Intelligent prediction method for suitable sowing time and sowing rate of winter wheat

Abstract

The invention relates to the technical field of agricultural planting, in particular to an intelligent prediction method for the suitable sowing time and sowing rate of winter wheat, and the method comprises the following steps: data acquisition, data cleaning, data analysis, planting simulation and iterative updating. The limitation that in the prior art, only meteorological or soil single factors are concerned is broken through, multi-aspect data fusion analysis is conducted, the analysis efficiency is improved, then PLS is adopted for data dimension reduction and denoising, PSO-RF optimization feature selection and LSTM capture time sequence rules, dynamic fusion is conducted on the three models through an entropy weight method, comprehensive analysis and processing of data are achieved, and the analysis efficiency is improved. The performance and accuracy of the whole model are improved, prediction errors are reduced, real-time feedback and online learning of the Internet of Things are combined, iteration updating is conducted after the model is planted in each season, and the adaptability of the model to the local place is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural planting, and particularly to an intelligent prediction method for the suitable sowing date and sowing rate of winter wheat. Background Art

[0002] Winter wheat is a wheat variety sown in autumn and harvested in the following summer. Shandong is located in the North China Plain and is a core production area of winter wheat in China (with an annual output accounting for 14% of the country), ranking third in the country in terms of grain sown area and total output; at the same time, Shandong Province is also a typical planting area of North China winter wheat, with the wheat planting area ranking second in the country, accounting for about 14% of the country. However, in recent years, affected by global warming, the autumn and winter temperatures in Shandong Province and some surrounding areas have continued to rise, and the sunshine hours have decreased, making winter wheat prone to frost damage, winter drought and spring drought during the jointing and booting stages.

[0003] Therefore, in order to ensure the grain yield and reduce losses, through a planting method of dryland wheat disclosed in an invention application with the publication number CN113317147A and an intelligent prediction method and system for the growth state of wheat crops disclosed in an invention application with the publication number CN115511219A, etc., the future growth situation of wheat is predicted according to weather information and the production situation of wheat.

[0004] However, it is found in the use process that the existing wheat planting prediction models still lack in predicting the sowing date and sowing rate of wheat. On the one hand, most models only analyze the weather and predict based on traditional experience, resulting in low accuracy; on the other hand, the models for long-term prediction of the sowing date and sowing rate of wheat are still incomplete, and moreover, it is not convenient to accurately monitor the weather of the planting area, resulting in poor practicability. Therefore, there is an urgent need for an intelligent prediction method for the suitable sowing date and sowing rate of winter wheat to improve the above problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides an intelligent prediction method for the suitable sowing date and sowing rate of winter wheat, which breaks through the limitation of the existing technology that only focuses on single factors such as meteorology or soil, improves the analysis efficiency through multi-faceted data fusion analysis, secondly uses PLS for data dimensionality reduction and noise reduction, PSO-RF for optimized feature selection, LSTM for capturing time series patterns, and the three models are dynamically fused through the entropy weight method to achieve comprehensive analysis and processing of data, and combined with Internet of Things real-time feedback and online learning, enabling the model to be iteratively updated after each season of planting, improving the adaptability of the model to the local area.

[0006] An intelligent prediction method for the suitable sowing date and sowing rate of winter wheat according to the present invention includes the following steps:

[0007] S1. Data acquisition:

[0008] Obtain weather data, soil data, and wheat variety data;

[0009] S2. Data cleaning:

[0010] Screen and standardize the obtained data;

[0011] S3. Data analysis:

[0012] Use PLS for data dimensionality reduction and denoising, PSO-RF for optimized feature selection, LSTM to capture time series patterns, and the three models are dynamically fused by the entropy weight method to achieve data analysis and processing, predict the sowing date and sowing rate of wheat, and formulate a wheat planting plan at the same time;

[0013] S4. Planting simulation:

[0014] According to the analysis results and the formulated planting plan, simulate the winter wheat planting to facilitate the user's intuitive judgment of the planting plan;

