Environment assessment monitoring method and device based on big data analysis

Abstract

The invention discloses an environment evaluation monitoring method and device based on big data analysis, and relates to the technical field of big data analysis. Acquiring environment monitoring data of the target greenhouse in a preset time period through a plurality of sensors; acquiring an operation state of regulation and control equipment of the target greenhouse in a preset time period; obtaining a feature matrix according to the environment monitoring data and the operation state of the regulation and control equipment; and taking the feature matrix as the input of a pre-trained long-short-term memory network model to obtain a prediction result. By cleaning and fusing the acquired environment monitoring data, unnecessary information can be removed, and high-quality and reliable data can be obtained; the long short-term memory network can capture the long-term dependency relationship in the time sequence data, and the prediction result of the environment can be obtained by inputting the feature matrix into the pre-trained long short-term memory network; based on the prediction result, the operation parameters of the regulation and control equipment can be adjusted in advance to stabilize the environmental parameters in the greenhouse.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to an environmental assessment and monitoring method and device based on big data analysis. Background Art

[0002] Greenhouse cultivation is a modern agricultural technology that provides a suitable growth environment for crops by building a closed or semi-closed environment and artificially controlling temperature, humidity, light, ventilation and other conditions. This method of cultivation is not restricted by seasons and climate and can stably produce high-quality and high-yield crops throughout the year. Greenhouse cultivation not only improves agricultural production efficiency, but also reduces the impact of natural disasters on crops. It is an important direction for the development of modern agriculture towards intelligence and refinement.

[0003] Precise control of greenhouse planting environment conditions plays a vital role in increasing crop yields and improving crop quality. Patent CN112416036B discloses a greenhouse environment monitoring system that controls the start and stop of execution node terminals to adjust the greenhouse environment by judging whether real-time environmental data exceeds a preset environmental parameter control threshold.

[0004] However, traditional environmental monitoring methods often rely on the instantaneous readings of a single sensor, which makes it difficult to fully and accurately reflect the dynamic changes in the greenhouse environment, and cannot effectively predict future environmental trends, resulting in a lag in environmental regulation. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of hysteresis in environmental regulation mentioned in the above background technology, and to propose an environmental assessment and monitoring method and device based on big data analysis.

[0006] The first aspect of the present invention provides an environmental assessment and monitoring method based on big data analysis, the method comprising:

[0007] Obtaining environmental monitoring data of the target greenhouse within a preset period of time; the environmental monitoring data includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneously monitoring multiple sensors of the same type;

[0008] Cleaning and fusing the environmental monitoring data to obtain a first feature matrix;

[0009] Acquire the operating status of the control equipment of the target greenhouse within the preset time period, and obtain a second characteristic matrix according to the operating status;

[0010] Merging the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix;

[0011] The third characteristic matrix is ​​used as the input of the pre-trained long short-term memory network model to obtain the prediction result of the target greenhouse at the target time point, and the operating parameters of the control equipment are adjusted according to the prediction result; the prediction result includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

[0012] Optionally, the cleaning and fusing of the environmental monitoring data to obtain a first feature matrix includes:

[0013] According to the monitoring data of each sensor, determine whether each sensor is normal;

[0014] For normal sensors, high confidence data at the target time is calculated based on monitoring data of the same type of sensor at the target time; the preset time period includes multiple sampling times; the target time is any one of the multiple sampling times;

[0015] According to the high confidence data of the same type at different times, the first eigenvector is obtained;

[0016] According to the first eigenvectors corresponding to the different types of sensors, a fourth eigenmatrix is ​​obtained;

[0017] The fourth characteristic matrix is ​​normalized to obtain a first characteristic matrix.

