Intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation

Abstract

The invention discloses an intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation. The method comprises the following steps: firstly, acquiring historical day-by-day weather data of a farmland area, calculating the evapotranspiration of a reference crop, and preprocessing the evapotranspiration; secondly, weather factors are screened to serve as input features, reference crop evapotranspiration serves as a target variable, and a reference crop evapotranspiration estimation model is constructed and trained; and then, through a reference crop evapotranspiration estimation model, obtaining reference crop evapotranspiration estimation results of the current day and the future n days. And finally, according to an evapotranspiration estimation result in combination with a crop coefficient corresponding to the growth stage of the rice, evapotranspiration of the current day and n days in the future is calculated, an irrigation decision is made, and the water storage capacity of the reservoir is adjusted based on the irrigation water consumption of the current day and the irrigation water consumption of the n days in the future. The invention provides an efficient, economic and sustainable technical scheme for intelligent irrigation management of large-scale farmland, and has important value for promoting agricultural precision and ecological development.

Description

Technical Field

[0001] The invention belongs to the technical field of agricultural irrigation, and in particular relates to an intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation. Background Art

[0002] Globally, crop irrigation is the largest consumer of freshwater. Therefore, the control of irrigation water volume is particularly important, but traditional strategies are almost impossible to achieve precise control. It is impossible to carry out precise irrigation operations and save water based on the actual evapotranspiration of crops at different growth stages. The precision irrigation system can accurately calculate the water requirements of crops at different growth stages, and accurately control the irrigation time, water volume and optimize the irrigation strategy accordingly. To further achieve precision irrigation, the irrigation strategy can be combined with the water balance equation to accurately calculate the irrigation water requirements of crops. In precision irrigation operations, water conservation requires accurate estimation of the main elements of the water balance equation, especially evapotranspiration. The value of evapotranspiration can be obtained through direct measurement technology, including lysimeter methods, etc., but the cost of these methods is too expensive, and estimating evapotranspiration through mathematical models will face the problem of missing weather data, which is very difficult. In addition, due to water shortage or insufficient water supply in mountainous and arid areas, water needs to be stored and pumped from the reservoir before irrigation, and it is difficult to determine the amount of stored water. To this end, we proposed an intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation. Summary of the invention

[0003] In response to the above problems, the present invention proposes an intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation. This method uses historical daily weather data (average temperature, air pressure, average wind speed, net solar radiation, sunshine time, relative humidity, precipitation, etc.) and reference crop evapotranspiration to screen key weather factors to train the CNN-LSTM-Attention reference crop evapotranspiration estimation model, calculates the daily evapotranspiration of rice in combination with the crop coefficient, constructs a water balance equation based on key weather factors of the day and the next 5 days, and makes irrigation decisions in combination with irrigation strategies, calculates the irrigation water consumption of farmland, and then arranges irrigation and water storage.

[0004] The intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation specifically comprises the following steps: S1: Obtain historical daily weather data in the farmland area to calculate reference crop evapotranspiration and perform preprocessing.

[0005] S2: Based on the weather data and reference crop evapotranspiration in S1, key weather factors are selected as input features, reference crop evapotranspiration is used as the target variable, and a CNN-LSTM-Attention reference crop evapotranspiration estimation model is constructed and trained.

[0006] S3: Obtain the key weather factors of the current day and the next n days as input features and input them into the trained reference crop evapotranspiration estimation model to obtain the reference crop evapotranspiration estimation results of the current day and the next n days respectively.

[0007] S4: Calculate the evapotranspiration for the current day and the next n days based on the reference crop evapotranspiration estimation results for the current day and the next n days in S3 combined with the crop coefficient corresponding to the growth stage of rice.

[0008] S5: construct a water balance equation based on the evapotranspiration of the day in S4 and make irrigation decisions in combination with the irrigation strategy, and calculate the irrigation water consumption of the day in combination with the farmland area to arrange irrigation for the next day.

[0009] S6: Based on the evapotranspiration in the next n days in S4, the irrigation water consumption in the next n days is calculated in the same way as S5, and water is stored in the reservoir in advance.

[0010] S7: Repeat S3 to S6 every day, and adjust the water storage capacity of the reservoir based on the irrigation water consumption of the day and the irrigation water consumption of the next n days.

