Greenhouse environment control method and system based on artificial intelligence

Abstract

The invention relates to the field of artificial intelligence, in particular to a greenhouse environment control method and system based on artificial intelligence. The method comprises the following steps: acquiring greenhouse environment data and plant growth stage data in a sensor, performing data preprocessing on the greenhouse environment data and the plant growth stage data, and performing correlation analysis on the plant growth environment data by using an Apriori correlation analysis algorithm; the method comprises the steps of training and testing a CNN-LSTM greenhouse environment prediction model by using initial plant growth environment data, optimizing the initial CNN-LSTM greenhouse environment prediction model by using an IPSO improved particle swarm optimization algorithm, generating a greenhouse adjustment instruction according to a greenhouse environment prediction state, and performing real-time control on a sensor according to the greenhouse adjustment instruction. Accurate prediction and abnormal early warning of plant growth are achieved, a planter can formulate scientific and reasonable cultivation management measures according to a personalized model, and the yield and quality of crops are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of greenhouse control, and in particular to an artificial intelligence-based greenhouse environment control method and system. Background Art

[0002] In traditional greenhouse cultivation, environmental control often relies on fixed experience values ​​and manual operations, which makes it difficult to accurately adapt to the dynamic needs of plant growth and complex and changing environmental factors in real time. With the development of science and technology, although sensor technology can obtain greenhouse environmental data, how to effectively analyze and use this data to achieve precise control is still a challenge. Existing greenhouse control methods are insufficient in data processing and model prediction accuracy, and cannot fully consider the complex correlation between environmental factors and the dynamic changes in plant growth stages, resulting in energy waste, irrational resource utilization, and unstable crop growth quality. Therefore, how to control the greenhouse more intelligently, accurately predict environmental changes in the greenhouse and achieve precise automatic control, and realize energy regulation are technical problems that need to be solved at this stage. Summary of the invention

[0003] The purpose of the present invention is to solve the above problems and to design a greenhouse environment control method and system based on artificial intelligence.

[0004] To achieve the above object, the technical solution of the present invention is that, further, in the above-mentioned greenhouse environment control method based on artificial intelligence, the greenhouse environment control method comprises the following steps:

[0005] Acquire greenhouse environment data and plant growth stage data from the sensor, perform data preprocessing on the greenhouse environment data and plant growth stage data, and obtain plant growth environment data;

[0006] Performing association analysis on the plant growth environment data using the Apriori association analysis algorithm to obtain initial plant growth environment data;

[0007] A CNN-LSTM greenhouse environment prediction model is established based on a CNN convolutional neural network and an LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model;

[0008] The IPSO improved particle swarm algorithm is used to optimize the initial CNN-LSTM greenhouse environment prediction model to obtain an IPSO-CNN-LSTM greenhouse environment prediction model;

[0009] Acquire real-time plant growth environment data in the greenhouse, input the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtain a greenhouse environment prediction state;

[0010] A greenhouse adjustment instruction is generated according to the greenhouse environment prediction state, and the sensor is controlled in real time according to the greenhouse adjustment instruction.

[0011] Furthermore, in the above-mentioned greenhouse environment control method based on artificial intelligence, the greenhouse environment data and plant growth stage data in the sensor are acquired, and data preprocessing is performed on the greenhouse environment data and plant growth stage data to obtain plant growth environment data, including:

[0012] By using various sensors arranged in the greenhouse, greenhouse environmental data is obtained in real time, wherein the greenhouse environmental data at least includes temperature data, humidity data, light intensity data and carbon dioxide concentration data;

[0013] Recording growth stage data of plants in the greenhouse, wherein the growth stage data at least includes germination stage, seedling stage, growth stage, flowering stage and fruiting stage;

[0014] Cleaning the greenhouse environment data and the growth stage data to remove abnormal values ​​and noise values ​​in the data;

[0015] The data normalization method is used to map data of different dimensions to the interval [0,1] to obtain plant growth environment data.

[0016] Furthermore, in the above-mentioned greenhouse environment control method based on artificial intelligence, the Apriori association analysis algorithm is used to perform association analysis on the plant growth environment data to obtain initial plant growth environment data, including:

[0017] The minimum support threshold and the minimum confidence threshold are set, and the plant growth environment data are scanned using the Apriori association analysis algorithm, and the number of occurrences of each item set is counted to obtain a frequent 1-item set;

[0018] Assume that the plant growth environment data contains N transactions. For an item set X, its support calculation formula is as follows:

[0019]

[0020] in, Represents the number of transactions that contain item set X in the data set;

[0021] For the association rule X→Y, the confidence calculation formula is:

[0022]

[0023] Generate candidate frequent 2-item sets by concatenation and pruning, scan the data set again, calculate the support of candidate frequent 2-item sets, and screen out frequent 2-item sets;

[0024] Continue to generate higher-order frequent item sets until no item sets that meet the minimum support threshold are found;

[0025] The association rules that meet the minimum confidence threshold are extracted from the frequent item sets to obtain the initial plant growth environment data.

