. Dynamic monitoring of agricultural water use efficiency allows for timely adjustments to irrigation strategies, ensuring normal crop growth and agricultural output while achieving water conservation and sustainable development in agriculture, and reducing resource waste. Therefore, dynamic monitoring of agricultural water use efficiency is receiving increasing attention in agricultural production and has become a key evaluation standard for sustainable agricultural development.

[0048] In this embodiment, a representative point is selected in the field to be monitored. To reduce local interference, the representative point should be located away from irrigation points, edges, depressions, etc. TDR sensors are installed at multiple depth layers based on crop root growth depth and water infiltration depth. According to conventional agricultural deployment depths, this embodiment specifies that sensors are installed at depth layers of 10cm, 30cm, 60cm, and 100cm (i.e., four monitoring locations). Since soil moisture does not change significantly in a short period, the sensors in this embodiment monitor soil moisture content data every 15 minutes. The depth layers where the sensors are installed and the monitoring frequency can be set according to the specific implementation scenario and are not limited here.

[0049] Since the prediction method for soil moisture content data at each monitoring location in the monitored field is the same, this embodiment takes any monitoring location as an example. It acquires the soil moisture content data and multidimensional environmental data of any monitoring location at the current moment, as well as historical soil moisture content data and multidimensional historical environmental data for each moment within a preset historical time period. This data is used to predict the soil moisture content data of any monitoring location at the next moment from the current moment. Based on the predicted soil moisture content data of each monitoring location in the monitored field at the next moment from the current moment, the irrigation strategy for the monitored field is adjusted. To clearly reflect the influence of the environment on changes in soil moisture content, sufficient and diverse historical monitoring data is needed for analysis. Since the water requirements of plants vary at different growth stages, this embodiment sets the preset historical time period to one month prior to the current moment. This is not a limitation and can be set according to the specific implementation scenario.

[0050] The multidimensional environmental data includes temperature, humidity, and wind speed data, which were obtained from the meteorological platform in the area where the monitored fields are located. Because the data ranges for soil moisture content, temperature, humidity, and wind speed vary considerably, the monitored data need to be normalized before further analysis.

[0051] Existing technologies typically use prediction algorithms (such as LSTM prediction algorithms) to predict soil moisture content based on historical data, in order to adjust irrigation strategies in a timely manner. However, when LSTM prediction algorithms predict soil moisture content based on historical data, the weights of the input data at different time steps are the same, meaning that each historical data point has the same weight. Environmental changes are random; when the weather changes abruptly, the soil moisture content also changes. For example, when there is a sudden gust of wind, the soil moisture content drops rapidly. In this case, using the same weight for each historical data point fails to reflect the impact of environmental changes on soil moisture content, leading to inaccurate predictions and affecting the effectiveness of irrigation strategies.

[0052] Therefore, this embodiment first obtains the correlation between each environment and soil moisture content. Then, when using the LSTM prediction algorithm for prediction, it obtains the prediction time period. Based on the correlation between each environment and soil moisture content, and the difference between each multidimensional historical environmental data and the current multidimensional environmental data within the prediction time period, it obtains the weight of each historical soil moisture data. It then predicts the soil moisture data for the next moment from the current moment, improving the accuracy of soil moisture prediction. Finally, based on the prediction results of the soil moisture data for each monitoring location in the field to be monitored at the next moment from the current moment, it adjusts the irrigation strategy for the field to be monitored to ensure the normal growth of crops.

[0053] Step S102: Obtain at least one time period of water content change within a preset historical time period, and obtain the correlation between each environment and soil water content based on the data change characteristics of historical soil water content data and multidimensional historical environmental data within each time period of water content change.

[0054] After irrigation, not all water is absorbed into the soil for crop growth; some is used for soil evaporation and plant transpiration. Under normal circumstances, the absorption and evaporation rates of soil moisture by plants follow a certain pattern, meaning soil moisture content changes according to a predictable pattern. However, weather changes such as wind, high temperatures, and low humidity can exacerbate soil moisture evaporation and transpiration, leading to increased soil moisture consumption and disrupting the original pattern. Therefore, irrigation strategies need to be adjusted promptly based on actual soil moisture consumption. Thus, it is essential to first analyze the impact of each environmental factor on changes in soil moisture content.