[0011] In S1, historical daily weather data in the farmland area is obtained, including daily average temperature, air pressure, average wind speed, net solar radiation, sunshine time, relative humidity, precipitation and other weather data, and the data set is divided into a training set and a test set; the reference crop evapotranspiration is calculated using the Penman-Monteith (PM) formula and used as the model target variable.

[0012] The preprocessing adopts the maximum-minimum standardization method to map the values ​​of all features and target variables to the range of [0,1]. The formula of the maximum-minimum standardization method is:

[0013] is the standardized data, is the original data, is the minimum value in the data, is the maximum value in the data.

[0014] In S2, based on the weather data and reference crop evapotranspiration in S1, key weather factors are selected as input features. The correlation strength between different weather data and reference crop evapotranspiration needs to be quantified through grey correlation analysis. The analysis process selects reference crop evapotranspiration as the reference sequence, and a comparison sequence composed of multiple weather data (including air pressure, average temperature, precipitation, relative humidity, sunshine time, average wind speed, etc.) is selected. The grey correlation formula is:

[0015] in, is the data of the reference sequence, To compare sequence data, is the degree of association between the reference sequence and the comparison sequence, ρ is the resolution coefficient, which is taken as 0.5.

[0016] Weather data with high correlation are selected as key weather factors, and the Z weather data with the highest correlation are selected as input features of the reference crop evapotranspiration estimation model; the reference crop evapotranspiration estimation model integrates the local feature extraction capability of the convolutional network CNN, the long-term dependency processing capability of the long short-term memory network LSTM, and the weight allocation capability of the attention mechanism Attention. The reference crop evapotranspiration estimation model first extracts the local features of the input sequence through a one-dimensional convolutional layer (Conv1D), and then uses a pooling layer (MaxPooling) to reduce the feature dimension and retain the significant information in the sequence. Then, the local features are input into the LSTM layer to process the long-term dependency relationship in the long sequence data and the generalization ability of the model is improved through the Dropout regularization technology. Finally, the key features are extracted through the Attention layer, and a fully connected layer is used as the output layer to output the reference crop evapotranspiration estimation value.

[0017] The key weather factors of the day in S3 can be obtained using the local weather station, and the key weather factors of the next n days can be obtained using the weather forecast, and the corresponding reference crop evapotranspiration estimation results are obtained through the model.

[0018] The rice growth stage in S4 divides the entire growth period of rice into greening stage, tillering stage, jointing and booting stage, heading and flowering stage, milky stage, and yellow stage; the crop coefficient corresponding to the growth stage is obtained by consulting materials (such as FAO-56) or localized field test data; the rice evapotranspiration on the same day is obtained by the reference crop evapotranspiration on the same day in S3 and the crop coefficient corresponding to the growth stage on the same day, and the evapotranspiration in the next n days is obtained by the reference crop evapotranspiration in the next n days in S3 and the crop coefficient corresponding to the growth stage in the next n days. The growth stage in the next n days is roughly determined by the growth stage in the same period in history. Evapotranspiration is calculated as:

[0019] In the formula is crop evapotranspiration (unit: mm / d); is the crop coefficient, and the corresponding constant value is selected according to the different growth stages of rice; is the reference crop evapotranspiration (unit: mm / d).

[0020] In S5, a water balance equation is constructed based on the daily evapotranspiration and an irrigation decision is made in combination with the irrigation strategy. The water balance equation is:

[0021] In the formula , are the depth of field surface water layer at the end of day t and day t-1, respectively; The amount of irrigation water; is the rainfall; is the leakage amount on day t; For displacement.

[0022] The irrigation strategy controls the depth of the water layer in the field and keeps the soil in a shallow water layer or moist state to reduce water consumption and improve water use efficiency. In practical applications, the water layer depth at the end of each day is calculated according to the water balance equation. If it is less than the lower limit of the suitable water layer height or the lower limit of the water content, irrigation will be arranged at the beginning of the next day to irrigate to the upper limit of the suitable water layer height. From the water balance equation, the water layer height calculation formula is:

[0023] When heavy rain causes the water level to rise, drainage is required. Greater than the allowable water storage capacity When the water is discharged to the aquifer, the water is discharged to the aquifer. The drainage volume calculation formula is:

[0024] The calculation formula for irrigation water volume is:

[0025] In the formula, The upper limit of the suitable water layer is is the lower limit of the suitable water layer. is the lower limit of water content, is the planned wetting depth on day t, To allow for water storage depth, is the saturated water content, is the soil moisture content on day t.

