A greenhouse water-saving irrigation method based on multi-information fusion

Abstract

The present invention discloses a greenhouse water-saving irrigation method based on multi-information fusion, which is applied to an irrigation system. The greenhouse water-saving irrigation method based on multi-information fusion includes: collecting data on light intensity, air temperature, air humidity, and soil water content in the greenhouse at fixed intervals. Using weighted normalization, multi-objective centroid method and average silhouette coefficient to optimize clustering for the growth characteristics of greenhouse crops, and obtaining the change in soil water content corresponding to the centroid of each cluster, that is, the water demand of crops in different environments in the greenhouse; establishing a soil water balance model suitable for the greenhouse, determining crop coefficients in different periods according to different crop growth cycles, calculating more accurately the water lost due to crop evapotranspiration, and using the results obtained by optimal clustering as the reference input of the model predictive controller to avoid irrigation waste caused by relying on experience.

Description

Technical Field

[0001] The present invention belongs to the field of agriculture, and particularly relates to a greenhouse water-saving irrigation method based on multi-information fusion. Background Technique

[0002] Agriculture is an important pillar of economic development, making a significant contribution to the gross production value and ensuring food security. Agricultural organizations predict that in a few years, the irrigated food production will increase by more than 50%. With the improvement of water productivity, an additional 10% of agricultural water is needed to achieve this goal. However, the land area for growing food has not expanded. Therefore, it is crucial to improve water use efficiency and use irrigation methods that significantly save water and increase yields.

[0003] The greenhouse environment is a direct factor affecting the yield and quality of internal crops. At present, the irrigation methods of the vast majority of domestic greenhouses only rely on farmers' production experience for regular irrigation. Some greenhouses use soil moisture sensors to monitor soil moisture and set a fixed soil water content threshold for irrigation. The irrigation method based on soil moisture sensors has a low cost, but the soil moisture sensors need to be calibrated, and the data collected by the sensors buried in the soil for a long time may have errors, which will affect the formulation of irrigation plans and cause water resource waste. Some greenhouse irrigation systems measure or calculate the evapotranspiration in the greenhouse to determine the water consumption of crops within a certain period of time and irrigate according to the water consumption of crops. However, the cost of directly measuring the evapotranspiration in the greenhouse is relatively high, making it difficult to promote on a large scale. Merely relying on the evapotranspiration mechanism model to determine the water demand of crops will also produce relatively large errors, resulting in waste of water resources. The current main irrigation methods do not consider the impact of the greenhouse environment on crops, resulting in low water production efficiency. Summary of the Invention

[0004] The purpose of the present invention is to propose a greenhouse water-saving irrigation method based on multi-information fusion in view of the problems raised in the background technique.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] A greenhouse water-saving irrigation method based on multi-information fusion proposed by the present invention is applied to an irrigation system, and is characterized in that: the greenhouse water-saving irrigation method based on multi-information fusion includes:

[0007] S1. Collect data on the light intensity, air temperature, air humidity, and soil water content in the greenhouse at fixed intervals, and calculate the average light intensity, average air temperature, average air humidity, and soil water content change amount in the greenhouse every day to obtain the daily environmental information in the greenhouse.

[0008] S2. Perform weighted normalization on the average light intensity, average air temperature, average air humidity, and soil water content change amount per day to obtain a standard greenhouse environment dataset.

[0009] S3. Cluster the standard greenhouse environment dataset. Take each day in the standard greenhouse environment dataset as a sample, set the range of the target number of centroids, and define the initial target centroid as the minimum boundary value in the range of the target number of centroids.

[0010] S4. Sequentially select centroids from the standard greenhouse environment dataset until the number of selected centroids reaches the minimum boundary value in the range of the target number of centroids, and then calculate the average silhouette coefficient of all samples in the standard greenhouse environment dataset.

[0011] S5. Increase the target number of centroids by one, repeat S4 until the target number of centroids reaches the maximum boundary value in the range of the target number of centroids, and select the target centroid with the largest average silhouette coefficient as the optimal clustering.