[0015] S5. Iterative update:

[0016] Use the Internet of Things method to store the obtained data, analysis results, planting plan, and actual planting situation and yield in real time, and use them as the training data of the model to enable the model to perform online learning and continuous iteration;

[0017] Preferably, in S1, weather acquisition is divided into two aspects, that is, obtaining weather data released by the National Meteorological Administration and obtaining weather data of the planting area;

[0018] Soil data is obtained by deploying low-cost Internet of Things nodes to form a soil monitoring network and monitoring the basic soil fertility before wheat sowing;

[0019] Seed gene data is obtained by accessing agricultural planting platforms.

[0020] Preferably, in S3, it is divided into pre-sowing analysis and post-sowing analysis;

[0021] Before sowing, according to the obtained weather, soil, and seed gene data, judge the variety and the corresponding sowing date and sowing rate, and generate a planting plan;

[0022] After sowing, according to the real-time obtained weather data and soil data, predict the growth of wheat, fine-tune the planting plan, and enable the user to take timely measures to ensure the wheat yield.

[0023] Preferably, a climate attention layer is added to the LSTM model in S3 to strengthen the model's response to extreme weather events.

[0024] Preferably, in the planting simulation process of S4, first, the area, geographical location, and soil conditions of the planting land are provided through human-computer interaction, and then combined with the obtained weather data and soil data, the generated planting plan is simulated to facilitate the user to observe the implementation effect of the planting plan and adjust the unreasonable positions in the planting plan.

[0025] Preferably, in S5, when the actual yield deviates from the predicted value by more than 10%, the stored data is used to participate in model training to enable the model to learn online and fine-tune the weights.

[0026] Preferably, the low-cost Internet of Things node in S1 is composed of a low-power soil multi-parameter sensor, an edge computing gateway, a solar power supply module, and a wireless network transmission module.

[0027] Preferably, the weather data of the planting land is obtained by using a planting land weather data acquisition module.

[0028] Preferably, the planting area weather data acquisition module includes a receiving cylinder, a working box, a steel drill, a support base, a helium gas cylinder, a first electric control valve, a winding shaft, a driving motor, a silk thread, an exhaust pipe, a pressure sensor, a bottom plate, a weather balloon, a rubber pad, a second electric control valve, a one-way valve, a weather monitor, and a direction adjustment mechanism. The receiving cylinder is installed on the top of the working box, the steel drill is installed on the bottom of the working box, the support base is installed on the working box, the helium gas cylinder is installed on the receiving cylinder, the first electric control valve is installed on the working box, and the exhaust port of the helium gas cylinder is connected to the first electric control valve. The winding shaft is rotatably installed in the working box, the driving motor is fixedly installed in the working box, and the output shaft of the driving motor is connected to the side end of the winding shaft. The silk thread is wound on the winding shaft. The exhaust pipe and the pressure sensor are both installed at the bottom of the receiving cylinder, and the receiving cylinder is communicated with the inside of the working box through the exhaust pipe. The weather balloon is installed on the top of the bottom plate, and both the weather balloon and the bottom plate are located in the receiving cylinder. The rubber pad and the second electric control valve are both installed on the top of the weather balloon. The one-way valve and the weather monitor are both installed at the bottom of the bottom plate, and one end of the silk thread passes through the exhaust pipe and is connected to the bottom of the one-way valve. The direction adjustment mechanism is installed at the bottom of the bottom plate. By hammering the steel drill, multiple steel drills are inserted into the ground to fix the receiving cylinder and the working box. When it is necessary to monitor the weather, the driving motor runs in reverse to make the winding shaft roll up the silk thread until the bottom of the one-way valve is butted against the exhaust pipe. By opening the first electric control valve, the helium gas in the helium gas cylinder flows through the first electric control valve, the working box, the exhaust pipe, and the one-way valve in sequence and enters the weather balloon, causing the weather balloon to expand. After the weather balloon expands, the one-way valve is closed. At the same time, the driving motor runs to release the silk thread, so that helium gas continuously enters the receiving cylinder, and the air pressure in the receiving cylinder is detected by the pressure sensor. The weather balloon is pushed out of the receiving cylinder by the helium gas in the receiving cylinder. The winding shaft continuously releases the silk thread, so that the weather balloon floats upward, and the position of the weather balloon is adjusted by the direction adjustment mechanism. The weather is monitored by the weather monitor. After the detection is completed, the winding shaft rolls up the silk thread, and at the same time, by opening the second electric control valve, the helium gas in the weather balloon is slowly released, and the weather balloon is put into the receiving cylinder, thereby improving the practicability of the weather data acquisition module.