[0026] Furthermore, in the above-mentioned greenhouse environment control method based on artificial intelligence, a CNN-LSTM greenhouse environment prediction model is established based on a CNN convolutional neural network and an LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model, including:

[0027] The CNN-LSTM greenhouse environment prediction model is established based on the CNN convolutional neural network and the LSTM long short-term memory network;

[0028] The calculation formula of the forget gate of the LSTM network in the CNN-LSTM greenhouse environment prediction model is as follows:

[0029] f t =σ(W f ·[H t-1 , X t ]+b f )

[0030] Among them, f t is the forget gate, σ is the sigmoid activation function, X t is the input at time t, H t-1 is the short-term memory output at time t-1, W f is the forget gate weight, b f Bias for the forget gate;

[0031] The input gate is based on the X of the network output node t and H t-1 To decide what new content needs to be added to memory, the calculation formula is as follows:

[0032] i t =σ(W i ·[H t-1 , X t ]+b i )

[0033]

[0034] Among them, it is the input gate, is the candidate vector, tanh is the hyperbolic tangent activation function, C t represents the long-term memory unit at time t, W i and W c are the weight parameters of the input gate and candidate vector, respectively, i and b c is the corresponding bias term;

[0035] The output gate is based on the long short-term memory unit C t , input node X t and H t-1 The output value is Ht, and its calculation formula is as follows:

[0036] ot=σ(W o ·[H t-1 , X t ]+b o )

[0037] Ht=ot×tanh(Ct)

[0038] Where ot is the output gate, W o is the output gate weight, b o is the output gate bias term, Ht is all the outputs of the model at time t;

[0039] The initial plant growth environment data is divided into a training set and a test set according to a ratio of 9:1, and the established CNN-LSTM greenhouse environment prediction model is trained and tested to obtain an initial CNN-LSTM greenhouse environment prediction model.

[0040] Further, in the above-mentioned greenhouse environment control method based on artificial intelligence, the IPSO improved particle swarm algorithm is used to optimize the initial CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model, including:

[0041] The initial CNN-LSTM greenhouse environment prediction model is optimized by using the IPSO improved particle swarm algorithm, and the particle population size, number of iterations, learning factor, position and speed of the IPSO improved particle swarm algorithm are determined;

[0042] Initialize the position and velocity of the particles and randomly generate a population x i =(θ, β, a1, a2), where θ is the number of iterations of the CNN-LSTM greenhouse environment prediction model, β is the batch size, and a1 and a2 are the number of neurons in the first hidden layer and the second hidden layer respectively;

[0043] The fitness function of the IPSO improved particle swarm algorithm is determined, and the CNN-LSTM greenhouse environment prediction model is constructed using the initialized particle swarm parameters. The fitness function δ is defined as follows:

[0044]

[0045] Among them, N and M are the number of samples for training and testing, y i t and i v are training data and test data, and is the predicted value of the training sample and the test sample;

[0046] Update the speed and position of the particle, and determine the local optimal value P according to the fitness of the particle t i and the global optimal value G t ;

[0047] The optimal number of iterations, batch size, and number of neurons in the hidden layer found by the IPSO improved particle swarm algorithm are brought into the CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model.

[0048] Furthermore, in the above-mentioned greenhouse environment control method based on artificial intelligence, the IPSO improved particle swarm algorithm includes:

[0049] The learning factor c in the original PSO particle swarm algorithm 1 Set from large to small, learning factor c 2 Assuming the position is from small to large, the formula is as follows:

[0050]

[0051] The inertia coefficient ω in the PSO particle swarm algorithm is set as a nonlinear inertia weight, and the improved form of the inertia coefficient ω is as follows:

[0052]

[0053] Among them, ω max is the maximum value of the inertia coefficient, ω min is the minimum value of the inertia coefficient, t is the current iteration number, t max is the maximum number of iterations;

[0054] Set the adaptive mutation mechanism in the PSO particle swarm algorithm. After each particle update, set the threshold ρ. When the particle meets the mutation condition, mutation occurs. The calculation formula of the adaptive mutation mechanism is as follows:

[0055]

[0056] rand>ρ

[0057] Among them, ρ is the threshold of the adaptive mutation mechanism, rand is a random number between [0,1]; after learning factor c 1 , learning factor c 2 After the inertia coefficient ω and the adaptive mutation mechanism are improved, the IPSO improved particle swarm algorithm is obtained.

[0058] Furthermore, in the above-mentioned greenhouse environment control method based on artificial intelligence, the greenhouse adjustment instruction is generated according to the greenhouse environment prediction state, and the sensor is controlled in real time according to the greenhouse adjustment instruction, including:

[0059] Generate a greenhouse adjustment instruction according to the greenhouse environment prediction state, wherein the greenhouse environment prediction state includes at least a temperature prediction value, a humidity prediction value, a light intensity prediction value, and a carbon dioxide concentration prediction value;

[0060] The absolute deviation is used to calculate the difference between the predicted value and the boundary value of the suitable range, and the environmental parameters are prioritized according to the size of the difference;

[0061] If the difference in the temperature prediction value is large, a ventilation equipment opening control instruction is generated, and the ventilation equipment is used for air convection to reduce the temperature to a temperature range of 22.5°C-25°C;

[0062] If the light difference is large, a sunshade control instruction is generated to use the sunshade to block the light. If the light is insufficient, the fill light is turned on to supplement the light.

[0063] If the humidity difference is large, turn on the ventilation equipment for air convection. If the humidity is too low, start the irrigation system to spray water appropriately.

[0064] If the difference in carbon dioxide concentration is large, the carbon dioxide generator control instruction is issued by increasing the release of carbon dioxide from the carbon dioxide generator.

[0065] Furthermore, in the above-mentioned greenhouse environment control system based on artificial intelligence, the communication network traffic optimization system includes the following modules:

[0066] A data acquisition module, used to acquire greenhouse environment data and plant growth stage data from the sensor, and perform data preprocessing on the greenhouse environment data and plant growth stage data to obtain plant growth environment data;

[0067] An association analysis module is used to perform association analysis on the plant growth environment data using an Apriori association analysis algorithm to obtain initial plant growth environment data;

[0068] A model building module is used to establish a CNN-LSTM greenhouse environment prediction model based on a CNN convolutional neural network and an LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model;

[0069] A model optimization module is used to optimize the initial CNN-LSTM greenhouse environment prediction model by using an IPSO improved particle swarm algorithm to obtain an IPSO-CNN-LSTM greenhouse environment prediction model;

[0070] An environmental prediction module is used to obtain real-time plant growth environment data in the greenhouse, input the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtain a greenhouse environment prediction state;

[0071] The environment control module is used to generate a greenhouse adjustment instruction according to the predicted state of the greenhouse environment, and to control the sensor in real time according to the greenhouse adjustment instruction.