[0012] S6. Establish a greenhouse soil water balance model. Determine which cluster of the optimal clustering result it belongs to according to the greenhouse environment information of the previous day in the greenhouse, and select the soil water content change amount of the corresponding centroid as a reference value to input into the model prediction controller of the irrigation system for prediction. The obtained value is input into the greenhouse soil water balance model, and the result calculated by the greenhouse soil water balance model is used as the actual irrigation amount.

[0013] Preferably, performing weighted normalization on the average light intensity, average air temperature, average air humidity, and soil water content change amount per day to obtain a standard greenhouse environment dataset, including:

[0014] Let the average light intensity per day be L m , the average air temperature per day be T m , the average air humidity per day be H m , the soil water content change amount per day be M d , the environmental information set per day be D, and the weights of the average light intensity, average air temperature, average air humidity, and soil water content change amount per day be w l , w t , w h , w m , respectively, and the weight of the environmental information set per day be w d , and the standard greenhouse environment dataset be E. Then the environmental information D of one day in the environmental information set D i = {L mi , T mi , H mi , M di}.

[0015] Weighted normalization is performed on the daily average light intensity, average air temperature, average air humidity, and soil water content change amount to obtain a standard greenhouse environment dataset, including:

[0016] E = {L e , T e , H e , M e};

[0017]

[0018] Among them, L e , T e , H e , M e successively represent the average light intensity, average air temperature, average air humidity, and soil water content change amount after weighted normalization processing. L mi , T mi , H mi , M di respectively represent a certain element in D. L mmin , T mmin , H mmin , M dmin successively represent the minimum values of the corresponding similar elements in D. L mmax , T mmax , H mmax , M dmax successively represent the maximum values of the corresponding similar elements in D. L ei , T ei , H ei , M ei successively represent the values corresponding to the average light intensity, average air temperature, average air humidity, and soil water content change amount of one day in D in E.

[0019] Preferably, clustering is performed on the standard greenhouse environment dataset. Taking each day in the standard greenhouse environment dataset as a sample, set the range of the target centroid number, and define the initial target centroid as the minimum boundary value in the range of the target centroid number, including:

[0020] Set the range of the target centroid number as K1 - K n , and define the initial target centroid as the minimum boundary value K1 in the range of the target centroid number.

[0021] Preferably, centroids are successively selected from the standard greenhouse environment dataset until the number of selected centroids reaches the minimum boundary value in the range of the target centroid number, including:

[0022] Select a sample with the average light intensity closest to the average value of the average light intensities of all samples from the standard greenhouse environment dataset as the first centroid;

[0023] Select the first centroid c1 using the probability calculation formula:

[0024]

[0025] Among them, select the sample corresponding to the largest value of P(c1) as the first centroid c1, represents the sum of the average light intensities of all samples, L ei represents the average light intensity of sample i in the standard greenhouse environment dataset E, L ej represents the average light intensity of sample j in the standard greenhouse environment dataset E, and n represents the total number of samples.

[0026] Then calculate the distance of each sample in the standard greenhouse environment dataset from the first centroid, and select the sample with the farthest distance as the second centroid.

[0027] Let the distance of each sample in the standard greenhouse environment dataset from the first centroid c1 be denoted as d(E i , c1), and:

[0028]

[0029] Select the second centroid c2 using the probability calculation formula:

[0030]

[0031] Among them, select the sample corresponding to the largest value of P(c2) as the second centroid c2, represents the sum of the distances of all samples from the first centroid c1, E i represents sample i in the standard greenhouse environment dataset E, E j represents the standard greenhouse environment dataset E d sample j in.

[0032] Then calculate the distances of each sample in the standard greenhouse environment dataset from the first centroid and the second centroid respectively, and assign each sample to the nearest centroid to form their respective clusters.

[0033] Let the distance of each sample in the standard greenhouse environment dataset from the second centroid c2 be denoted as d(E i , c2), and

[0034]

[0035] Continue to select the remaining centroids from the standard greenhouse environment dataset, calculate the distances from each sample in the standard greenhouse environment dataset to each centroid, and assign each sample to the nearest centroid to form their respective clusters. The selection method for the remaining centroids is as follows: Calculate the sample with the farthest distance from its centroid in each determined cluster as the next centroid.