[0029] Preferably, the direction adjustment mechanism includes a position receiver, a position transmitter, an annular guide rail, an electric slider, a propeller, and a counterweight. The position receiver is installed on the bottom plate, the position transmitter is installed in the working box, the annular guide rail is fixedly installed at the bottom of the bottom plate, the electric slider is slidably installed on the annular guide rail, and the propeller and the counterweight are both installed on the electric slider. By detecting the position of the position transmitter by the position receiver, when the weather balloon deviates from the receiving cylinder beyond the threshold value, the electric slider slides on the annular guide rail to adjust the position of the propeller. At the same time, by running the propeller, the weather balloon is driven to move to adjust the position of the weather balloon, thereby improving the practicability of the direction adjustment mechanism.

[0030] The beneficial effects of the present invention compared with the prior art are as follows:

[0031] 1. Break through the limitation of the prior art that only focuses on single factors such as meteorology or soil. Through multi-faceted data fusion analysis, the analysis efficiency is improved.

[0032] 2. Use PLS for data dimensionality reduction and noise reduction, PSO-RF for optimized feature selection, and LSTM to capture temporal patterns. And the three models are dynamically fused by the entropy weight method to achieve comprehensive analysis and processing of data, improve the performance and accuracy of the overall model, and reduce prediction errors.

[0033] 3. Combine the real-time feedback of the Internet of Things with online learning, so that the model can be iteratively updated after each season of planting, improving the adaptability of the model to the local area.

[0034] 4. Adopt a weather data acquisition module for the planting area, and obtain the local weather in the field through high-altitude monitoring. Combine with the weather data released by the national meteorological bureau to improve the weather prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic flowchart of the present invention;

[0036] Figure 2 is a first axonometric structural schematic diagram of the weather data acquisition module for the planting area of the present invention;

[0037] Figure 3 is a second axonometric structural schematic diagram of the weather data acquisition module for the planting area of the present invention;

[0038] Figure 4 is a front view structural schematic diagram of the weather data acquisition module for the planting area of the present invention;

[0039] Figure 5 is a front view sectional structural schematic diagram of the weather data acquisition module for the planting area of the present invention;

[0040] Figure 6 is a front view sectional structural schematic diagram of the direction adjustment mechanism of the present invention;

[0041] Figure 7 is an axonometric sectional structural schematic diagram of the weather data acquisition module for the planting area of the present invention;

[0042] Figure 8 is a structural schematic diagram when the weather balloon is launched of the present invention.

[0043] Reference numerals in the drawings: 1, receiving cylinder; 2, working box; 3, steel drill rod; 4, support base; 5, helium gas cylinder; 6, first electric control valve; 7, winding shaft; 8, drive motor; 9, wire; 10, exhaust pipe; 11, pressure sensor; 12, bottom plate; 13, weather balloon; 14, rubber pad; 15, second electric control valve; 16, one-way valve; 17, weather monitor; 18, position receiver; 19, position transmitter; 20, annular guide rail; 21, electric slider; 22, propeller; 23, counterweight. Detailed implementation manners

[0044] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. The present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0045] As Figures 1 to 8 shown, a method for intelligent prediction of the suitable sowing date and sowing rate of winter wheat includes the following steps:

[0046] S1. Data acquisition:

[0047] Obtain weather data, soil data and wheat variety data;