[0018] Optionally, judging whether each sensor is normal according to the monitoring data of each sensor includes:

[0019] Determining a first change threshold value according to a type of the target sensor;

[0020] Counting the monitoring data of the target sensor to obtain a first data change curve;

[0021] Set the sliding window according to the preset window size, and calculate the absolute value of the change between the starting point and the end point of the sliding window;

[0022] If the absolute value of the change in any sliding window is greater than or equal to the first change threshold, it is determined that the target sensor is failed; otherwise, it is determined that the target sensor is normal.

[0023] Optionally, the calculating high confidence data at the target time according to the monitoring data of the same type of sensor at the target time includes:

[0024] According to the monitoring data of the same type of sensors at the target time, the difference matrix is ​​calculated:

[0025] A ij =|M i -M j |

[0026] Wherein, Aij is the difference between the monitoring values ​​of the i-th sensor and the j-th sensor at the target time, and is the element of the i-th row and j-th column of the difference matrix; Mi is the monitoring value of the i-th sensor at the target time; Mj is the monitoring value of the j-th sensor at the target time;

[0027] According to the difference matrix, the first difference mean of each sensor is calculated:

[0028]

[0029] in, is the first difference mean of the i-th sensor; n is the number of sensors of the same type;

[0030] According to the first difference mean of each sensor, high confidence data at the target time is obtained.

[0031] Optionally, obtaining high confidence data at the target time according to the first difference mean of each sensor includes:

[0032] According to the first difference mean of each sensor, the second difference mean is calculated:

[0033]

[0034] in, is the desired second difference mean;

[0035] Putting the monitoring values ​​of the sensors whose first difference mean is less than the second difference mean into the first valid data set;

[0036] The elements in the first valid data set are averaged to obtain high confidence data at the target time:

[0037]

[0038] Among them, C is the high confidence data at the target moment; r is the number of elements in the first valid data set; Mp is the pth element value in the first valid data set.

[0039] Optionally, the long short-term memory network model includes an input layer, a first LSTM layer, a first normalization layer, a second LSTM layer, a second normalization layer, a multi-head attention mechanism layer, a fully connected layer and multiple output layers;

[0040] The step of using the third feature matrix as an input of a pre-trained long short-term memory network model to obtain an environment prediction result includes:

[0041] Using the third feature matrix as input of the input layer to obtain first representation data;

[0042] Using the first representation data as input of the first LSTM layer to obtain a first feature;

[0043] Using the first feature as input of the first normalization layer to obtain a second feature;

[0044] Using the second feature as the input of the second LSTM layer to obtain a third feature;

[0045] Using the third feature as the input of the second normalization layer to obtain a fourth feature;

[0046] Using the fourth feature as the input of the multi-head attention mechanism layer to obtain a fifth feature;

[0047] Using the fifth feature as an input of the fully connected layer to obtain a sixth feature;

[0048] The sixth feature is used as the input of each output layer to obtain a prediction result.

[0049] Optionally, during the model training phase, a sparrow search algorithm is used to optimize the hyperparameters of the long short-term memory network model; the hyperparameters include the number of neurons, the number of iterations, the input batch, and the learning rate;

[0050] The method further comprises:

[0051] Step 1: Determine the search range of the hyperparameter according to a preset range;

[0052] Step 2, initializing the parameters of the sparrow population and classifying the population; the parameters of the sparrow population include population size, initial position, optimal position and global optimal fitness; the classification types include discoverers, followers and vigilants;

[0053] Step 3: Calculate the fitness value of each individual sparrow based on the initial position;

[0054] Step 4: Update the locations of the discoverer, follower, and sentinel;

[0055] Step 5, constraining the positions of the discoverer, the follower and the alerter according to the search range;

[0056] Step 6: Calculate the fitness of each individual sparrow at the updated position to obtain the optimal fitness of this iteration;

[0057] Step 7: If the best fitness of this round of iteration is greater than the global best fitness, the best fitness of this round of iteration is used as the global best fitness, and the position of the sparrow corresponding to the best fitness of this round of iteration is used as the best position;

[0058] Step eight, determining whether the current round of iteration has reached the stopping condition, if not, returning to step four; otherwise, stopping the iteration, and determining the hyperparameters of the long short-term memory network model according to the optimal position.