[0055] During and for a period after irrigation, soil moisture content gradually increases until it stabilizes as water gradually infiltrates the soil. Subsequently, due to evaporation, transpiration, seepage, and absorption by plant roots, soil moisture content gradually decreases again. To more accurately analyze the impact of environmental factors on soil moisture content changes, a historical soil moisture content curve is constructed. The horizontal axis of this curve represents time, and the vertical axis represents the historical soil moisture content data at each time point. Only data segments showing a downward trend in the soil moisture content curve are analyzed. Specifically, the slope of the tangent line corresponding to each data point at each time point is obtained from the historical soil moisture content curve. Data points with negative slopes are designated as data points to be analyzed, and the time periods corresponding to consecutive data points to be analyzed are designated as the moisture content change time periods.

[0056] Furthermore, based on the data change characteristics of historical soil moisture content data and multidimensional historical environmental data within each moisture content change period, the correlation between each environment and soil moisture content is obtained, as follows:

[0057] (1) Obtain the rate of change of water content at each moment within each water content change period.

[0058] Specifically, for any moment within any time period of moisture content change, the absolute value of the difference between the historical soil moisture content data of the time and the next moment is obtained to obtain the moisture content difference. The duration between the time and the next moment is obtained, and the ratio of the moisture content difference to the duration is calculated to obtain the moisture content change rate at any moment.

[0059] In one embodiment, taking the i-th moment within any time period of moisture content change as an example, the formula for calculating the rate of change of moisture content at the i-th moment is:

[0060]

[0061] in, Let be the rate of change of water content at time i; This represents the historical soil moisture content data at time i. The historical soil moisture content data for the next moment after the i-th moment within a preset historical time period; Let be the duration between the i-th time point and the next time point. In this embodiment... It lasts for 15 minutes; It is the absolute value symbol.

[0062] (2) Multidimensional historical environmental data includes historical temperature data, historical humidity data and historical wind speed data. Multidimensional historical environmental data of all times within all time periods of water content change are combined into a multidimensional historical environmental data sequence. The multidimensional historical environmental data sequence is divided into three subsequences according to the dimensions. For any subsequence, other environmental standard degrees of each historical environmental data in any subsequence are obtained.

[0063] To analyze the influence of an environment on changes in soil moisture content, it is first necessary to minimize the influence of other environments. Therefore, a two-dimensional environmental coordinate map is constructed using historical environmental data from two subsequences other than the aforementioned subsequence as the x and y axes, respectively.

[0064] The data points in the two-dimensional environmental coordinate map are clustered using the DBSCAN clustering method to obtain at least one cluster. The DBSCAN clustering method is an existing technology and will not be described in detail here. The cluster center of the cluster containing the most data points is recorded as the standard data point.

[0065] For any historical environmental data in any subsequence, the distance between the data point corresponding to any historical environmental data and the standard data point is obtained to obtain the distance difference. The reciprocal of the sum of the distance difference and a preset constant is obtained to obtain the other environmental standard degree of any historical environmental data.

[0066] In one implementation, taking the j-th historical environmental data in any subsequence as an example, the formula for calculating the degree of other environmental standards for the j-th historical environmental data is as follows:

[0067]

[0068] in, Other environmental standard degrees for the j-th historical environmental data; Let j be the distance between the j-th historical environmental data point and the standard data point; As a preset constant, this embodiment sets This is used to ensure that the fraction is meaningful. There are no restrictions here, and it can be set according to the specific implementation scenario.

[0069] It should be noted that, The smaller the value, the closer the other two historical environmental data points of the j-th historical environmental data point are to the environmental data corresponding to the standard data point. The larger it is.

[0070] (3) Construct an environmental moisture content coordinate map. In the environmental moisture content coordinate map, according to the preset environmental data interval, the horizontal axis of the environmental moisture content coordinate map is divided into at least two data segments, and the representative data points of each data segment and their moisture content change characteristic values ​​are obtained.