[0048] The weather data such as daily maximum temperature, daily minimum temperature, solar radiation and precipitation;

[0049] S2. Data cleaning:

[0050] Screen and standardize the acquired data;

[0051] S3. Data analysis:

[0052] Use PLS (partial least squares method) for data dimensionality reduction and noise reduction, PSO-RF (particle swarm optimization-based random forest) for optimized feature selection, LSTM (long short-term memory network) to capture time series patterns, and the three models are dynamically fused by the entropy weight method (weight formula: w i = ∑(1 - E i ) / 1 - E i , E i is the information entropy of each model), to realize the analysis and processing of the data, predict the sowing date and sowing rate of wheat, and at the same time formulate a wheat planting plan;

[0053] S4. Planting simulation:

[0054] According to the analysis results and the formulated planting plan, simulate the planting of winter wheat to facilitate the user's intuitive judgment of the planting plan;

[0055] S5. Iterative update:

[0056] In the Internet of Things (IoT) approach, the acquired data, analysis results, planting plans, and actual planting conditions and yields are stored in real time and used as training data for the model, enabling the model to perform online learning and continuous iteration.

[0057] In step S1, weather acquisition is divided into two aspects: obtaining weather data released by the national meteorological bureau and obtaining weather data at the planting site.

[0058] Soil data is obtained by deploying low-cost IoT nodes to form a soil monitoring network. Before sowing wheat, the basic soil fertility is monitored, and then the type of land is determined.

[0059] Seed gene data is obtained by accessing agricultural planting platforms.

[0060] In step S3, it is divided into pre-sowing analysis and post-sowing analysis.

[0061] Before sowing, based on the acquired weather, soil, and seed gene data, the variety and corresponding sowing date and seeding rate are determined, and a planting plan is generated.

[0062] After sowing, based on the real-time acquired weather data and soil data, the growth trend of wheat is predicted, and the planting plan is fine-tuned to enable users to take timely measures to ensure the wheat yield.

[0063] In the LSTM model in step S3, a climate attention layer is added to enhance the model's response to extreme weather events.

[0064] The LSTM network structure enhanced by the climate attention mechanism includes:

[0065] Input layer: Receiving meteorological warning signals for the next 15 days and historical time series data.

[0066] Attention calculation layer: Generating climate event attention weights through the following formula:

[0067] α t = softmax(W q ·h t + W k ·c t )

[0068] where h t is the LSTM hidden state, and c t is the meteorological warning feature vector.

[0069] Output layer: Fusing the attention weights and the LSTM output to generate the sowing date prediction value.

[0070] In the planting simulation process of S4, first, the area, geographical location, and soil conditions of the planting land are provided through human-computer interaction. Then, combined with the obtained weather data and soil data, the generated planting plan is simulated to facilitate the user's observation of the implementation effect of the planting plan and adjustment of unreasonable positions in the planting plan.

[0071] In S5, when the actual yield deviates from the predicted value by more than 10%, the stored data is used to participate in model training to enable the model to learn online and fine-tune the weights.

[0072] The loss function of the online learning mechanism is as follows:

[0073] Loss = 0.7·MSE(Y pre d,Y rea l)+0.3·Huber Loss(C pre d,C rea l)

[0074] The low-cost IoT node in S1 is composed of a low-power soil multi-parameter sensor, an edge computing gateway, a solar power supply module, and a wireless network transmission module.

[0075] The weather data of the planting land is obtained by a planting land weather data acquisition module.