[0059] A second aspect of the present invention provides an environmental assessment and monitoring device based on big data analysis, the device comprising:

[0060] The first data acquisition module is used to acquire environmental monitoring data of the target greenhouse within a preset period of time; the environmental monitoring data includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneously monitoring multiple sensors of the same type;

[0061] A data preprocessing module, used for cleaning and fusing the environmental monitoring data to obtain a first feature matrix;

[0062] A second data acquisition module, used to acquire the operating state of the control equipment of the target greenhouse within the preset time period, and obtain a second characteristic matrix according to the operating state;

[0063] A data merging module, used for merging the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix;

[0064] A prediction module is used to use the third feature matrix as an input of a pre-trained long short-term memory network model to obtain a prediction result of the target greenhouse at a target time point, and adjust the operating parameters of the control device according to the prediction result; the prediction result includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

[0065] Beneficial effects of the present invention:

[0066] The present invention proposes an environmental assessment and monitoring method based on big data analysis, which includes: obtaining environmental monitoring data of a target greenhouse within a preset time period; the environmental monitoring data include air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by monitoring multiple sensors of the same type at the same time; cleaning and fusing the environmental monitoring data to obtain a first feature matrix; obtaining the operating status of the control equipment of the target greenhouse within a preset time period, and obtaining a second feature matrix according to the operating status; merging the first feature matrix and the second feature matrix to obtain a third feature matrix; using the third feature matrix as the input of a pre-trained long short-term memory network model to obtain a prediction result of the target greenhouse at a target time point, and adjusting the operating parameters of the control equipment according to the prediction result; the prediction result includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

[0067] By cleaning and fusing the acquired environmental monitoring data, unnecessary information can be removed to obtain high-quality and reliable data, thereby improving the reliability of the prediction results; the long short-term memory network can capture the long-term dependencies in time series data, and by inputting the feature matrix into the pre-trained long short-term memory network, the environmental prediction results can be obtained; based on the prediction results, the operating parameters of the control equipment can be adjusted in advance to stabilize the environmental parameters in the greenhouse, thereby avoiding the problems of delayed and untimely adjustment in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The present invention will be further described below in conjunction with the accompanying drawings.

[0069] Figure 1 A flowchart of an environmental assessment and monitoring method based on big data analysis is provided for an embodiment of the present invention;

[0070] Figure 2 A structural diagram of an environmental assessment and monitoring device based on big data analysis is provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] The embodiment of the present invention provides an environmental assessment and monitoring method based on big data analysis. Figure 1 , Figure 1 A flowchart of an environmental assessment and monitoring method based on big data analysis provided by an embodiment of the present invention. The method comprises the following steps:

[0073] S101, obtaining environmental monitoring data of a target greenhouse within a preset period of time.

[0074] S102, cleaning and integrating the environmental monitoring data to obtain a first feature matrix.

[0075] S103, obtaining the operating status of the control equipment of the target greenhouse within a preset period of time, and obtaining a second characteristic matrix according to the operating status.

[0076] S104, merging the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix.

[0077] S105, using the third feature matrix as the input of the pre-trained long short-term memory network model to obtain the prediction result of the target greenhouse at the target time point, and adjusting the operating parameters of the control equipment according to the prediction result.

[0078] Among them, environmental monitoring data include air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneous monitoring by multiple sensors of the same type; the prediction results include air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

[0079] An environmental assessment and monitoring method based on big data analysis provided by an embodiment of the present invention can remove unnecessary information by cleaning and fusing the acquired environmental monitoring data, and obtain high-quality and reliable data, thereby improving the reliability of the prediction results; the long short-term memory network can capture the long-term dependencies in the time series data, and the environmental prediction results can be obtained by inputting the feature matrix into the pre-trained long short-term memory network; based on the prediction results, the operating parameters of the control equipment can be adjusted in advance to stabilize the environmental parameters in the greenhouse, thereby avoiding the problems of delayed and untimely adjustment in traditional methods.