[0071] Specifically, historical soil moisture data at all times within all time periods of moisture content variation are compiled into a historical soil moisture data sequence. An environmental moisture coordinate graph is constructed using historical environmental data from any subsequence as the x-axis and historical soil moisture data from the historical soil moisture data sequence as the y-axis. In this environmental moisture coordinate graph, the x-axis is divided into at least two data segments according to a preset environmental data interval. Since fluctuations in historical environmental data within a small range have little impact on evaporation, transpiration, and plant water absorption, the preset environmental data interval for historical temperature data is set to 3℃, for historical humidity data to 5%, and for historical wind speed data to 1m / s in this embodiment. These settings are not limited here and can be adjusted according to the specific implementation scenario.

[0072] For any data segment, obtain the median of the horizontal coordinate of the data segment, and denote it as the representative value of the horizontal coordinate;

[0073] For any data point in any data segment, obtain the absolute value of the difference between the horizontal coordinate of the data point and the representative value of the horizontal coordinate to obtain the horizontal coordinate difference. Obtain the reciprocal of the sum of the horizontal coordinate difference and a preset constant to obtain the horizontal coordinate proximity. Obtain the product between the other environmental standard degree of the historical environmental data corresponding to the data point and the horizontal coordinate proximity to obtain the representativeness of the data point.

[0074] In one implementation, taking the k-th data point of the u-th data segment as an example, the formula for calculating the representativeness of the k-th data point is:

[0075]

[0076] in, The representativeness of the k-th data point in the u-th data segment; Other environmental standard degrees for the historical environmental data corresponding to the k-th data point of the u-th data segment; Let x be the x-coordinate of the k-th data point in the u-th data segment; Let x be the x-coordinate of the u-th data segment; It is the absolute value symbol; As a preset constant, this embodiment sets This is used to ensure that the fraction is meaningful. There are no restrictions here, and it can be set according to the specific implementation scenario.

[0077] It should be noted that, The larger the value, the closer the other two historical environmental data points corresponding to the k-th data point in the u-th data segment are to the environmental data corresponding to the standard data point. Therefore, the historical environmental data corresponding to the k-th data point in the u-th data segment is more representative of the historical environmental data of the u-th data segment. The larger it is; The smaller the value, the closer the k-th data point in the u-th data segment is to the value represented by the x-axis. The larger it is.

[0078] Furthermore, the representativeness of each data point is obtained, and the representativeness of each data point is used as the weighting coefficient of the rate of change of water content at the time corresponding to the historical environmental data of each data point. The weighted average of the rate of change of water content at the time corresponding to the historical environmental data of each data point is obtained and recorded as the characteristic value of water content change.

[0079] In one embodiment, taking the u-th data segment as an example, the formula for calculating the characteristic value of the water content change in the u-th data segment is:

[0080]

[0081] in, This represents the characteristic value of water content change in the u-th data segment; Let be the rate of change of water content at the moment corresponding to the k-th data point of the historical environmental data in the u-th data segment; The representativeness of the k-th data point in the u-th data segment; Let be the number of data points in the u-th data segment.

[0082] It should be noted that, The larger the value, the greater the rate of change of water content at the k-th data point in the historical environmental data at that time. The larger the value, the more representative the k-th data point of the u-th data segment is of the water content variation characteristics of the u-th data segment. The larger it is.

[0083] Finally, the representative data point of the u-th data segment is formed by using the representative value of the horizontal axis as the horizontal axis and the characteristic value of the water content change as the vertical axis. The characteristic value of the water content change of the u-th data segment is recorded as the characteristic value of the water content change of the representative data point of the u-th data segment.

[0084] Based on the method for obtaining the u-th data segment and its characteristic value of water content change, obtain the representative data points and their characteristic values ​​of water content change for each data segment in the environmental water content coordinate graph.

[0085] (4) Based on the degree of other environmental standards of the historical environmental data corresponding to the data points in each data segment, obtain the degree of influence of water content change of the representative data points in each data segment.