[0026] The irrigation water consumption is determined by making an irrigation decision based on the irrigation strategy, that is, judging whether irrigation is needed, and then calculating the irrigation water consumption when irrigation is needed based on the evapotranspiration and the area of ​​farmland to be irrigated. The irrigation water consumption calculation formula is:

[0027] In the formula, is the irrigation area (㎡), Irrigation water volume (mm), Irrigation water consumption (m³).

[0028] Preferably, high-precision liquid level sensors are installed in the farmland to monitor the water layer height in real time during irrigation. When the water layer height reaches the upper limit of the suitable water layer during irrigation, irrigation is stopped to reduce the impact of sudden situations such as irrigation water volume errors and water loss in water pipelines caused by calculation errors in the reference crop evapotranspiration estimation model.

[0029] In S6, based on the evapotranspiration of the next n days, the irrigation water consumption of the next n days is calculated in the same way as in S5, and water is stored in the reservoir in advance. The evapotranspiration of the next n days is calculated, and irrigation decisions are made daily in combination with the irrigation strategy to calculate the daily irrigation water consumption; water is stored in the reservoir in advance according to the total irrigation water consumption of the current day and the next n days.

[0030] The water storage capacity of the reservoir is adjusted in S7 based on the irrigation water consumption of the day and the irrigation water consumption of the next n days, the irrigation water consumption of the day is calculated every day, and the irrigation water consumption of the next n days is updated and calculated; according to the water storage capacity of the reservoir on the day and the total irrigation water consumption of the next n days, the water storage is replenished in time to avoid water shortage or untimely water supply in mountainous and arid areas; the water tank is installed with a liquid level sensor, and the water level of the reservoir is calibrated before irrigation. When the water volume in the reservoir is less than the total irrigation water consumption, water is replenished to avoid the influence of sudden situations such as evaporation of the reservoir and accidental water loss of the reservoir. The water storage capacity calculation formula of the reservoir is:

[0031] In the formula, is the water storage capacity of the reservoir, is the reservoir area, is the water level of the reservoir.

[0032] The beneficial effect of the present invention is that the present invention proposes an intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation. The CNN-LSTM-Attention reference crop evapotranspiration estimation model integrates the local feature extraction capability of CNN, the time series modeling advantage of LSTM and the key feature screening function of the attention mechanism, and combines the grey correlation analysis to screen key weather factors, which significantly improves the accuracy of evapotranspiration estimation and reduces the dependence on traditional high-cost physical measurement equipment or complete weather data. By dynamically associating the crop coefficients and water balance equations of each growth stage of rice, real-time monitoring of changes in field water layer depth, and intelligent triggering of irrigation or drainage decisions, it is possible to accurately control the irrigation water volume on demand, effectively optimize the utilization rate of water resources, and save a lot of water compared to traditional irrigation. In addition, adjusting the water storage capacity of the reservoir based on the irrigation water consumption of the day and the irrigation water consumption of the next n days avoids the problem of untimely irrigation in mountainous and arid areas due to water shortage or insufficient water supply. The present invention provides feasible guidance for achieving precise irrigation, avoiding the unnecessary waste of water resources due to excessive irrigation, and the adverse effects on crop growth due to insufficient irrigation; it provides an efficient, economical and sustainable technical solution for the intelligent irrigation management of large-scale farmland, which is of great value in promoting the precise and ecological development of agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of the intelligent irrigation dynamic decision-making method based on the reference crop evapotranspiration estimation; Figure 2 The calculation results of reference crop evapotranspiration; Figure 3 This is the structure diagram of the CNN-LSTM-Attention model; Figure 4 This is a map of the estimated crop evapotranspiration for reference in 2022 and 2023; Figure 5 This is a graph showing changes in irrigation water volume and water layer height in 2022; Figure 6 This is a graph showing changes in irrigation water volume and water layer height in 2023; Figure 7 This is a comparison chart of water consumption between the irrigation strategy of the present invention and the traditional irrigation strategy. DETAILED DESCRIPTION

[0034] In order to more clearly illustrate the purpose, technical solution and advantages of the embodiment of the present invention, the technical solution of the embodiment of the present invention will be further clearly and completely described in conjunction with the accompanying drawings of the embodiment of the present invention. The described embodiment is a part of the embodiment of the present invention, not all of the embodiments.