[0080] In one implementation, environmental monitoring data may be sampled every 15 minutes. The first characteristic matrix may have six columns, representing air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration, and light intensity, respectively; each row represents a monitoring value at a sampling moment.

[0081] The second feature matrix may have 15 columns, representing the state of the heating device, heating time, heating temperature, the state of the cooling device, cooling time, cooling temperature, the state of the ventilation device, ventilation time, ventilation rate, the state of the irrigation device, irrigation time, irrigation rate, the lighting device, lighting time and lighting intensity. Each row represents the operating state of a 15-minute period.

[0082] The first feature matrix and the second feature matrix are merged row by row to obtain a third feature matrix for predicting the environmental state at a target time point; the target time point may be the fifth minute after the current moment, and the preset time period may be four hours before the current moment. At this time, the first feature matrix has 16 rows and 6 columns, the second feature matrix has 16 rows and 15 columns, and the third feature matrix has 16 rows and 21 columns.

[0083] In one embodiment, step S102 includes:

[0084] Step 1: Determine whether each sensor is normal based on the monitoring data of each sensor.

[0085] Step 2: For normal sensors, high confidence data at the target time is calculated based on the monitoring data of the same type of sensors at the target time.

[0086] Step three, obtain the first eigenvector based on the same type of high confidence data at different times.

[0087] Step 4: Obtain a fourth characteristic matrix according to the first characteristic vectors corresponding to different types of sensors.

[0088] Step five, normalize the fourth characteristic matrix to obtain the first characteristic matrix.

[0089] The preset time period includes multiple sampling moments; the target moment is any moment among the multiple sampling moments.

[0090] In one implementation, by determining whether the sensor is normal, the data of abnormal or faulty sensors can be eliminated to ensure that the data used is more accurate and reliable. Calculating high-confidence data for the data of normal sensors can reduce occasional errors in the data of a single sensor and improve the overall accuracy of the data.

[0091] In one implementation, after the fourth feature matrix is ​​normalized, the numerical range of the first feature matrix is ​​unified, which is helpful for the prediction of the subsequent model.

[0092] In one embodiment, judging whether each sensor is normal according to the monitoring data of each sensor includes:

[0093] Step 1: Determine a first change threshold according to the type of the target sensor.

[0094] Step 2: Count the monitoring data of the target sensor to obtain a first data change curve.

[0095] Step 3: Set the sliding window according to the preset window size and calculate the absolute value of the change between the starting point and the end point of the sliding window.

[0096] Step 4: If the absolute value of the change in any sliding window is greater than or equal to the first change threshold, the target sensor is judged to be invalid; otherwise, the target sensor is judged to be normal.

[0097] In one implementation, the change of each parameter of the environment should not exceed a certain value within a certain time step. For example, the preset window size can be set to 5 seconds. The change of air temperature after a 5-second interval should not exceed 0.5 degrees Celsius; the change of air humidity after a 5-second interval should not exceed 0.5%; the change of CO2 concentration after a 5-second interval should not exceed 100mg / m3, etc. Determining the first change threshold of different types of sensors in this way can help determine which sensors have failed and can be repaired or replaced in time; it can also help determine which sensors are normal, so that reliable data can be used for prediction, making the prediction results more accurate.

[0098] In one embodiment, based on the monitoring data of the same type of sensor at the target time, the high confidence data at the target time is calculated including:

[0099] Step 1: Based on the monitoring data of the same type of sensors at the target time, the difference matrix is ​​calculated:

[0100] A ij =|M i -M j |

[0101] Among them, Aij is the difference between the monitoring values ​​of the i-th sensor and the j-th sensor at the target time, and is the element of the i-th row and j-th column of the difference matrix; Mi is the monitoring value of the i-th sensor at the target time; Mj is the monitoring value of the j-th sensor at the target time.