[0072] Furthermore, in the above-mentioned greenhouse environment control system based on artificial intelligence, the data acquisition module includes the following submodules:

[0073] An acquisition submodule is used to acquire greenhouse environmental data in real time through various sensors arranged in the greenhouse, wherein the greenhouse environmental data at least includes temperature data, humidity data, light intensity data and carbon dioxide concentration data;

[0074] A recording submodule, used to record the growth stage data of plants in the greenhouse, wherein the growth stage data at least includes the germination stage, the seedling stage, the growth stage, the flowering stage and the fruiting stage;

[0075] A cleaning submodule, used to clean the greenhouse environment data and the growth stage data to remove abnormal values ​​and noise values ​​in the data;

[0076] The mapping submodule is used to map data of different dimensions to the [0,1] interval using a data normalization method to obtain plant growth environment data.

[0077] Furthermore, in the above-mentioned greenhouse environment control system based on artificial intelligence, the association analysis module includes the following submodules:

[0078] A setting submodule is used to set a minimum support threshold and a minimum confidence threshold, use the Apriori association analysis algorithm to scan the plant growth environment data, count the number of occurrences of each item set, and obtain a frequent 1-item set;

[0079] The calculation submodule is used to assume that the plant growth environment data contains N transactions. For an item set X, its support calculation formula is as follows:

[0080]

[0081] in, Represents the number of transactions that contain item set X in the data set;

[0082] For the association rule X→Y, the confidence calculation formula is:

[0083]

[0084] The generation submodule is used to generate candidate frequent 2-item sets by concatenation and pruning, and scan the data set again to calculate the support of the candidate frequent 2-item sets and filter out the frequent 2-item sets;

[0085] Iteration submodule, used to continuously generate higher-order frequent itemsets until no itemsets that meet the minimum support threshold are found;

[0086] The extraction submodule is used to extract association rules that meet the minimum confidence threshold from the frequent item set to obtain initial plant growth environment data.

[0087] The invention has the beneficial effects of obtaining greenhouse environment data and plant growth stage data from a sensor, performing data preprocessing on the greenhouse environment data and plant growth stage data, and obtaining plant growth environment data; performing association analysis on the plant growth environment data using an Apriori association analysis algorithm, and obtaining initial plant growth environment data; establishing a CNN-LSTM greenhouse environment prediction model based on a CNN convolutional neural network and an LSTM long short-term memory network; training and testing the CNN-LSTM greenhouse environment prediction model using the initial plant growth environment data, and obtaining an initial CNN-LSTM greenhouse environment prediction model; optimizing the initial CNN-LSTM greenhouse environment prediction model using an IPSO improved particle swarm algorithm, and obtaining an IPSO-CNN-LSTM greenhouse environment prediction model; obtaining real-time plant growth environment data in a greenhouse, inputting the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtaining a greenhouse environment prediction state; generating a greenhouse adjustment instruction according to the greenhouse environment prediction state, and controlling the sensor in real time according to the greenhouse adjustment instruction. 1. It realizes accurate prediction of plant growth and early warning of abnormalities. Growers can formulate scientific and reasonable cultivation management measures according to personalized models, significantly improving crop yield and quality. 2. The intelligent optimization model of energy consumption based on artificial intelligence algorithm can adjust equipment operation in real time according to the actual situation of the greenhouse, minimize energy consumption, improve energy utilization efficiency, and reduce greenhouse operating costs. 3. Precise management of water resources realizes accurate irrigation decision-making through the analysis of multi-source data by deep learning model, avoids the waste of water resources, and reduces the problems of soil nutrient loss and disease spread caused by excessive irrigation, which is beneficial to protecting the environment and healthy plant growth. 4. The greenhouse intelligent control system of the present invention improves the overall automation and intelligence level of greenhouse production, reduces the intensity of manual labor, improves production efficiency and economic benefits, and provides strong technical support for the development of modern agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0089] Figure 1 This is a schematic diagram of a first embodiment of a greenhouse environment control method based on artificial intelligence in an embodiment of the present invention;

[0090] Figure 2 Schematic diagram of a second embodiment of a greenhouse environment control method based on artificial intelligence in an embodiment of the present invention;

[0091] Figure 3 Schematic diagram of a third embodiment of a greenhouse environment control method based on artificial intelligence in an embodiment of the present invention;

[0092] Figure 4 This is a schematic diagram of a first embodiment of a greenhouse environment control system based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0093] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0094] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0095] The present invention will be described in detail below in conjunction with the accompanying drawings. Figure 1 As shown, a greenhouse environment control method based on artificial intelligence, the greenhouse environment control method comprises the following steps:

[0096] Step 101, obtaining greenhouse environment data and plant growth stage data from a sensor, performing data preprocessing on the greenhouse environment data and the plant growth stage data, and obtaining plant growth environment data;

[0097] Specifically, in this embodiment, various sensors arranged in the greenhouse are used to obtain greenhouse environmental data in real time, and the greenhouse environmental data includes at least temperature data, humidity data, light intensity data and carbon dioxide concentration data;

[0098] Record the growth stage data of plants in the greenhouse, which at least includes the germination stage, seedling stage, growth stage, flowering stage and fruiting stage;

[0099] Clean the greenhouse environment data and growth stage data to remove outliers and noise values ​​in the data;

[0100] The data normalization method is used to map data of different dimensions to the interval [0,1] to obtain plant growth environment data.