[0086] Specifically, for any data segment, the mean of other environmental standard degrees of historical environmental data corresponding to the data points in the data segment is obtained and denoted as the mean of other environmental standard degrees. The mean of other environmental standard degrees is normalized to obtain the degree of influence of water content change on the representative data points of the data segment.

[0087] In one embodiment, taking a representative data point of the u-th data segment as an example, the formula for calculating the degree of influence of the change in water content of the representative data point of the u-th data segment is as follows:

[0088]

[0089] in, The degree of influence of water content change on the representative data point of the u-th data segment; Other environmental standard degrees for the historical environmental data corresponding to the k-th data point of the u-th data segment; Let be the number of data points in the u-th data segment; This is the normalization function.

[0090] Following the method for obtaining the degree of influence of water content change on representative data points of the u-th data segment, obtain the degree of influence of water content change on representative data points of each data segment.

[0091] (5) The influence of the change in water content of the representative data point of each data segment is used as the weight of the characteristic value of the change in water content of the representative data point of each data segment. The weighted Pearson correlation coefficient between any subsequence and the historical soil water content data sequence is obtained as the degree of correlation between the environment and soil water content of any subsequence. The weighted Pearson correlation coefficient is an existing technology and will not be described in detail here.

[0092] By using the method for obtaining the correlation between the environment and soil moisture content of any subsequence, the correlation between each environment and soil moisture content is obtained, thus obtaining the correlation between each environment and soil moisture content.

[0093] Step S103: When using the LSTM prediction algorithm to predict the soil moisture content data of the next moment from the current moment, the prediction time period before the next moment from the current moment is obtained. For any moment within the prediction time period, based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data of any moment and the multidimensional environmental data of the current moment, the weight of the historical soil moisture content data of any moment is obtained. Using the historical soil moisture content data and its weight at each moment within the prediction time period, the soil moisture content data of the next moment from the current moment is predicted, and the prediction result of the soil moisture content data of any monitoring location at the next moment from the current moment is obtained.

[0094] After obtaining the correlation between each environment and soil moisture content, the soil moisture content data can be predicted based on the obtained correlation between each environment and soil moisture content. When using the LSTM prediction algorithm to predict the soil moisture content data of the next moment from the current moment, since the historical soil moisture content data with a long time interval from the moment to be predicted (i.e. the moment after the current moment) has a low degree of performance on the soil moisture content data of the moment to be predicted, in this embodiment, the time before the moment to be predicted is used as the prediction time period for predicting the soil moisture content data of the moment to be predicted. There is no restriction here, and it can be set according to the specific implementation scenario.

[0095] Since the influence of the environment on soil moisture content changes is similar when environmental conditions are similar, that is, the trend of soil moisture content changes is similar, the closer the environment at the time corresponding to the historical soil moisture content data is to the current environment, the better the change pattern of soil moisture content in the historical soil moisture content data can reflect the change of soil moisture content at the time to be predicted.

[0096] Therefore, for any moment within the prediction time period, based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data at any moment and the multidimensional environmental data at the current moment, the weight of the historical soil moisture content data at any moment is obtained. Then, using the historical soil moisture content data and its weight at each moment within the prediction time period, the soil moisture content data at the moment to be predicted is predicted.

[0097] The method for obtaining the weight of the historical soil moisture content data at any given time, based on the correlation between each environmental factor and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, is as follows:

[0098] (1) Based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, obtain the feature value of the difference in moisture content change between any given time and the current time.

[0099] Specifically, for any dimension of the multidimensional historical environmental data at any given time, the difference between the historical environmental data of any dimension and the environmental data of the same dimension at the current time is obtained to obtain the environmental data difference.

[0100] The degree of correlation between the environment to which the historical environmental data of any dimension belongs and the soil moisture content is recorded as the degree of correlation corresponding to the historical environmental data of any dimension.

[0101] The environmental factor impact value of the historical environmental data in any dimension is obtained by multiplying the difference in the environmental data with the degree of correlation between the historical environmental data in any dimension.