[0102] Step 2: Calculate the first difference mean of each sensor according to the difference matrix:

[0103]

[0104] in, is the first difference mean of the i-th sensor; n is the number of sensors of the same type.

[0105] Step three: obtain high confidence data at the target time according to the first difference mean of each sensor.

[0106] In one implementation, the difference matrix can be used to calculate the distance between a sensor's monitoring value and the overall monitoring value, i.e., the first difference mean. This helps to select the most supported data of the same type of sensor. For example, the monitoring value of the sensor with the smallest first difference mean can be used as the high confidence data at the target time.

[0107] In one embodiment, obtaining high confidence data at the target time according to the first difference mean of each sensor includes:

[0108] Step 1: Calculate the second difference mean according to the first difference mean of each sensor:

[0109]

[0110] in, is the desired second difference mean.

[0111] Step 2: Put the monitoring values ​​of the sensors whose first difference mean is smaller than the second difference mean into the first valid data set.

[0112] Step 3: Average the elements in the first valid data set to obtain high confidence data at the target time:

[0113]

[0114] Wherein, C is the high confidence data at the target moment; r is the number of elements in the first valid data set; Mp is the value of the pth element in the first valid data set.

[0115] In one implementation, the data in the first valid data set is the result of two difference mean screenings, and these data have high guarantees in terms of consistency and reliability. This multiple screening process can effectively reduce the impact of outliers and noise data, ensuring that the final data has high consistency. Based on the average value of multiple reliable sensor data, the occasional errors of individual sensors can be smoothed out, improving the credibility of the data.

[0116] In one embodiment, the long short-term memory network model includes an input layer, a first LSTM layer, a first normalization layer, a second LSTM layer, a second normalization layer, a multi-head attention mechanism layer, a fully connected layer, and multiple output layers;

[0117] The third feature matrix is ​​used as the input of the pre-trained long short-term memory network model, and the environmental prediction results include:

[0118] Step 1: Use the third feature matrix as the input of the input layer to obtain the first representation data.

[0119] Step 2: Use the first representation data as the input of the first LSTM layer to obtain the first feature.

[0120] Step three, use the first feature as the input of the first normalization layer to obtain the second feature.

[0121] Step 4: Use the second feature as the input of the second LSTM layer to obtain the third feature.

[0122] Step five, use the third feature as the input of the second normalization layer to obtain the fourth feature.

[0123] Step 6: Use the fourth feature as the input of the multi-head attention mechanism layer to obtain the fifth feature.

[0124] Step seven, use the fifth feature as the input of the fully connected layer to obtain the sixth feature.

[0125] Step 8: Use the sixth feature as the input of each output layer to obtain the prediction result.

[0126] In one implementation, multi-layer LSTM can capture complex patterns and long- and short-term dependencies in time series data, improving the depth and breadth of feature extraction. The multi-head attention mechanism can focus on important features in the input data, further enhancing the model's ability to capture key information.

[0127] In one implementation, adding a normalization layer after the LSTM layer can effectively prevent the gradient from vanishing or exploding, stabilize model training, and improve the model's convergence speed and prediction accuracy. Normalization can also unify the data range of different features and reduce the impact of magnitude differences between features on the model.

[0128] In one implementation, the prediction results for each type of environmental data are given by a separate output layer.

[0129] In one embodiment, during the model training phase, the sparrow search algorithm is used to optimize the hyperparameters of the long short-term memory network model; the hyperparameters include the number of neurons, the number of iterations, the input batch, and the learning rate;

[0130] The parameter optimization process includes:

[0131] Step 1: Determine the search range of hyperparameters based on the preset range.

[0132] Step 2: Initialize the sparrow population parameters and classify the population.

[0133] Step 3: Calculate the fitness value of each sparrow individual based on the initial position.