[0101] Step 102: Perform association analysis on the plant growth environment data using the Apriori association analysis algorithm to obtain initial plant growth environment data;

[0102] Specifically, in this embodiment, a minimum support threshold and a minimum confidence threshold are set, and the plant growth environment data are scanned using the Apriori association analysis algorithm to count the number of occurrences of each item set to obtain a frequent 1-item set;

[0103] Assume that the plant growth environment data contains N transactions. For an item set X, its support calculation formula is as follows:

[0104]

[0105] in, Represents the number of transactions that contain item set X in the data set;

[0106] For the association rule X→Y, the confidence calculation formula is:

[0107]

[0108] Generate candidate frequent 2-item sets by concatenation and pruning, scan the data set again, calculate the support of candidate frequent 2-item sets, and screen out frequent 2-item sets;

[0109] Continue to generate higher-order frequent item sets until no item sets that meet the minimum support threshold are found;

[0110] The association rules that meet the minimum confidence threshold are extracted from the frequent item sets to obtain the initial plant growth environment data.

[0111] The generated association rules are interpreted and analyzed to understand the potential relationship between the environmental factors represented by each rule and the plant growth stage. For example, a rule "When the temperature is in the interval [X1, X2] and the light intensity is in the interval [Y1, Y2], the probability of the plant being in the flowering stage is high" can provide guidance for greenhouse environmental regulation, that is, under such environmental conditions, corresponding measures may need to be taken to promote plant flowering.

[0112] Based on the results of the association rules, greenhouse management strategies can be developed. For example, if a certain growth stage is found to be strongly associated with a specific combination of environmental factors, the focus can be on regulating these environmental factors during that growth stage, and by adjusting heating, ventilation, lighting and other equipment, environmental conditions can be made more in line with plant growth needs to improve crop yield and quality.

[0113] The results of association analysis are applied to the greenhouse intelligent control system to achieve automated environmental regulation. For example, when the sensor detects that the current environmental data meets the prerequisites of a certain association rule, the system can automatically issue control instructions to adjust the operating parameters of the greenhouse equipment to create an environment more conducive to plant growth.

[0114] Step 103: Establish a CNN-LSTM greenhouse environment prediction model based on the CNN convolutional neural network and the LSTM long short-term memory network; use the initial plant growth environment data to train and test the CNN-LSTM greenhouse environment prediction model to obtain an initial CNN-LSTM greenhouse environment prediction model;

[0115] Specifically, in this embodiment, a CNN-LSTM greenhouse environment prediction model is established based on a CNN convolutional neural network and an LSTM long short-term memory network;

[0116] The calculation formula of the forget gate of the LSTM network in the CNN-LSTM greenhouse environment prediction model is as follows:

[0117] f t =σ(W f ·[H t-1 , X t ]+b f )

[0118] Among them, f t is the forget gate, σ is the sigmoid activation function, X t is the input at time t, H t-1 is the short-term memory output at time t-1, W f is the forget gate weight, b f Bias for the forget gate;

[0119] The input gate is based on the X of the network output node t and H t-1 To decide what new content needs to be added to memory, the calculation formula is as follows:

[0120] i t =σ(W i ·[H t-1 , X t ]+b i )

[0121]

[0122] Among them, i t is the input gate, is the candidate vector, tanh is the hyperbolic tangent activation function, C t represents the long-term memory unit at time t, W i and W c are the weight parameters of the input gate and candidate vector, respectively, i and b c is the corresponding bias term;

[0123] The output gate is based on the long short-term memory unit C t , input node X t and Ht-1 The output value is Ht, and its calculation formula is as follows:

[0124] 0t=σ(W o ·[H t-1 , X t ]+b o )

[0125] Ht=ot×tanh(Ct)

[0126] Where ot is the output gate, W o is the output gate weight, b o is the output gate bias term, Ht is all the outputs of the model at time t;

[0127] The initial plant growth environment data were divided into training set and test set according to a ratio of 9:1, and the established CNN-LSTM greenhouse environment prediction model was trained and tested to obtain the initial CNN-LSTM greenhouse environment prediction model.

[0128] Step 104, using the IPSO improved particle swarm algorithm to optimize the initial CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model;

[0129] Specifically, in this embodiment, the learning factor c in the original PSO particle swarm algorithm is 1 Set from large to small, learning factor c 2 Assuming the position is from small to large, the formula is as follows:

[0130]

[0131] The inertia coefficient ω in the PSO particle swarm algorithm is set as a nonlinear inertia weight, and the improved form of the inertia coefficient ω is as follows:

[0132]

[0133] Among them, ω max is the maximum value of the inertia coefficient, ω min is the minimum value of the inertia coefficient, t is the current iteration number, t max is the maximum number of iterations;

[0134] Set the adaptive mutation mechanism in the PSO particle swarm algorithm. After each particle update, set the threshold ρ. When the particle meets the mutation condition, mutation occurs. The calculation formula of the adaptive mutation mechanism is as follows:

[0135]

[0136] rand>ρ

[0137] Among them, ρ is the threshold of the adaptive mutation mechanism, rand is a random number between [0,1], which can be 0, 0.1, 0.2, 0.3; after learning factor c 1 , learning factor c 2 After the inertia coefficient ω and the adaptive mutation mechanism are improved, the IPSO improved particle swarm algorithm is obtained.