[0102] The average value of the environmental factor influence value of each dimension of the historical environmental data at any given time is obtained to obtain the characteristic value of the difference in water content change between any given time and the current time.

[0103] In one embodiment, taking the m-th time point in the prediction time period as an example, the formula for calculating the characteristic value of the difference in water content change between the m-th time point and the current time is as follows:

[0104]

[0105] in, Let be the characteristic value representing the difference in water content change between the m-th time and the current time; For the historical temperature data in the multidimensional historical environmental data at time m; This refers to the temperature data at the current moment. The degree of correlation between temperature and soil moisture content; The historical wind speed data is from the multidimensional historical environmental data at time m. This represents the wind speed data at the current moment; The correlation between wind speed and soil moisture content; The historical humidity data is from the multidimensional historical environmental data at time m. This represents the humidity data at the current moment. The degree of correlation between humidity and soil moisture content.

[0106] It should be noted that, The larger the value, the greater the difference in environmental factors between the historical temperature data at time m and the current temperature data. In this case, if... The larger the value, the greater the correlation between the temperature at time m and the soil moisture content, indicating a greater difference between the rate of change of moisture content corresponding to the historical temperature data at time m and the rate of change of moisture content at the current time. The larger it is; The larger the value, the greater the difference in environmental factors between the historical wind speed data at time m and the wind speed data at the current time. In this case, if... The larger the value, the greater the correlation between wind speed and soil moisture content at time m, indicating a greater difference between the historical wind speed data at time m and the current rate of change in soil moisture content. The larger it is; The larger the value, the greater the difference in environmental factors between the historical humidity data at time m and the humidity data at the current time. In this case, if... The larger the value, the greater the difference between the rate of change of moisture content at the historical humidity time m and the rate of change of moisture content at the current time. The larger it is.

[0107] (2) Obtain the difference between constant 1 and the characteristic value of the difference in water content change, and obtain the weight of the historical soil water content data at any time.

[0108] In one implementation, taking the m-th time point in the prediction time period as an example, the formula for calculating the weight of the historical soil moisture content data at the m-th time point is:

[0109]

[0110] in, The weights of the historical soil moisture content data at time m are given. Let be the characteristic value representing the difference in water content change between the m-th time point and the current time point.

[0111] It should be noted that, The larger the value, the greater the difference between the rate of change of water content at time m and the rate of change of water content at the current time. This means it is less representative of the current soil moisture content change and should therefore have a smaller weight in the LSTM prediction algorithm. The smaller it is.

[0112] Similarly, the weights of historical soil moisture content data at each moment in the prediction period are obtained. When using the LSTM prediction algorithm to predict the soil moisture content data at the next moment of the current moment, the weights of historical soil moisture content data at each moment in the prediction period are used as the weights of historical soil moisture content data in the prediction period in the input gate of the LSTM prediction algorithm to predict the soil moisture content data at the next moment of the current moment, thus obtaining the prediction result of the soil moisture content data at any monitoring location at the next moment of the current moment.

[0113] Thus, the predicted soil moisture content data for any of the monitored locations at the current time is obtained.

[0114] Step S104: Obtain the prediction result of the next time for each monitoring location in the field to be monitored at the current time, and use it to adjust the irrigation strategy.

[0115] Following the method described above for obtaining the predicted soil moisture content data of any monitored location at the next time step, the predicted results for each monitored location in the monitored field at the next time step are obtained. Based on these predicted results and the root depth of crops in the monitored field, the irrigation frequency and intensity are dynamically adjusted to ensure normal crop growth. The method of dynamically adjusting the irrigation frequency and intensity based on the predicted results of each monitored location at the next time step and the root depth of crops in the monitored field is existing technology and will not be elaborated upon here.