[0035] like Figure 1As shown, a smart irrigation dynamic decision-making method based on reference crop evapotranspiration estimation specifically includes the following steps: S1: Obtain historical daily weather data in the farmland area to calculate reference crop evapotranspiration and perform preprocessing.

[0036] The historical daily weather data of Shaoxing City was obtained, including daily average temperature, air pressure, average wind speed, net solar radiation, sunshine time, relative humidity, precipitation and other weather data from January 1, 2013 to December 31, 2023. The data set was divided into a training set (daily weather data from 2013 to 2021) and a test set (daily weather data from 2022 to 2023); the reference crop evapotranspiration was calculated using the Penman (PM) formula and used as the model target variable. The calculation results of the reference crop evapotranspiration are as follows: Figure 2 As shown. Penman's formula is:

[0037] in, is the net solar radiation (unit: MJ / m² / day), is the soil heat flux (unit: MJ / m² / day, usually negligible in the short term); is the average wind speed at 2m (unit: m / s), is the psychrometric constant (usually taken as 0.066), is the slope of the saturated water vapor pressure curve (unit: kPa / °C), is the saturated water vapor pressure (unit: kPa), is the actual water vapor pressure (unit: kPa), is the average temperature (unit: ℃).

[0038] The preprocessing adopts the maximum-minimum standardization method to map the values ​​of all features and target variables to the range of [0,1]. The formula of the maximum-minimum standardization method is:

[0039] is the standardized data, is the original data, is the minimum value in the data, is the maximum value in the data.

[0040] S2: Based on the weather data and reference crop evapotranspiration in S1, key weather factors are selected as input features, reference crop evapotranspiration is used as the target variable, and a CNN-LSTM-Attention reference crop evapotranspiration estimation model is constructed and trained.

[0041] The screening of key weather factors mentioned above requires quantifying the correlation strength between different weather data and reference crop evapotranspiration through grey correlation analysis. The analysis process selects reference crop evapotranspiration as the reference sequence and a comparison sequence consisting of multiple weather data (including air pressure, average temperature, precipitation, relative humidity, sunshine time, average wind speed, etc.). The grey correlation formula is:

[0042] in, is the data of the reference sequence, To compare sequence data, is the correlation between the reference sequence and the comparison sequence, and ρ is the resolution coefficient, which is taken as 0.5.

[0043] The results of grey correlation analysis are shown in Table 1. Weather data with higher correlation were selected as key weather factors. The influencing factors were ranked as sunshine time > average temperature > average wind speed > precipitation > air pressure > relative humidity. Therefore, sunshine time, average temperature and average wind speed were taken as key weather factors and used as input features of the CNN-LSTM-Attention reference crop evapotranspiration estimation model.

[0044] Table 1 Correlation coefficient

[0045] The CNN-LSTM-Attention reference crop evapotranspiration estimation model combines the local feature extraction capability of CNN, the long-term dependency processing capability of LSTM, and the weight allocation capability of the attention mechanism Attention. The first layer of the model is a one-dimensional convolution layer (Conv1D) to extract local features of the input sequence, followed by a pooling layer (MaxPooling) to reduce the feature dimension and retain significant information in the sequence. The data is then input into the LSTM layer to process the long-term dependencies in the long sequence data and the Dropout regularization technique is used to improve the generalization ability of the model. Finally, the key features are extracted through the Attention layer, and a fully connected layer is used as the output layer. See the CNN-LSTM-Attention model structure diagram for details. Figure 3 .

[0046] In this embodiment, the CNN-LSTM-Attention model adopts the Pytorch deep learning framework, and the main parameters are set as follows: the number of convolution kernels in the convolution layer is 32, the convolution kernel size is 3, and the pooling window size of the pooling layer is 1; the number of neurons in the LSTM layer is 32; the number of training rounds is 100, the batch size is 32, the optimizer is Adam, the learning rate is 0.01, and the dropout rate is 0.2.