[0036] The selection of the next centroid c is represented by the probability calculation formula q :

[0037]

[0038] Among them, select the sample corresponding to the largest value of P(c q ) as the next centroid c q , d(E r , c q ) represents the distance between a sample in the confirmed cluster and its centroid. represents the sum of the distances from all samples in the confirmed cluster to their centroids, F q represents all confirmed clusters, E r represents the sample r in the confirmed cluster, E t represents the sample t in the confirmed cluster.

[0039] Let each centroid be represented as c i .

[0040] According to formulas (1) and (2), the distances from each sample to each centroid are represented by the formula:

[0041]

[0042] Until the number of selected centroids reaches the minimum boundary value K1 within the range of the target number of centroids.

[0043] Preferably, the maximum boundary value within the range of the target number of centroids is K n .

[0044] Preferably, establish a greenhouse soil moisture balance model, including:

[0045] The greenhouse soil moisture balance model is as follows:

[0046]

[0047] And,

[0048] ET c = K c ·ET o ;

[0049]

[0050] Among them, I(t) represents the irrigation amount, with the unit of cubic centimeter per hour, and ET c (t) represents the water lost due to crop evapotranspiration, with the unit of millimeter per hour, θ(t) represents the soil water content, with the unit of cubic centimeter per hour, q1, q2, and q3 are estimated by directly measuring the initial soil irrigation amount, the water lost due to crop evapotranspiration, and the water content and using the least squares method. ET o represents the crop reference evapotranspiration, and K c represents the crop coefficient, d, d1, and d2 are all the days of crop growth, and K ini is the crop coefficient at the initial stage of the recommended crop generation, and K mid is the crop coefficient at the middle stage of the recommended crop generation, and K end is the crop coefficient at the final stage of the recommended crop generation. RH represents the average daily relative humidity in the greenhouse, h represents the average height of the crop growth at the initial, middle, and final stages, with the unit of meter, and R a represents the extraterrestrial radiation, with the unit of megajoule per square meter per day, τ represents the ratio of solar radiation inside and outside, and T, T max and T min respectively represent the average, maximum, and minimum greenhouse air temperatures, with the unit of degree Celsius.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] 1. Optimize clustering for greenhouse crop growth characteristics by using weighted normalization, multi-objective centroid method, and average silhouette coefficient, and obtain the change amount of soil water content corresponding to the centroid of each cluster, that is, the water demand of crops in different environments in the greenhouse;

[0053] 2. Establish a soil water balance model suitable for the greenhouse, determine the crop coefficients in different periods according to the different crop growth cycles, calculate more accurately the water lost due to crop evapotranspiration, and use the results obtained by optimal clustering as the reference input of the model predictive controller to avoid irrigation waste caused by relying on experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a block diagram of the greenhouse water-saving irrigation method based on multi-information fusion of the present invention;

[0055] Figure 2 is a schematic diagram of the irrigation system of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0057] It should be noted that when a component is referred to as being "connected" to another component, it can be directly connected to the other component or there can also be an intermediate component. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs. The terms used in the specification of the present application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0058] As Figure 1-2 shown, a greenhouse water-saving irrigation method based on multi-information fusion includes:

[0059] S1. Collect data on the light intensity, air temperature, air humidity and soil water content in the greenhouse at fixed intervals, and calculate the average light intensity, average air temperature, average air humidity and soil water content change amount in the greenhouse every day to obtain the daily environmental information in the greenhouse.

[0060] It should be noted that after the irrigation system of this method is completed, the irrigation system includes an outdoor weather station, an air temperature sensor, an air humidity sensor and a soil moisture sensor in the greenhouse, which successively collect data on the light intensity (the outdoor weather station collects the light intensity outside the greenhouse, and then multiplies it by the solar radiation ratio to obtain the light intensity in the greenhouse), air temperature, air humidity and soil water content in the greenhouse. And the irrigation system also includes an upper PC, a lower computer and an irrigation solenoid valve. The upper PC includes a model predictive controller (the model predictive controller belongs to a module in the upper PC).