[0116] In summary, this embodiment of the invention acquires soil moisture content data and multidimensional environmental data at any monitoring location in the field to be monitored at the current moment, as well as historical soil moisture content data and multidimensional historical environmental data for each moment in a preset historical period; within the preset historical period, it acquires at least one time period of moisture content change, and based on the data change characteristics of historical soil moisture content data and multidimensional historical environmental data within each time period of moisture content change, it acquires the correlation between each environmental factor and soil moisture content; when using the LSTM prediction algorithm to predict the soil moisture content data for the next moment from the current moment, it acquires the data before the next moment from the current moment. The prediction period involves, for any given moment within that period, determining the weight of historical soil moisture data at any given moment based on the correlation between each environmental factor and soil moisture content, and considering the difference between historical and current multidimensional environmental data at that moment. Using the historical soil moisture data and its weights at each moment within the prediction period, the soil moisture data for the next moment is predicted, resulting in a prediction of the soil moisture content at any monitored location at the next moment. This prediction result is then used to adjust irrigation strategies. Specifically, by combining the influence of environmental factors on soil moisture changes, the accuracy of soil moisture prediction is improved, allowing for timely adjustments to irrigation strategies and ensuring normal crop growth.

[0117] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: For any location in the field to be monitored, the soil moisture content data and multidimensional environmental data of the location at the current moment are obtained, as well as the historical soil moisture content data and multidimensional historical environmental data for each moment of a preset historical period. Within a preset historical period, at least one time period of water content change is obtained. Based on the data change characteristics of historical soil water content data and multidimensional historical environmental data within each time period of water content change, the correlation between each environment and soil water content is obtained. When using the LSTM prediction algorithm to predict the soil moisture content data of the next moment from the current moment, the prediction time period before the next moment from the current moment is obtained. For any moment within the prediction time period, the weight of the historical soil moisture content data at any moment is obtained based on the correlation between each environment and soil moisture content, as well as the difference between the multidimensional historical environmental data at any moment and the multidimensional environmental data at the current moment. Using the historical soil moisture content data and its weight at each moment within the prediction time period, the soil moisture content data of the next moment from the current moment is predicted, and the prediction result of the soil moisture content data of any monitoring location at the next moment from the current moment is obtained. Obtain the prediction results of each monitoring location in the field to be monitored at the next time step, and use them to adjust the irrigation strategy; The method involves obtaining the correlation between soil moisture content and environmental conditions based on the data change characteristics of historical soil moisture content data and multidimensional historical environmental data within each time period of moisture content variation, including: For any moment within any time period of moisture content change, obtain the absolute value of the difference between the historical soil moisture content data of the time and the next moment to obtain the moisture content difference; obtain the duration between the time and the next moment; calculate the ratio of the moisture content difference to the duration to obtain the moisture content change rate at any moment. Multidimensional historical environmental data includes historical temperature data, historical humidity data, and historical wind speed data. Multidimensional historical environmental data from all moments within all water content change periods are combined into a multidimensional historical environmental data sequence. This multidimensional historical environmental data sequence is divided into three subsequences according to dimensions. For any subsequence, a two-dimensional environmental coordinate graph is constructed using the historical environmental data from the two subsequences other than the first subsequence as the x-axis and y-axis, respectively. Based on the distance differences between data points in the two-dimensional environmental coordinate graph, the degree of other environmental standards for each historical environmental data in any subsequence is obtained. Historical soil moisture data at all times within all time periods of moisture content variation are compiled into a historical soil moisture data sequence. An environmental moisture coordinate map is constructed using historical environmental data in any subsequence as the x-axis and historical soil moisture data in the historical soil moisture data sequence as the y-axis. In the environmental moisture coordinate map, the x-axis is divided into at least two data segments according to a preset environmental data interval. Obtain representative data points and their water content change characteristic values ​​for each data segment. Based on the degree of other environmental standards of the historical environmental data corresponding to the data points in each data segment, obtain the degree of influence of water content change on the representative data points of each data segment. The influence of water content change on representative data points of each data segment is used as the weight of water content change feature value of representative data points of each data segment. The weighted Pearson correlation coefficient between any subsequence and the historical soil water content data sequence is obtained as the degree of correlation between the environment and soil water content of any subsequence. The step of obtaining the weight of the historical soil moisture content data at any given time based on the correlation between each environmental condition and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, includes: Based on the correlation between each environment and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, the feature value of the difference in moisture content change between any given time and the current time is obtained. The difference between constant 1 and the characteristic value of the difference in water content change is obtained to obtain the weight of the historical soil water content data at any given time. The step of obtaining the characteristic value of the difference in water content change between any given time and the current time based on the correlation between each environmental condition and soil moisture content, and the difference between the multidimensional historical environmental data at any given time and the multidimensional environmental data at the current time, includes: For any dimension of the multidimensional historical environmental data at any given time, the difference between the historical environmental data of that dimension and the environmental data of the same dimension at the current time is obtained to obtain the environmental data difference. The degree of correlation between the environment to which the historical environmental data of any dimension belongs and the soil moisture content is recorded as the degree of correlation corresponding to the historical environmental data of any dimension. The environmental factor impact value of the historical environmental data in any dimension is obtained by multiplying the difference in the environmental data with the degree of correlation between the historical environmental data in any dimension. The average value of the environmental factor influence value of each dimension of the historical environmental data at any given time is obtained to obtain the characteristic value of the difference in water content change between any given time and the current time.