[0138] Specifically, in this embodiment, the IPSO improved particle swarm algorithm is used to optimize the initial CNN-LSTM greenhouse environment prediction model, and the particle population size, number of iterations, learning factor, position and speed of the IPSO improved particle swarm algorithm are determined;

[0139] Initialize the position and velocity of the particles and randomly generate a population x i =(θ,β,a1,a2), where θ is the number of iterations of the CNN-LSTM greenhouse environment prediction model, β is the batch size, and a1 and a2 are the number of neurons in the first hidden layer and the second hidden layer respectively;

[0140] Determine the fitness function of the IPSO improved particle swarm algorithm, and use the initialized particle swarm parameters to build the CNN-LSTM greenhouse environment prediction model. The fitness function δ is defined as follows:

[0141]

[0142] Among them, N and M are the number of samples for training and testing, y i t and i v are training data and test data, and is the predicted value of the training sample and the test sample;

[0143] Update the speed and position of the particle, and determine the local optimal value P according to the fitness of the particle t i and the global optimal value G t ;

[0144] The optimal number of iterations, batch size, and number of neurons in the hidden layer found by the IPSO improved particle swarm algorithm were brought into the CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model.

[0145] Step 105, obtaining real-time plant growth environment data in the greenhouse, inputting the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtaining a greenhouse environment prediction state;

[0146] Specifically, in this embodiment, real-time plant growth environment data is input into the IPSO-RNN-LSTM greenhouse environment prediction model for training. During the training process, the model calculates the predicted value based on the input data and the current parameters, and then calculates the loss between the predicted value and the actual value through a loss function (such as a mean square error function). According to the loss value, the gradient is calculated using the back propagation algorithm, and the parameters of the model are updated through the Adam optimizer to reduce the loss. This process is repeated until the performance of the model on the training set and the validation set is stable or meets the preset requirements.

[0147] After the training is completed, the initial plant growth environment data in the test set is input into the trained model to obtain the greenhouse environment prediction state. The prediction state includes the change trend and prediction value of greenhouse environment parameters (temperature, humidity, light, etc.) in the future.

[0148] Step 106: Generate a greenhouse adjustment instruction according to the predicted state of the greenhouse environment, and control the sensor in real time according to the greenhouse adjustment instruction.

[0149] Specifically, in this embodiment, a greenhouse adjustment instruction is generated according to the greenhouse environment prediction state, and the greenhouse environment prediction state includes at least a temperature prediction value, a humidity prediction value, a light intensity prediction value, and a carbon dioxide concentration prediction value;

[0150] The absolute deviation is used to calculate the difference between the predicted value and the boundary value of the suitable range, and the environmental parameters are prioritized according to the size of the difference;

[0151] If the difference in the temperature prediction value is large, a ventilation equipment opening control instruction is generated, and the ventilation equipment is used for air convection to reduce the temperature to a temperature range of 22.5°C-25°C;

[0152] If the light difference is large, a sunshade control instruction is generated to use the sunshade to block the light. If the light is insufficient, the fill light is turned on to supplement the light.

[0153] If the humidity difference is large, turn on the ventilation equipment for air convection. If the humidity is too low, start the irrigation system to spray water appropriately.

[0154] If the difference in carbon dioxide concentration is large, the carbon dioxide generator control instruction is issued by increasing the release of carbon dioxide from the carbon dioxide generator.

[0155] Evaluate the degree of difference

[0156] For each environmental parameter, calculate the degree of deviation between its predicted value and the suitable range. You can use the absolute deviation or relative deviation calculation method. The absolute deviation is the absolute value of the difference between the predicted value and the boundary value of the suitable range. For example, if the temperature prediction value is 28℃ and the upper limit of the suitable temperature range is 25℃, then the absolute deviation of the temperature is |28-25|=3℃. The relative deviation is the ratio of the absolute deviation to the width of the suitable range. Taking temperature as an example, assuming that the suitable temperature range is [20,25]℃ and the width is 25-20=5℃, then the relative deviation is 3℃ / 5℃=0.6.

[0157] Prioritize environmental parameters based on the magnitude of the deviation. Parameters with larger deviations mean that the impact on plant growth may be more significant and need to be adjusted first. For example, if the deviation of temperature is larger than the deviation of humidity and light intensity, then temperature adjustment may be placed at a higher priority when generating greenhouse adjustment instructions.

[0158] Instruction transmission and execution

[0159] The generated greenhouse adjustment instruction set is transmitted to the execution module of the greenhouse control system through a suitable communication method. The communication method can be wired communication (such as serial communication, Ethernet communication, etc.) or wireless communication (such as Wi-Fi, Bluetooth, ZigBee, etc.), and the specific choice depends on the actual layout and equipment configuration of the greenhouse.

[0160] After receiving the instruction, the execution module controls the corresponding greenhouse equipment according to the instruction content. For example, the speed of the ventilator, the solenoid valve switch of the irrigation system, the power switch of the heating device, etc. are controlled by the control circuit. During the execution process, the operation of the equipment is ensured to be accurate and the operating status of the equipment is monitored in real time, such as whether it is started normally and whether the operating parameters meet the instruction requirements.

[0161] Environmental monitoring and data feedback

[0162] During the operation of the greenhouse equipment, the sensor continuously monitors the greenhouse environment in real time and collects environmental data, including temperature, humidity, light intensity, carbon dioxide concentration, etc. The sensor transmits the collected data back to the data processing module of the greenhouse control system in real time at a certain time interval (such as every minute, every five minutes, etc.).

[0163] The data processing module analyzes and processes the environmental data fed back, compares it with the previous greenhouse environmental forecast status, and evaluates the execution effect of the greenhouse adjustment instructions. For example, it checks whether the temperature drops after the ventilation and cooling instructions are executed, and whether the humidity rises after irrigation.

[0164] Feedback Regulation and Optimization

[0165] According to the feedback results of environmental data, if it is found that there is still a large deviation between the actual environmental parameters and the expected adjustment target, or the change trend of the environmental parameters does not meet expectations, the system will conduct feedback adjustment. Feedback adjustment can include adjusting the execution parameters of the instruction (such as increasing the running time of the ventilator, increasing the amount of irrigation water, etc.), or regenerating new adjustment instructions.