[0061] Taking the tomato production process as an example, as Figure 2 shown, the outdoor weather station, the air temperature sensor, the air humidity sensor and the soil moisture sensor in the greenhouse collect data on light intensity, air temperature, air humidity and soil water content once every hour in sequence. The outdoor weather station, the air temperature sensor, the air humidity sensor and the soil moisture sensor in the greenhouse send the data to the lower computer (the lower computer is the stm32f107 lower computer) through RS-485 communication. The lower computer sends the information to the upper PC through TCP communication and stores it in the MySQL database of the upper PC. During the irrigation process, the upper PC sends irrigation instructions to the lower computer, and the lower computer realizes irrigation control by controlling the opening and closing of the irrigation solenoid valve.

[0062] The time range of the average light intensity, average air temperature, average air humidity and soil water content change amount in the greenhouse every day is from 6:00 the previous day to 6:00 on the current day.

[0063] S2. Perform weighted normalization on the average light intensity, average air temperature, average air humidity and soil water content change amount every day to obtain a standard greenhouse environment data set.

[0064] Specifically, because the dimensions of the light intensity, air temperature, air humidity and soil water content change amount are different, it is necessary to normalize all the collected data; since different environmental information has different effects on crop transpiration and soil water content change amount, weights need to be set for different elements in the environmental information during the normalization process.

[0065] Let the average light intensity per day be L m , the average air temperature per day be T m , the average air humidity per day be H m , the soil water content change amount per day be M d , the environmental information set per day be D, and the weights of the average light intensity, average air temperature, average air humidity, and soil water content change amount per day be w l , w t , w h , w n , and the weight of the environmental information set per day be w d . If the standard greenhouse environment data set is E, then the environmental information D of one day in the environmental information set D i = {L mi , T mi , H mi , M di}.

[0066] Perform weighted normalization on the average light intensity, average air temperature, average air humidity and soil water content change amount every day to obtain a standard greenhouse environment data set, including:

[0067] E = {L e , T e , H e , M e};

[0068]

[0069] Among them, L e , T e , H e , M e respectively represent the average light intensity, average air temperature, average air humidity and soil water content change amount after weighted normalization, Lmi , T mi , H mi , M di respectively represent a certain element in D, L mmin , T mmin , H mmin , M dmin successively represent the minimum value of the corresponding similar elements in D, L mmax , T mmax , H mmax , M dmax successively represent the maximum value of the corresponding similar elements in D, L ei , T ei , H ei , M ei successively represent the corresponding values in E of the average light intensity, average air temperature, average air humidity and change amount of soil water content in one day in D.

[0070] S3. Cluster the standard greenhouse environment data set. Take each day in the standard greenhouse environment data set as a sample, set the range of the target number of centroids, and define the initial target centroid as the minimum boundary value in the range of the target number of centroids.

[0071] Specifically, set the range of the target number of centroids as K1 - K n , and define the initial target centroid as the minimum boundary value K1 in the range of the target number of centroids.

[0072] In this embodiment, the K - means++ algorithm is used to cluster the greenhouse environment information data set E. Set the value range of the target number of centroids K as 3 - 5, and select the initial target number of centroids as 3 (the range of the target centroid and the target centroid are not restricted and can be set according to actual needs). One sample is {L ei , T ei , H ei , M ei}.

[0073] S4. Successively select centroids from the standard greenhouse environment data set until the number of selected centroids reaches the minimum boundary value in the range of the target number of centroids, and then calculate the average silhouette coefficient of all samples in the standard greenhouse environment data set.

[0074] Specifically, successively select centroids from the standard greenhouse environment data set as follows:

[0075] Select the sample with the average light intensity closest to the average value of the average light intensity of all samples from the standard greenhouse environment data set as the first centroid.