[0134] Step 4: Update the locations of the discoverer, follower, and sentinel.

[0135] Step 5: Constrain the locations of the discoverer, follower, and sentinel based on the search scope.

[0136] Step 6: Calculate the fitness of each sparrow individual at the updated position to obtain the optimal fitness of this iteration.

[0137] Step 7: If the best fitness of this round of iteration is greater than the global best fitness, the best fitness of this round of iteration is used as the global best fitness, and the sparrow position corresponding to the best fitness of this round of iteration is used as the best position.

[0138] Step 8: Determine whether the current iteration has reached the stopping condition. If not, return to step 4; otherwise, stop the iteration and determine the hyperparameters of the long short-term memory network model based on the optimal position.

[0139] Among them, the sparrow population parameters include population size, initial position, optimal position and global optimal fitness; the classification types include discoverers, followers and guards.

[0140] In one implementation, the sparrow search algorithm can automatically search and optimize the hyperparameters of the LSTM model (such as the number of neurons, number of iterations, input batch, and learning rate) to find the best combination and improve the predictive performance of the model.

[0141] The embodiment of the present invention provides an environmental assessment and monitoring method based on big data analysis. Figure 2 , Figure 2 A structural diagram of an environmental assessment and monitoring device based on big data analysis provided by an embodiment of the present invention. The device includes:

[0142] The first data acquisition module is used to acquire the environmental monitoring data of the target greenhouse within a preset time period.

[0143] The data preprocessing module is used to clean and fuse the environmental monitoring data to obtain the first feature matrix.

[0144] The second data acquisition module is used to acquire the operating status of the control equipment of the target greenhouse within a preset time period, and obtain a second characteristic matrix according to the operating status.

[0145] The data merging module is used to merge the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix.

[0146] The prediction module is used to use the third feature matrix as the input of the pre-trained long short-term memory network model to obtain the prediction result of the target greenhouse at the target time point, and adjust the operating parameters of the control equipment according to the prediction result.

[0147] Among them, environmental monitoring data include air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneous monitoring by multiple sensors of the same type; the prediction results include air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

[0148] An environmental assessment and monitoring device based on big data analysis provided by an embodiment of the present invention can remove unnecessary information by cleaning and fusing the acquired environmental monitoring data, and obtain high-quality and reliable data, thereby improving the reliability of the prediction results; the long short-term memory network can capture the long-term dependencies in the time series data, and the environmental prediction results can be obtained by inputting the feature matrix into the pre-trained long short-term memory network; based on the prediction results, the operating parameters of the control equipment can be adjusted in advance to stabilize the environmental parameters in the greenhouse, thereby avoiding the problems of delayed and untimely adjustment in traditional methods.

[0149] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An environmental assessment and monitoring method based on big data analysis, characterized in that: The method comprises: Obtaining environmental monitoring data of the target greenhouse within a preset period of time; the environmental monitoring data includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneously monitoring multiple sensors of the same type; Cleaning and fusing the environmental monitoring data to obtain a first feature matrix; Acquire the operating status of the control equipment of the target greenhouse within the preset time period, and obtain a second characteristic matrix according to the operating status; Merging the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix; The third characteristic matrix is ​​used as the input of the pre-trained long short-term memory network model to obtain the prediction result of the target greenhouse at the target time point, and the operating parameters of the control equipment are adjusted according to the prediction result; the prediction result includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

2. The environmental assessment and monitoring method based on big data analysis according to claim 1 is characterized in that: The cleaning and fusion of the environmental monitoring data to obtain a first feature matrix includes: According to the monitoring data of each sensor, determine whether each sensor is normal; For normal sensors, high confidence data at the target time is calculated based on monitoring data of the same type of sensor at the target time; the preset time period includes multiple sampling times; the target time is any one of the multiple sampling times; According to the high confidence data of the same type at different times, the first eigenvector is obtained; According to the first eigenvectors corresponding to the different types of sensors, a fourth eigenmatrix is ​​obtained; The fourth characteristic matrix is ​​normalized to obtain a first characteristic matrix.