[0166] For example, if the temperature is still too high after executing a cooling command, the system can further increase the speed of the fan or extend the ventilation time. At the same time, the system will continuously optimize the control strategy based on the results of feedback regulation and accumulate experience data so that it can more accurately formulate greenhouse adjustment instructions in future environmental regulation to improve control effects and energy efficiency.

[0167] The continuous cycle of real-time control and feedback adjustment ensures that the greenhouse environment always remains in a state suitable for plant growth, or is constantly approaching a suitable state. Through continuous optimization and adjustment, the greenhouse intelligent control system can better adapt to different climatic conditions, plant growth stages and external interference factors, and achieve accurate and efficient greenhouse environment control.

[0168] Its beneficial effects are that, by obtaining greenhouse environment data and plant growth stage data from sensors, data preprocessing is performed on greenhouse environment data and plant growth stage data to obtain plant growth environment data; Apriori association analysis algorithm is used to perform association analysis on plant growth environment data to obtain initial plant growth environment data; a CNN-LSTM greenhouse environment prediction model is established based on CNN convolutional neural network and LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model; the initial CNN-LSTM greenhouse environment prediction model is optimized using IPSO improved particle swarm algorithm to obtain an IPSO-CNN-LSTM greenhouse environment prediction model; real-time plant growth environment data in the greenhouse is obtained, and the real-time plant growth environment data is input into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis to obtain a greenhouse environment prediction state; a greenhouse adjustment instruction is generated according to the greenhouse environment prediction state, and the sensor is controlled in real time according to the greenhouse adjustment instruction. 1. Accurate prediction of plant growth and early warning of abnormalities are realized, and growers can formulate scientific and reasonable cultivation management measures according to personalized models to significantly improve crop yield and quality. 2. The intelligent optimization model for energy consumption based on artificial intelligence algorithms can adjust equipment operation in real time according to the actual situation of the greenhouse, minimize energy consumption, improve energy utilization efficiency, and reduce greenhouse operating costs. 3. Precise water resource management realizes precise irrigation decisions through the analysis of multi-source data by deep learning models, avoids waste of water resources, and reduces problems such as soil nutrient loss and disease spread caused by excessive irrigation, which is beneficial to protecting the environment and healthy plant growth. 4. The greenhouse intelligent control system of the present invention improves the overall automation and intelligence level of greenhouse production, reduces manual labor intensity, improves production efficiency and economic benefits, and provides strong technical support for the development of modern agriculture.

[0169] In this embodiment, please refer to Figure 2 In a second embodiment of a greenhouse environment control method based on artificial intelligence in an embodiment of the present invention, obtaining greenhouse environment data and plant growth stage data from a sensor, performing data preprocessing on the greenhouse environment data and the plant growth stage data, and obtaining plant growth environment data includes the following steps:

[0170] Step 201: Acquire greenhouse environmental data in real time through various sensors arranged in the greenhouse, where the greenhouse environmental data at least includes temperature data, humidity data, light intensity data and carbon dioxide concentration data;

[0171] Step 202, recording the growth stage data of the plants in the greenhouse, the growth stage data at least including the germination stage, the seedling stage, the growth stage, the flowering stage and the fruiting stage;

[0172] Step 203: Clean the greenhouse environment data and growth stage data to remove abnormal values ​​and noise values ​​in the data;

[0173] Step 204: Map the data of different dimensions to the interval [0, 1] using a data normalization method to obtain plant growth environment data.

[0174] In this embodiment, please refer to Figure 3 In a third embodiment of a greenhouse environment control method based on artificial intelligence in an embodiment of the present invention, the target communication network traffic data is monitored, and the data change range of the target communication network traffic data is obtained in real time. If the data change range is greater than a set threshold, a second network traffic management strategy is generated, including the following steps:

[0175] Step 301, using the IPSO improved particle swarm algorithm to optimize the initial CNN-LSTM greenhouse environment prediction model, and determining the particle population size, number of iterations, learning factor, position and speed of the IPSO improved particle swarm algorithm;

[0176] Step 302, initializing the position and velocity of particles and randomly generating a population;

[0177] Population x i =(θ,β,a1,a2), where θ is the number of iterations of the CNN-LSTM greenhouse environment prediction model, β is the batch size, and a1 and a2 are the number of neurons in the first hidden layer and the second hidden layer respectively;

[0178] Step 303: determine the fitness function of the IPSO improved particle swarm algorithm, and use the initialized particle swarm parameters to build a CNN-LSTM greenhouse environment prediction model. The fitness function δ is defined as follows:

[0179]

[0180] Among them, N and M are the number of samples for training and testing, y i t and i v are training data and test data, and is the predicted value of the training sample and the test sample;

[0181] Step 304: Update the speed and position of the particle and determine the local optimal value P according to the fitness of the particle. t i and the global optimal value G t ;

[0182] Step 305: The optimal number of iterations, batch size, and number of neurons in the hidden layer found by the IPSO improved particle swarm algorithm are brought into the CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model.

[0183] The above describes a greenhouse environment control method based on artificial intelligence provided by an embodiment of the present invention. The following describes a greenhouse environment control system based on artificial intelligence according to an embodiment of the present invention. Figure 4 , an embodiment of the communication network traffic optimization system in the embodiment of the present invention includes:

[0184] A data acquisition module is used to acquire greenhouse environment data and plant growth stage data from the sensor, and perform data preprocessing on the greenhouse environment data and plant growth stage data to obtain plant growth environment data;

[0185] The association analysis module is used to perform association analysis on the plant growth environment data using the Apriori association analysis algorithm to obtain initial plant growth environment data;

[0186] The model building module is used to build a CNN-LSTM greenhouse environment prediction model based on the CNN convolutional neural network and the LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model;

[0187] The model optimization module is used to optimize the initial CNN-LSTM greenhouse environment prediction model using the IPSO improved particle swarm algorithm to obtain the IPSO-CNN-LSTM greenhouse environment prediction model;

[0188] The environmental prediction module is used to obtain the real-time plant growth environment data in the greenhouse, input the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtain the greenhouse environment prediction status;

[0189] The environment control module is used to generate greenhouse adjustment instructions according to the predicted state of the greenhouse environment and to control the sensor in real time according to the greenhouse adjustment instructions.