[0076] Represent the selection of the first centroid c1 with the probability calculation formula:

[0077]

[0078] Among them, the sample corresponding to the largest value of P(c1) is selected as the first centroid c1 (that is, the sample with the average light intensity closest to the average light intensity of all samples is selected from the standard greenhouse environment dataset as the first centroid). represents the sum of the average light intensities of all samples, L ei represents the average light intensity of sample i in the standard greenhouse environment dataset E, L ej represents the average light intensity of sample j in the standard greenhouse environment dataset E, and n represents the total number of samples.

[0079] Then, calculate the distance of each sample in the standard greenhouse environment dataset from the first centroid, and select the sample with the farthest distance as the second centroid.

[0080] Let the distance of each sample in the standard greenhouse environment dataset from the first centroid c1 be denoted as d(E i , c1), and:

[0081]

[0082] Use the probability calculation formula to represent the selection of the second centroid c2:

[0083]

[0084] Among them, the sample corresponding to the largest value of P(c2) is selected as the second centroid c2 (that is, calculate the distance of each sample in the standard greenhouse environment dataset from the first centroid, and select the sample with the farthest distance as the second centroid). represents the sum of the distances of all samples from the first centroid c1, E i represents sample i in the standard greenhouse environment dataset E, E j represents the standard greenhouse environment dataset E d in sample j.

[0085] Then, calculate the distances of each sample in the standard greenhouse environment dataset from the first centroid and the second centroid respectively, and assign each sample to the centroid with the closest distance to form their respective clusters.

[0086] Let the distance of each sample in the standard greenhouse environment dataset from the second centroid c2 be denoted as d(E i , c2), and

[0087]

[0088] Continue to select the remaining centroids from the standard greenhouse environment dataset, calculate the distances from each sample in the standard greenhouse environment dataset to each centroid, and assign each sample to the nearest centroid to form their respective clusters. The selection method for the remaining centroids is as follows: Calculate the sample that is farthest from its centroid in each determined cluster as the next centroid.

[0089] Represent the selection of the next centroid c using the probability calculation formula q :

[0090]

[0091] Among them, select the sample corresponding to the largest value of P(c q ) as the next centroid c q (Calculate the sample that is farthest from its centroid in each determined cluster as the next centroid), d(E r , c q ) represents the distance from a sample in the confirmed cluster to its centroid, represents the sum of the distances from all samples in the confirmed cluster to their centroids, E q represents all confirmed clusters, E r represents the sample r in the confirmed cluster, E t represents the sample t in the confirmed cluster;

[0092] Assume that each centroid is represented as c i ;

[0093] According to formulas (1) and (2), the distances from each sample to each centroid are expressed by the formula:

[0094]

[0095] Until the number of selected centroids reaches the minimum boundary value of 3 within the range of the target number of centroids.

[0096] S5. Increase the target number of centroids by one, repeat S4 until the target number of centroids reaches the maximum boundary value within the range of the target number of centroids, and select the target centroid with the largest average silhouette coefficient as the optimal clustering.

[0097]

[0098] And,

[0099] -1 ≤ S i ≤ 1;

[0100] Among them, S i represents the silhouette coefficient of sample i (the closer the silhouette coefficient is to 1, the relatively better the cohesion and separation), a i represents the average distance of a sample to other samples within its cluster, bi represents the minimum value of the average distance between a sample and the samples of other clusters, max(a i , b i ) represents selecting the maximum value between a i and b i .

[0101] Calculate the average silhouette coefficient of all samples in the standard greenhouse environment dataset through the silhouette coefficient, and then increase the target centroid number by one. At this time, the target centroid is 4. Continue to repeat step S4 until the target centroid number reaches the maximum boundary value 5 within the range of the target centroid number, and select the target centroid with the largest average silhouette coefficient as the optimal clustering.

[0102] S6. Establish a greenhouse soil water balance model. Determine which cluster of the optimal clustering result it belongs to according to the greenhouse environment information of the previous day in the greenhouse, and select the change amount of soil water content corresponding to the centroid as the reference value and input it into the model prediction controller of the irrigation system for prediction. The obtained value is input into the greenhouse soil water balance model, and the result calculated by the greenhouse soil water balance model is used as the actual irrigation amount.