[0076] The weather data acquisition module includes a receiving cylinder 1, a working box 2, a steel drill 3, a support base 4, a helium gas cylinder 5, a first electric control valve 6, a winding shaft 7, a driving motor 8, a silk thread 9, an exhaust pipe 10, a pressure sensor 11, a bottom plate 12, a weather balloon 13, a rubber pad 14, a second electric control valve 15, a one-way valve 16, a weather monitor 17, and a direction adjustment mechanism. The receiving cylinder 1 is installed on the top of the working box 2, the steel drill 3 is installed on the bottom of the working box 2, the support base 4 is installed on the working box 2, the helium gas cylinder 5 is installed on the receiving cylinder 1, the first electric control valve 6 is installed on the working box 2, and the exhaust port of the helium gas cylinder 5 is connected to the first electric control valve 6. The winding shaft 7 is rotatably installed in the working box 2, the driving motor 8 is fixedly installed in the working box 2, and the output shaft of the driving motor 8 is connected to the side end of the winding shaft 7. The silk thread 9 is wound on the winding shaft 7. The exhaust pipe 10 and the pressure sensor 11 are both installed at the bottom of the receiving cylinder 1, and through the exhaust pipe 10, the receiving cylinder 1 is communicated with the inside of the working box 2. The weather balloon 13 is installed on the top of the bottom plate 12, and both the weather balloon 13 and the bottom plate 12 are located in the receiving cylinder 1. The rubber pad 14 and the second electric control valve 15 are both installed on the top of the weather balloon 13. The one-way valve 16 and the weather monitor 17 are both installed at the bottom of the bottom plate 12, and one end of the silk thread 9 passes through the exhaust pipe 10 and is connected to the bottom of the one-way valve 16. The direction adjustment mechanism is installed at the bottom of the bottom plate 12.

[0077] The direction adjusting mechanism includes a position receiver 18, a position transmitter 19, an annular guide rail 20, an electric slider 21, a propeller 22 and a counterweight 23. The position receiver 18 is installed on the bottom plate 12, the position transmitter 19 is installed in the working box 2, the annular guide rail 20 is fixedly installed at the bottom of the bottom plate 12, the electric slider 21 is slidably installed on the annular guide rail 20, and both the propeller 22 and the counterweight 23 are installed on the electric slider 21.

[0078] Examples and effect verification:

[0079] Comparative test in the hilly area of central Shandong:

[0080] Index Traditional experience-based decision-making The model of the present invention Improvement range Sowing date accuracy (days) ±6.2 ±1.3 79.0％ Average yield per mu (kg) 612.5 689.4 12.6％ Incidence of freezing injury 18.7％ 5.2％ 72.2％ Nitrogen fertilizer utilization rate 32.1％ 41.5％ 29.3％

[0081] The intelligent prediction method for the suitable sowing date and sowing rate of winter wheat of the present invention, its installation method, connection method or setting method are all common mechanical methods, and any method that can achieve its beneficial effects can be implemented; the helium gas cylinder 5, the first electric control valve 6, the drive motor 8, the pressure sensor 11, the weather balloon 13, the second electric control valve 15, the weather monitor 17, the position receiver 18, the position transmitter 19, the electric slider 21 and the propeller 22 of the intelligent prediction method for the suitable sowing date and sowing rate of winter wheat of the present invention are purchased on the market, and those skilled in the art only need to install and operate according to the attached user manual, without the need for those skilled in the art to make creative efforts.

[0082] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art in this technical field, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent prediction method for suitable sowing period and sowing amount of winter wheat, characterized in that: The following steps are involved: S1. Data acquisition: Obtain weather data, soil data, and wheat variety data; S2. Data cleaning: Filter and standardize the acquired data; S3. Data analysis: PLS is used for data dimension reduction and denoising, PSO-RF optimizes feature selection, LSTM captures time series rules, and the three models are dynamically integrated through the entropy weight method to realize data analysis and processing, predict wheat sowing time and sowing amount, and formulate wheat planting plans; S4. Planting simulation: Based on the analysis results and the formulated planting plan, winter wheat planting is simulated to facilitate users to make intuitive judgments on the planting plan; S5, Iteration Update: The Internet of Things is used to store acquired data, analysis results, planting plans, actual planting conditions and yields in real time, and use them as training data for the model, allowing the model to learn online and iterate continuously.

2. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 1, characterized in that: In S1, weather acquisition is divided into two aspects, namely, obtaining weather data released by the National Meteorological Bureau and obtaining weather data of the planting area; Soil data is formed by deploying low-cost IoT nodes to form a soil monitoring network, which monitors basic soil fertility before wheat sowing; Seed genetic data is obtained by accessing agricultural planting platforms.

3. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 1, characterized in that: The S3 is divided into pre-sowing analysis and post-sowing analysis; Before sowing, the variety and corresponding sowing time and amount are determined based on the weather, soil and seed gene data obtained, and a planting plan is generated; After sowing, the growth of wheat is predicted based on real-time weather and soil data, and the planting plan is fine-tuned, allowing users to take timely measures to ensure wheat yields.

4. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 1, characterized in that: A climate attention layer is added to the LSTM model in S3 to enhance the model's response to extreme weather events.

5. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 1, characterized in that: During the planting simulation process in S4, the area, geographical location and soil conditions of the planting site are first provided through human-computer interaction, and then the generated planting plan is simulated in combination with the acquired weather data and soil data, so that the user can observe the implementation effect of the planting plan and adjust the unreasonable positions in the planting plan.

6. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 1, characterized in that: In S5, when the actual output deviates from the predicted value by more than 10%, the stored data is used to participate in the model training, so that the model can learn online and the weights can be fine-tuned.

7. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat as claimed in claim 2, characterized in that: The low-cost Internet of Things node in S1 is composed of a low-power soil multi-parameter sensor, an edge computing gateway, a solar power supply module and a wireless network transmission module.

8. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat as claimed in claim 2, characterized in that: The weather data of the planting site is acquired by using a planting site weather data acquisition module.

9. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 8, characterized in that: The planting site weather data acquisition module comprises a receiving tube (1), a working box (2), a steel chisel (3), a support seat (4), a helium bottle (5), a first electric control valve (6), a winding shaft (7), a driving motor (8), a silk thread (9), an exhaust pipe (10), a pressure sensor (11), a bottom plate (12), a weather balloon (13), a rubber pad (14), a second electric control valve (15), a one-way valve (16), a weather monitor (17) and a direction adjustment mechanism. The receiving tube (1) is installed on the top of the working box (2), the steel chisel (3) is installed on the bottom of the working box (2), the support seat (4) is installed on the working box (2), the helium bottle (5) is installed on the receiving tube (1), the first electric control valve (6) is installed on the working box (2), and the exhaust port of the helium bottle (5) is connected to the first electric control valve (6). The winding shaft (7) is rotatably installed in the working box (2), and the driving motor (8) is connected to the first electric control valve (6). The driving motor (8) is fixedly installed in the working box (2), and the output shaft of the driving motor (8) is connected to the side end of the winding shaft (7), the silk thread (9) is wound on the winding shaft (7), the exhaust pipe (10) and the pressure sensor (11) are both installed at the bottom of the receiving tube (1), and the receiving tube (1) and the interior of the working box (2) are communicated through the exhaust pipe (10), the weather balloon (13) is installed on the top of the bottom plate (12), and the weather balloon (13) and the bottom plate (12) are both located in the receiving tube (1), the rubber pad (14) and the second electric control valve (15) are both installed on the top of the weather balloon (13), the one-way valve (16) and the weather monitor (17) are both installed on the bottom of the bottom plate (12), and one end of the silk thread (9) passes through the exhaust pipe (10) and is connected to the bottom of the one-way valve (16), and the direction adjustment mechanism is installed at the bottom of the bottom plate (12).

10. The intelligent prediction method for suitable sowing period and sowing amount of winter wheat according to claim 9, characterized in that: The direction adjustment mechanism comprises a position receiver (18), a position transmitter (19), an annular guide rail (20), an electric slider (21), a propeller (22) and a counterweight (23); the position receiver (18) is mounted on a base plate (12); the position transmitter (19) is mounted in a working box (2); the annular guide rail (20) is fixedly mounted on the bottom of the base plate (12); the electric slider (21) is slidably mounted on the annular guide rail (20); and the propeller (22) and the counterweight (23) are both mounted on the electric slider (21).

Images