[0190] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A greenhouse environment control method based on artificial intelligence, characterized in that: The greenhouse environment control method comprises the following steps: Acquire greenhouse environment data and plant growth stage data from the sensor, perform data preprocessing on the greenhouse environment data and plant growth stage data, and obtain plant growth environment data; Performing association analysis on the plant growth environment data using the Apriori association analysis algorithm to obtain initial plant growth environment data; A CNN-LSTM greenhouse environment prediction model is established based on a CNN convolutional neural network and an LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model; The IPSO improved particle swarm algorithm is used to optimize the initial CNN-LSTM greenhouse environment prediction model to obtain an IPSO-CNN-LSTM greenhouse environment prediction model; Acquire real-time plant growth environment data in the greenhouse, input the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtain a greenhouse environment prediction state; A greenhouse adjustment instruction is generated according to the greenhouse environment prediction state, and the sensor is controlled in real time according to the greenhouse adjustment instruction.

2. The greenhouse environment control method based on artificial intelligence according to claim 1, characterized in that: The step of acquiring greenhouse environment data and plant growth stage data from a sensor, and performing data preprocessing on the greenhouse environment data and plant growth stage data to obtain plant growth environment data includes: By using various sensors arranged in the greenhouse, greenhouse environmental data is obtained in real time, wherein the greenhouse environmental data at least includes temperature data, humidity data, light intensity data and carbon dioxide concentration data; Recording growth stage data of plants in the greenhouse, wherein the growth stage data at least includes germination stage, seedling stage, growth stage, flowering stage and fruiting stage; Cleaning the greenhouse environment data and the growth stage data to remove abnormal values ​​and noise values ​​in the data; The data normalization method is used to map data of different dimensions to the interval [0,1] to obtain plant growth environment data.

3. The greenhouse environment control method based on artificial intelligence according to claim 1, characterized in that: The method of performing association analysis on the plant growth environment data using the Apriori association analysis algorithm to obtain initial plant growth environment data includes: The minimum support threshold and the minimum confidence threshold are set, and the plant growth environment data are scanned using the Apriori association analysis algorithm, and the number of occurrences of each item set is counted to obtain a frequent 1-item set; Assume that the plant growth environment data contains N transactions. For an item set X, its support calculation formula is as follows: in, Represents the number of transactions that contain item set X in the data set; For the association rule X→Y, the confidence calculation formula is: Generate candidate frequent 2-item sets by concatenation and pruning, scan the data set again, calculate the support of candidate frequent 2-item sets, and screen out frequent 2-item sets; Continue to generate higher-order frequent item sets until no item sets that meet the minimum support threshold are found; The association rules that meet the minimum confidence threshold are extracted from the frequent item sets to obtain the initial plant growth environment data.

4. The greenhouse environment control method based on artificial intelligence according to claim 1, characterized in that: The CNN-LSTM greenhouse environment prediction model is established based on the CNN convolutional neural network and the LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model, including: The CNN-LSTM greenhouse environment prediction model is established based on the CNN convolutional neural network and the LSTM long short-term memory network; The calculation formula of the forget gate of the LSTM network in the CNN-LSTM greenhouse environment prediction model is as follows: f t =σ(W f ·[H t-1 ,X t ]+b f ) Among them, f t is the forget gate, σ is the sigmoid activation function, X t is the input at time t, H t-1 is the short-term memory output at time t-1, W f is the forget gate weight, b f Bias for the forget gate; The input gate is based on the X of the network output node t and H t-1 To decide what new content needs to be added to memory, the calculation formula is as follows: i t =σ(W i ·[H t-1 ,X t ]+b i ) Among them, i t is the input gate, is the candidate vector, tanh is the hyperbolic tangent activation function, C t represents the long-term memory unit at time t, W i and W c are the weight parameters of the input gate and candidate vector, respectively, i and b c is the corresponding bias term; The output gate is based on the long short-term memory unit C t , input node X t and H t-1 The output value is Ht, and its calculation formula is as follows: ot=σ(W o ·[H t-1 ,X t ]+b o ) Ht=ot×tanh(Ct) Where ot is the output gate, W o is the output gate weight, b o is the output gate bias term, Ht is all the outputs of the model at time t; The initial plant growth environment data is divided into a training set and a test set according to a ratio of 9:1, and the established CNN-LSTM greenhouse environment prediction model is trained and tested to obtain an initial CNN-LSTM greenhouse environment prediction model.

5. The greenhouse environment control method based on artificial intelligence according to claim 1, characterized in that: The IPSO improved particle swarm algorithm is used to optimize the initial CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model, including: The initial CNN-LSTM greenhouse environment prediction model is optimized by using the IPSO improved particle swarm algorithm, and the particle population size, number of iterations, learning factor, position and speed of the IPSO improved particle swarm algorithm are determined; Initialize the position and velocity of the particles and randomly generate a population x i =(θ,β,a1,a2), where θ is the number of iterations of the CNN-LSTM greenhouse environment prediction model, β is the batch size, and a1 and a2 are the number of neurons in the first hidden layer and the second hidden layer respectively; The fitness function of the IPSO improved particle swarm algorithm is determined, and the CNN-LSTM greenhouse environment prediction model is constructed using the initialized particle swarm parameters. The fitness function δ is defined as follows: Among them, N and M are the number of samples for training and testing, y i t and i v are training data and test data, and is the predicted value of the training sample and the test sample; Update the speed and position of the particle, and determine the local optimal value P according to the fitness of the particle t i and the global optimal value G t ; The optimal number of iterations, batch size, and number of neurons in the hidden layer found by the IPSO improved particle swarm algorithm are brought into the CNN-LSTM greenhouse environment prediction model to obtain the IPSO-CNN-LSTM greenhouse environment prediction model.