[0103] Specifically, the greenhouse soil water balance model is as follows:

[0104]

[0105] And,

[0106] ET c = K c ·ET o ;

[0107]

[0108]

[0109] Among them, I(t) represents the irrigation amount, with the unit of cubic centimeter per hour, ET c (t) represents the water lost due to crop evapotranspiration, with the unit of millimeter per hour, θ(t) represents the soil water content, with the unit of cubic centimeter per hour, q1, q2, q3 are estimated by the least squares method based on the directly measured initial soil irrigation amount, water lost due to crop evapotranspiration, and water content respectively, ET o represents the crop reference evapotranspiration, K c represents the crop coefficient, d, d1, d2 are all the days of crop growth, K ini is the crop coefficient at the initial stage of recommended crop growth, K mid is the crop coefficient at the middle stage of recommended crop growth, K endGenerate the crop coefficient at the end stage of the recommended crop. RH represents the average daily relative humidity in the greenhouse (calculated after passing through the air humidity sensor), h represents the average height of the crop growth in the initial, middle, and end stages, in meters, and R a represents the extraterrestrial radiation, in megajoules per square meter per day, τ represents the ratio of solar radiation inside and outside (for the experimental greenhouse with polyvinyl chloride film, τ is taken as 0.8), T, T max and T min respectively represent the average, maximum, and minimum greenhouse air temperatures, in degrees Celsius. Since the air flow in the greenhouse is less, the influence of the daily average wind speed u2 at a height of two meters on the crop coefficient is not considered. Taking tomatoes as an example, K ini = 0.5, K mid = 1.05, K end = 0.7, d1 = 30, d2 = 70.

[0110] It should be noted that when the soil water content is less than the maximum water holding capacity of the soil θ fc , almost no deep percolation occurs, and the soil water consumption can be approximately regarded as the absorption amount of the crops. Therefore, the change amount of the soil water content can be used as a reference value for the irrigation amount.

[0111] Adopt a model predictive controller to achieve precise control of the soil water content. According to the average light intensity, average air temperature, and average air humidity of the greenhouse the previous day, determine which cluster of the optimal clustering results it belongs to, and select the change amount of the soil water content at the centroid of this cluster as the crop water absorption amount of the previous day, that is, the irrigation amount of the current day, as the input reference quantity of the model predictive controller. The output result of the model predictive controller is input into the greenhouse soil water balance model for calculation, and the calculated result is fed back to the model predictive controller as the actual irrigation amount, and irrigation is carried out by controlling the irrigation solenoid valve through the upper computer and the lower computer. Repeat continuously until the actual change amount of the soil water content is equal to the reference value input into the model predictive controller.

[0112] When predicting, set three parameters for the model predictive controller, and the three parameters are the prediction horizon, control horizon, and sampling period respectively.

[0113] In this embodiment, the irrigation time is set at 6 o'clock every day. Determine that the prediction horizon P is 20, determine that the control horizon C is 10, and determine that the sampling period T s for irrigation is 5 minutes, and predict the system state for the next 20 control cycles in sequence. Among them, the values of P, C, and T s are determined according to the actual situation.

[0114] In order to prevent the soil water content from being too low, finally, calibration is carried out every 5 days. When the soil texture in the greenhouse is loam, if the soil water content is lower than the maximum water holding capacity of the soil θ fc , that is, exceeding 0.33 m3 ·m -3 When it is (generally specified), it is supplemented to the maximum water holding capacity of the soil. If the soil water content is higher than the maximum water holding capacity θ of the soil fc , no irrigation is carried out.

[0115] The weighted normalization, multi-objective centroid method and average silhouette coefficient are used to optimize the clustering for the growth characteristics of greenhouse crops, and the change amount of soil water content corresponding to the centroid of each cluster is obtained, that is, the water requirement of crops in different environments in the greenhouse; a soil water balance model suitable for the greenhouse is established, different crop coefficients are determined according to different growth periods of crops, the water lost by crop evapotranspiration is calculated more accurately, and the result obtained by the optimal clustering is used as the reference input of the model predictive controller to avoid irrigation waste caused by relying on experience.