[0047] In order to verify the effectiveness of the CNN-LSTM-Attention prediction model proposed in this paper, comparative experiments were conducted with the RNN prediction model, LSTM prediction model, CNN-LSTM prediction model, and CNN-LSTM-Attention model in the same data set with the same training set and validation set, and the prediction effect of the comparative model was analyzed based on the evaluation indicators of each model. The root mean square error (RMSE), mean absolute error (MAE), and determination coefficient (R2) were used as the evaluation indicators of the model, and the evaluation results are shown in Table 2.

[0048] Table 2 Evaluation results

[0049] S3: Obtain the key weather factors of the day and the next five days as input features and input them into the trained CNN-LSTM-Attention reference crop evapotranspiration estimation model to obtain the reference crop evapotranspiration estimation results of the day and the next five days.

[0050] The key weather factors of the day in S3 can be obtained using the local weather station, and the weather forecast for the next five days can be obtained using the weather forecast. The corresponding reference crop evapotranspiration estimation results are obtained through the model. Taking the key weather factors from 2022 to 2023 as an example, the key weather factors of a certain day during the period and the next five days are obtained as input features to estimate the daily reference crop evapotranspiration. The reference crop evapotranspiration estimation results for 2022 and 2023 are shown in the figure below. Figure 4 shown.

[0051] S4: Calculate the evapotranspiration for the current day and the next five days based on the estimated reference crop evapotranspiration for the current day and the next five days in S3 combined with the crop coefficient corresponding to the growth stage of rice.

[0052] The growth stage of rice in S4 divides the whole growth period of rice into greening stage, tillering stage, jointing and booting stage, heading and flowering stage, milky stage, and yellowing stage; the crop coefficients corresponding to the growth stages are obtained by consulting materials (such as FAO-56) or localized field test data. In this embodiment, the crop coefficients of rice greening stage, tillering stage, jointing and booting stage, heading and flowering stage, milky stage, and yellowing stage are 1.05, 1.10, 1.20, 1.15, 1.05, and 0.85; the evapotranspiration of rice on the same day is obtained by the reference crop evapotranspiration of the same day in S3 and the crop coefficient corresponding to the growth stage on that day, and the evapotranspiration in the next 5 days is obtained by the reference crop evapotranspiration of the next 5 days in S3 and the crop coefficient corresponding to the growth stage in the next 5 days. The growth stage in the next 5 days is roughly judged by the growth stage in the same period in history. The evapotranspiration is calculated as:

[0053] In the formula is crop evapotranspiration (unit: mm / d; is the crop coefficient, and the corresponding constant value is selected according to the different growth stages of rice; is the reference crop evapotranspiration (unit: mm / d).

[0054] Taking the key weather factors from 2022 to 2023 as an example, the daily evapotranspiration is calculated based on the daily reference crop evapotranspiration on a certain day during the period and the next 5 days and the growth stage of the corresponding date.

[0055] S5: construct a water balance equation based on the evapotranspiration of the day in S4 and make irrigation decisions in combination with the irrigation strategy, and calculate the irrigation water consumption of the day in combination with the farmland area to arrange irrigation for the next day.

[0056] In S5, a water balance equation is constructed based on the daily evapotranspiration and an irrigation decision is made in combination with the irrigation strategy. The water balance equation is:

[0057] In the formula , are the depth of field surface water layer at the end of day t and day t-1, respectively; The amount of irrigation water; is the rainfall; is the leakage amount on day t; For displacement.

[0058] The irrigation strategy controls the depth of the water layer in the field and keeps the soil in a shallow water layer or moist state to reduce water consumption and improve water use efficiency. The irrigation strategy is shown in Table 3 Table 3 Rice field irrigation strategies

[0059] In Table 3, saturated water content Take 40%.

[0060] In practical applications, the water layer depth at the end of each day is calculated based on the water balance equation. If it is less than the lower limit of the suitable water layer height or the lower limit of water content, irrigation will be arranged at the beginning of the next day to irrigate to the upper limit of the suitable water layer height. From the water balance equation, the water layer height calculation formula is:

[0061] When heavy rain causes the water level to rise, drainage is required. Greater than the allowable water storage capacity When the water is discharged to the aquifer, the water is discharged to the aquifer. The drainage volume calculation formula is:

[0062] The calculation formula for irrigation water volume is:

[0063] In the formula, The upper limit of the suitable water layer is is the lower limit of the suitable water layer, is the lower limit of water content, is the planned wetting depth on day t, To allow for water storage depth, is the saturated water content, is the soil moisture content on day t.