6. The greenhouse environment control method based on artificial intelligence as claimed in claim 5, characterized in that: The IPSO improved particle swarm algorithm includes: The learning factor c1 in the original PSO particle swarm algorithm is set from large to small, and the learning factor c2 is set from small to large. The formula is as follows: The inertia coefficient ω in the PSO particle swarm algorithm is set as a nonlinear inertia weight, and the improved form of the inertia coefficient ω is as follows: Among them, ω max is the maximum value of the inertia coefficient, ω min is the minimum value of the inertia coefficient, t is the current iteration number, t max is the maximum number of iterations; Set the adaptive mutation mechanism in the PSO particle swarm algorithm. After each particle update, set the threshold ρ. When the particle meets the mutation condition, mutation occurs. The calculation formula of the adaptive mutation mechanism is as follows: rand>ρ Among them, ρ is the threshold of the adaptive mutation mechanism, and rand is a random number between [0,1]. After the learning factor c1, learning factor c2, inertia coefficient ω and adaptive mutation mechanism are improved, the IPSO improved particle swarm algorithm is obtained.

7. The greenhouse environment control method based on artificial intelligence according to claim 1, characterized in that: The step of generating a greenhouse adjustment instruction according to the greenhouse environment prediction state, and controlling the sensor in real time according to the greenhouse adjustment instruction, comprises: Generate a greenhouse adjustment instruction according to the greenhouse environment prediction state, wherein the greenhouse environment prediction state includes at least a temperature prediction value, a humidity prediction value, a light intensity prediction value, and a carbon dioxide concentration prediction value; The absolute deviation is used to calculate the difference between the predicted value and the boundary value of the suitable range, and the environmental parameters are prioritized according to the size of the difference; If the difference in the temperature prediction value is large, a ventilation equipment opening control instruction is generated, and the ventilation equipment is used for air convection to reduce the temperature to a temperature range of 22.5°C-25°C; If the light difference is large, a sunshade control instruction is generated to use the sunshade to block the light. If the light is insufficient, the fill light is turned on to supplement the light. If the humidity difference is large, turn on the ventilation equipment for air convection. If the humidity is too low, start the irrigation system to spray water appropriately. If the difference in carbon dioxide concentration is large, the carbon dioxide generator control instruction is issued by increasing the release of carbon dioxide from the carbon dioxide generator.

8. A greenhouse environment control system based on artificial intelligence, characterized in that: The greenhouse environment control system comprises the following steps: A data acquisition module, used to acquire greenhouse environment data and plant growth stage data from the sensor, and perform data preprocessing on the greenhouse environment data and plant growth stage data to obtain plant growth environment data; An association analysis module is used to perform association analysis on the plant growth environment data using an Apriori association analysis algorithm to obtain initial plant growth environment data; A model building module is used to establish a CNN-LSTM greenhouse environment prediction model based on a CNN convolutional neural network and an LSTM long short-term memory network; the CNN-LSTM greenhouse environment prediction model is trained and tested using the initial plant growth environment data to obtain an initial CNN-LSTM greenhouse environment prediction model; A model optimization module is used to optimize the initial CNN-LSTM greenhouse environment prediction model by using an IPSO improved particle swarm algorithm to obtain an IPSO-CNN-LSTM greenhouse environment prediction model; An environmental prediction module is used to obtain real-time plant growth environment data in the greenhouse, input the real-time plant growth environment data into the IPSO-CNN-LSTM greenhouse environment prediction model for analysis, and obtain a greenhouse environment prediction state; The environment control module is used to generate a greenhouse adjustment instruction according to the predicted state of the greenhouse environment, and to control the sensor in real time according to the greenhouse adjustment instruction.

9. The greenhouse environment control system based on artificial intelligence as claimed in claim 8, characterized in that: The data acquisition module includes the following submodules: An acquisition submodule is used to acquire greenhouse environmental data in real time through various sensors arranged in the greenhouse, wherein the greenhouse environmental data at least includes temperature data, humidity data, light intensity data and carbon dioxide concentration data; A recording submodule, used to record the growth stage data of plants in the greenhouse, wherein the growth stage data at least includes the germination stage, the seedling stage, the growth stage, the flowering stage and the fruiting stage; A cleaning submodule, used to clean the greenhouse environment data and the growth stage data to remove abnormal values ​​and noise values ​​in the data; The mapping submodule is used to map data of different dimensions to the [0,1] interval using a data normalization method to obtain plant growth environment data.

10. The greenhouse environment control system based on artificial intelligence according to claim 8, characterized in that: The association analysis module includes the following submodules: A setting submodule is used to set a minimum support threshold and a minimum confidence threshold, use the Apriori association analysis algorithm to scan the plant growth environment data, count the number of occurrences of each item set, and obtain a frequent 1-item set; The calculation submodule is used to assume that the plant growth environment data contains N transactions. For an item set X, the support calculation formula is as follows: in, Represents the number of transactions that contain item set X in the data set; For the association rule X→Y, the confidence calculation formula is: The generation submodule is used to generate candidate frequent 2-item sets by concatenation and pruning, and scan the data set again to calculate the support of the candidate frequent 2-item sets and filter out the frequent 2-item sets; Iteration submodule, used to continuously generate higher-order frequent itemsets until no itemsets that meet the minimum support threshold are found; The extraction submodule is used to extract association rules that meet the minimum confidence threshold from the frequent item set to obtain initial plant growth environment data.

Images