[0116] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0117] The above-described embodiments only express the embodiments of the present application that are described more specifically and in detail, but cannot be understood as a limitation on the scope of the patent application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A greenhouse water-saving irrigation method based on multi-information fusion, applied to an irrigation system, characterized in that: The greenhouse water-saving irrigation method based on multi-information fusion includes: S1. Collect data on light intensity, air temperature, air humidity, and soil water content in the greenhouse at fixed intervals, and calculate the average light intensity, average air temperature, average air humidity, and soil water content change in the greenhouse every day to obtain the daily environmental information in the greenhouse. S2. Perform weighted normalization on the average light intensity, average air temperature, average air humidity, and soil water content change every day to obtain a standard greenhouse environment dataset. S3. Cluster the standard greenhouse environment dataset, taking each day in the standard greenhouse environment dataset as a sample, set the range of the target number of centroids, and define the initial target centroid as the minimum boundary value in the range of the target number of centroids. S4. Select centroids from the standard greenhouse environment dataset in sequence until the number of selected centroids reaches the minimum boundary value in the range of the target number of centroids, and then calculate the average silhouette coefficient of all samples in the standard greenhouse environment dataset. Among them, select the sample with the average light intensity closest to the average value of the average light intensity of all samples from the standard greenhouse environment dataset as the first centroid. S5. Increase the target number of centroids by one, repeat S4 until the target number of centroids reaches the maximum boundary value in the range of the target number of centroids, and select the target centroid with the largest average silhouette coefficient as the optimal clustering. S6. Establish a greenhouse soil water balance model, determine which cluster of the optimal clustering result it belongs to according to the greenhouse environmental information of the previous day in the greenhouse, and select the soil water content change of the corresponding centroid as a reference value to input into the model prediction controller of the irrigation system for prediction. The obtained value is input into the greenhouse soil water balance model, and the result calculated by the greenhouse soil water balance model is used as the actual irrigation amount. Among them, the establishment of the greenhouse soil water balance model includes: The greenhouse soil water balance model is as follows: And, ET c = K c ·ET o ; Among them, I(t) represents the irrigation amount, with the unit of cubic centimeter per hour, and ET c (t) represents the water lost due to crop evapotranspiration, with the unit of millimeter per hour, θ(t) represents the soil water content, with the unit of cubic centimeter per hour, q1, q2, and q3 are obtained by directly measuring the initial soil irrigation amount, the water lost due to crop evapotranspiration, and the water content and using the least squares method. ET o represents the crop reference evapotranspiration, and K c represents the crop coefficient, d, d1, and d2 are all the days of crop growth, and K ini is the crop coefficient at the initial stage of recommended crop growth, and K mid is the crop coefficient at the middle stage of recommended crop growth, and K end is the crop coefficient at the end stage of recommended crop growth. RH represents the average daily relative humidity in the greenhouse, h represents the average height of crop growth at the initial, middle, and end stages, with the unit of meter, and R a represents the extraterrestrial radiation, with the unit of megajoule per square meter per day, τ represents the ratio of solar radiation inside and outside, and T, T max and T min represent the average, maximum, and minimum greenhouse air temperatures respectively, with the unit of degree Celsius.