[0064] The irrigation water consumption is determined by making an irrigation decision based on the irrigation strategy, i.e. judging whether irrigation is needed, and then calculating the irrigation water consumption when irrigation is needed based on the evapotranspiration and the area of ​​farmland to be irrigated. The area of ​​farmland is shown in Table 4. The irrigation water consumption calculation formula is:

[0065] In the formula, is the irrigation area (㎡), Irrigation water volume (mm), Irrigation water consumption (m³).

[0066] Table 4 Farmland area

[0067] The changes in irrigation water volume and water layer height in 2022 and 2023 are shown in Figure 5 and Figure 6 During the planting period in 2022, rice was irrigated 18 times, with a total irrigation water volume of 3398.41 mm for farmland in the region, and an irrigation water consumption of approximately 202395.42 m³; during the planting period in 2023, rice was irrigated 8 times, with a total irrigation water volume of 1628.10 mm for farmland in the region, and a water consumption of approximately 96963.43 m³ for water pumps.

[0068] The irrigation quota of the traditional irrigation strategy is set at 70 mm. During the planting period of 2022, rice irrigation was carried out 9 times, and the total irrigation water consumption of farmland in the region was about 225,121.13 m³. During the planting period of 2023, rice irrigation was carried out 6 times, and the total irrigation water consumption of farmland in the region was about 150,080.75 m³. The comparison of water consumption between the irrigation strategy of the present invention and the traditional irrigation strategy is shown in Figure 7 .

[0069] Compared with the traditional irrigation strategy, the irrigation water consumption of each area of ​​farmland is effectively reduced by the irrigation strategy of the present invention. Comparative data show that the total irrigation water consumption of the strategy of the present invention in 2022 and 2023 was saved by about 10.1% and 35.4% respectively.

[0070] Preferably, a high-precision liquid level sensor is installed in the farmland to monitor the water layer height in real time during irrigation. When the water layer height reaches the upper limit of the suitable water layer during irrigation, irrigation is stopped to reduce the impact of sudden conditions such as irrigation water volume error caused by calculation error of the reference crop evapotranspiration estimation model and water loss in the water pipeline; S6: Based on the evapotranspiration in the next five days in S4, the irrigation water consumption in the next five days is calculated in the same way as in S5, and water is stored in the reservoir in advance.

[0071] In S6, the irrigation water consumption for the next five days is calculated based on the evapotranspiration for the next five days in the same way as in S5, and water is stored in the reservoir in advance. The evapotranspiration for the next five days is calculated, and irrigation decisions are made daily in combination with the irrigation strategy, and the daily irrigation water consumption is calculated; water is stored in the reservoir in advance based on the total irrigation water consumption for the current day and the next five days. In this embodiment, July 21, 2022 is the current day, and July 22 to July 26, 2022 are the next five days. Irrigation decisions are made daily in combination with the irrigation strategy based on the evapotranspiration for the current day and the next five days, and the daily irrigation water consumption is calculated. The irrigation decision table is shown in Table 5.

[0072] Table 5 Irrigation decision table for the current day and the next 5 days

[0073] On July 21, 2022, it is necessary to ensure that the water storage in the reservoir is not less than the total irrigation water consumption for that day and the next five days, which is 5977.93m³.

[0074] S7: Repeat S3 to S6 every day and adjust the water storage capacity of the reservoir based on the irrigation water consumption of the day and the irrigation water consumption of the next 5 days.

[0075] The water storage capacity of the water reservoir in S7 is adjusted based on the irrigation water consumption of the day and the irrigation water consumption of the next 5 days, the irrigation water consumption of the day is calculated every day, and the irrigation water consumption of the next 5 days is updated and calculated; according to the water storage capacity of the water reservoir on the day and the total irrigation water consumption of the next 5 days, the water storage is replenished in time to avoid water shortage or untimely water supply in mountainous and arid areas; the water reservoir is installed with a liquid level sensor, and the water level of the water reservoir is calibrated before irrigation. When the water volume in the water reservoir is less than the total irrigation water consumption, water is replenished to avoid the influence of sudden situations such as evaporation of the water reservoir and accidental water loss of the water reservoir. The water storage capacity calculation formula of the water reservoir is:

[0076] In the formula, is the water storage capacity of the reservoir, is the reservoir area, is the water level of the reservoir.