2. The machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency according to claim 1, characterized in that, The step of acquiring at least one time period of water content change within a preset historical time period includes: A historical soil moisture content curve is constructed, where the horizontal axis of the historical soil moisture content curve represents time, and the vertical axis represents the historical soil moisture content data at each time. In the historical soil moisture content curve, the slope of the tangent line corresponding to the data point at each time is obtained. Data points with negative slopes are recorded as data points to be analyzed, and the time period corresponding to consecutive data points to be analyzed is recorded as the time period of moisture content change.

3. The machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency according to claim 1, characterized in that, The step of obtaining other environmental standard degrees for each historical environmental data in any subsequence based on the distance differences of data points in the two-dimensional environmental coordinate graph includes: Cluster the data points in the two-dimensional environment coordinate map to obtain at least one cluster, and the cluster center of the cluster containing the most data points is recorded as the standard data point. For any historical environmental data in any subsequence, the distance between the data point corresponding to any historical environmental data and the standard data point is obtained to obtain the distance difference. The reciprocal of the sum of the distance difference and a preset constant is obtained to obtain the other environmental standard degree of any historical environmental data.

4. The machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency according to claim 1, characterized in that, The process of obtaining representative data points and their water content change characteristic values ​​for each data segment includes: For any data segment, obtain the median of the horizontal coordinate of the data segment, and denote it as the representative value of the horizontal coordinate; For any data point in any data segment, obtain the absolute value of the difference between the horizontal coordinate of the data point and the representative value of the horizontal coordinate to obtain the horizontal coordinate difference. Obtain the reciprocal of the sum of the horizontal coordinate difference and a preset constant to obtain the horizontal coordinate proximity. Obtain the product between the other environmental standard degree of the historical environmental data corresponding to the data point and the horizontal coordinate proximity to obtain the representativeness of the data point. The representativeness of each data point is obtained, and the representativeness of each data point is used as the weighting coefficient of the rate of change of water content at the time corresponding to the historical environmental data of each data point. The weighted average of the rate of change of water content at the time corresponding to the historical environmental data of each data point is obtained and recorded as the characteristic value of water content change. Using the horizontal axis as the representative value and the water content change characteristic value as the vertical axis, representative data points are formed for any data segment, and the water content change characteristic value is recorded as the water content change characteristic value of the representative data point.

5. The machine learning-based dynamic monitoring and optimization system for agricultural water use efficiency according to claim 1, characterized in that, The step of obtaining the degree of influence of water content change on representative data points in each data segment based on the degree of other environmental standards corresponding to historical environmental data in each data segment includes: For any data segment, the mean of other environmental standard degrees of historical environmental data corresponding to the data points in the data segment is obtained and denoted as the mean of other environmental standard degrees. The mean of other environmental standard degrees is normalized to obtain the degree of influence of water content change on the representative data points of the data segment.

Images