2. The greenhouse water-saving irrigation method based on multi-information fusion according to claim 1, characterized in that: The weighted normalization of the average light intensity, average air temperature, average air humidity, and soil water content change every day to obtain a standard greenhouse environment dataset includes: Let the average light intensity per day be L m and the average air temperature per day be T m and the average air humidity per day be H m and the change amount of soil water content per day be M d and the environmental information set per day be D, and the weights of the average light intensity, average air temperature, average air humidity, and the change amount of soil water content per day are w l , w t , w h , w m , respectively, and the weight of the environmental information set per day is w d . If the standard greenhouse environmental data set is E, then the environmental information D of one day in the environmental information set D i ={L mi , T mi , H mi , M di}; The weighted normalization of the average light intensity, average air temperature, average air humidity, and soil water content change every day to obtain a standard greenhouse environment dataset includes: E = {L e , T e , H e , M e}; Among them, L e 、 T e 、 H e 、 M e respectively represent the average light intensity, average air temperature, average air humidity, and change amount of soil water content after weighted normalization processing. L mi 、 T mi 、 H mi 、 M di respectively represent a certain element in D. L mmin 、 T mmin 、 H mmin 、 M dmin respectively represent the minimum values of the corresponding similar elements in D. L mmax 、 T mmax 、 H mmax 、 M dmax respectively represent the maximum values of the corresponding similar elements in D. L ei 、 T ei 、 H ei 、 M ei respectively represent the corresponding values in E of the average light intensity, average air temperature, average air humidity, and change amount of soil water content in D for one day.

3. The greenhouse water-saving irrigation method based on multi-information fusion according to claim 2, characterized in that: The clustering of the standard greenhouse environment dataset, taking each day in the standard greenhouse environment dataset as a sample, setting the range of the target number of centroids, and defining the initial target centroid as the minimum boundary value in the range of the target number of centroids includes: Set the range of the number of target centroids to K1 - K n , and define the initial target centroid as the minimum boundary value K1 in the range of the number of target centroids.

4. The greenhouse water-saving irrigation method based on multi-information fusion according to claim 3, characterized in that: The selection of centroids from the standard greenhouse environment dataset in sequence until the number of selected centroids reaches the minimum boundary value in the range of the target number of centroids includes: Express the selection of the first centroid c1 using a probability calculation formula: Among them, the sample corresponding to the largest value of P(c1) is selected as the first centroid c1. represents the sum of the average light intensities of all samples, L ei represents the average light intensity of sample i in the standard greenhouse environment dataset E, L ej represents the average light intensity of sample j in the standard greenhouse environment dataset E, and n represents the total number of samples; Then calculate the distance between each sample in the standard greenhouse environment dataset and the first centroid, and select the sample with the farthest distance as the second centroid. Let the distance of each sample in the standard greenhouse environment dataset from the first centroid c1 be denoted as d(E i , c1), and: Express the selection of the second centroid c2 using a probability calculation formula: Among them, the sample corresponding to the largest value of P(c2) is selected as the second centroid c2. represents the sum of the distances of all samples from the first centroid c1, E i represents the sample i in the standard greenhouse environment dataset E, E j represents the standard greenhouse environment dataset E d and the sample j in it; Then calculate the distances between each sample in the standard greenhouse environment dataset to the first centroid and the second centroid respectively, and assign each sample to the centroid with the closest distance to form their respective clusters. Let the distance of each sample in the standard greenhouse environment dataset from the second centroid c2 be denoted as d(E i , c2), and Continue to select the remaining centroids from the standard greenhouse environment dataset, calculate the distances from each sample in the standard greenhouse environment dataset to each centroid respectively, assign each sample to the nearest centroid to form their respective clusters, and the selection method of the remaining centroids is: calculate the sample with the farthest distance from its centroid in each determined cluster as the next centroid; Express the selection of the next centroid c using the probability calculation formula q : Among them, select the sample corresponding to the maximum value of P(c q ) as the next centroid c q , d(E r , c q ) represents the distance between a sample in the confirmed cluster and its centroid, represents the sum of the distances between all samples in the confirmed cluster and their centroids, E q represents all confirmed clusters, E r represents the sample r in the confirmed cluster, E t represents the sample t in the confirmed cluster; Let each centroid be denoted as c i ; The distances from each sample to each centroid are expressed by formulas according to formulas (1) and (2): Until the number of selected centroids reaches the minimum boundary value K1 in the range of the target number of centroids.

5. The greenhouse water-saving irrigation method based on multi-information fusion according to claim 3, characterized in that: The maximum boundary value in the range of the target centroid number is K n .

Images