Claims

1. An intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation, characterized in that: The following steps are involved: S1: Obtain historical daily weather data of farmland area to calculate reference crop evapotranspiration and perform preprocessing; S2: Based on weather data and reference crop evapotranspiration, weather factors were selected as input features, reference crop evapotranspiration was used as the target variable, and a reference crop evapotranspiration estimation model was constructed and trained; S3: Obtain key weather factors for the current day and the next n days, and obtain the reference crop evapotranspiration estimation results for the current day and the next n days respectively through the reference crop evapotranspiration estimation model; S4: Based on the evapotranspiration estimation results and the crop coefficient corresponding to the rice growth stage, calculate the evapotranspiration for the current day and the next n days, make irrigation decisions, and calculate the irrigation water consumption for the next n days; S5: Adjust the water storage capacity of the reservoir based on the irrigation water consumption on the day and the irrigation water consumption in the next n days.

2. The intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation according to claim 1, characterized in that: The S1 is specifically implemented as follows: obtaining historical daily weather data in the farmland area, including daily average temperature, air pressure, average wind speed, net solar radiation, sunshine time, relative humidity, and precipitation, and dividing the data set into a training set and a test set; The reference crop evapotranspiration is calculated using the Penman PM formula and used as the target variable; The preprocessing uses the maximum-minimum normalization method to map the values ​​of all features and target variables to the range of [0, 1].

3. The intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation according to claim 2 is characterized in that: The specific implementation process of S2 is as follows: Based on the weather data and reference crop evapotranspiration in S1, the correlation strength between different weather data and reference crop evapotranspiration is quantified by grey correlation analysis, wherein the reference crop evapotranspiration is selected as a reference sequence and a comparison sequence is composed of multiple weather data; The Z weather data with the highest correlation were selected as the input features of the reference crop evapotranspiration estimation model; The reference crop evapotranspiration estimation model combines the local feature extraction capability of the convolutional network CNN, the long-term dependency processing capability of the long short-term memory network LSTM, and the weight allocation capability of the attention mechanism Attention; the reference crop evapotranspiration estimation model first extracts the local features of the input sequence through a one-dimensional convolutional layer, then inputs the local features into the LSTM layer to process the long-term dependencies in the long sequence data, and finally extracts the key features through the Attention layer, and uses a fully connected layer as the output layer to output the reference crop evapotranspiration estimation value.

4. The intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation according to claim 3 is characterized in that: The calculation of the evapotranspiration for the current day and the next n days is specifically implemented as follows: the growth stage of rice divides the entire growth period of rice into the greening stage, tillering stage, jointing and booting stage, heading and flowering stage, milky stage, and yellow stage; the evapotranspiration of rice for the current day is obtained by multiplying the reference crop evapotranspiration for the current day in S3 by the crop coefficient corresponding to the growth stage of the current day, and the evapotranspiration for the next n days is obtained by multiplying the reference crop evapotranspiration for the next n days in S3 by the crop coefficient corresponding to the growth stage of the next n days, and the growth stage of the next n days is determined by the growth stage of the same period in history.

5. The intelligent irrigation dynamic decision-making method based on reference crop evapotranspiration estimation according to claim 4, characterized in that: The irrigation decision making in step S4 is specifically implemented as follows: Calculate the water depth at the end of each day based on the water balance equation If it is less than the lower limit of the water layer height or the lower limit of the water content, irrigation will be arranged on the next day to irrigate to the upper limit of the water layer height. The water layer height is obtained by the water balance equation. When heavy rain causes the water level to rise, drainage is required. Greater than the allowable water storage capacity When the water level reaches 10000 m, the water is drained to the aquifer that allows for storage; The irrigation water consumption is determined by making an irrigation decision based on the irrigation strategy, that is, determining whether to irrigate, and then calculating the irrigation water consumption during irrigation based on the evapotranspiration and the irrigated farmland area.

Images