3. The environmental assessment and monitoring method based on big data analysis according to claim 2 is characterized in that: The step of judging whether each sensor is normal according to the monitoring data of each sensor includes: Determining a first change threshold value according to a type of the target sensor; Counting the monitoring data of the target sensor to obtain a first data change curve; Set the sliding window according to the preset window size, and calculate the absolute value of the change between the starting point and the end point of the sliding window; If the absolute value of the change in any sliding window is greater than or equal to the first change threshold, it is determined that the target sensor is failed; otherwise, it is determined that the target sensor is normal.

4. The environmental assessment and monitoring method based on big data analysis according to claim 2 is characterized in that: The high confidence data at the target time is calculated based on the monitoring data of the same type of sensor at the target time, including: According to the monitoring data of the same type of sensors at the target time, the difference matrix is ​​calculated: A ij =|M i -M j | Among them, A ij is the difference between the monitoring values ​​of the i-th sensor and the j-th sensor at the target time, and is the i-th row and j-th column element of the difference matrix; M i is the monitoring value of the i-th sensor at the target time; M j is the monitoring value of the jth sensor at the target time; According to the difference matrix, the first difference mean of each sensor is calculated: in, is the first difference mean of the i-th sensor; n is the number of sensors of the same type; According to the first difference mean of each sensor, high confidence data at the target time is obtained.

5. The environmental assessment and monitoring method based on big data analysis according to claim 4 is characterized in that: The step of obtaining high confidence data at the target time according to the first difference mean of each sensor includes: According to the first difference mean of each sensor, the second difference mean is calculated: in, is the desired second difference mean; Putting the monitoring values ​​of the sensors whose first difference mean is less than the second difference mean into the first valid data set; The elements in the first valid data set are averaged to obtain high confidence data at the target time: Where C is the high confidence data at the target moment; r is the number of elements in the first valid data set; M p is the value of the pth element in the first valid data set.

6. The environmental assessment and monitoring method based on big data analysis according to claim 1 is characterized in that: The long short-term memory network model includes an input layer, a first LSTM layer, a first normalization layer, a second LSTM layer, a second normalization layer, a multi-head attention mechanism layer, a fully connected layer and multiple output layers; The step of using the third feature matrix as an input of a pre-trained long short-term memory network model to obtain an environment prediction result includes: Using the third feature matrix as input of the input layer to obtain first representation data; Using the first representation data as input of the first LSTM layer to obtain a first feature; Using the first feature as input of the first normalization layer to obtain a second feature; Using the second feature as the input of the second LSTM layer to obtain a third feature; Using the third feature as the input of the second normalization layer to obtain a fourth feature; Using the fourth feature as the input of the multi-head attention mechanism layer to obtain a fifth feature; Using the fifth feature as an input of the fully connected layer to obtain a sixth feature; The sixth feature is used as the input of each output layer to obtain a prediction result.

7. The environmental assessment and monitoring method based on big data analysis according to claim 1 is characterized in that: In the model training phase, the hyperparameters of the long short-term memory network model are optimized using a sparrow search algorithm; the hyperparameters include the number of neurons, the number of iterations, the input batch, and the learning rate; The method further comprises: Step 1: Determine the search range of the hyperparameter according to a preset range; Step 2, initializing the parameters of the sparrow population and classifying the population; the parameters of the sparrow population include population size, initial position, optimal position and global optimal fitness; the classification types include discoverers, followers and vigilants; Step 3: Calculate the fitness value of each individual sparrow based on the initial position; Step 4: Update the locations of the discoverer, follower, and sentinel; Step 5, constraining the positions of the discoverer, the follower and the alerter according to the search range; Step 6: Calculate the fitness of each individual sparrow at the updated position to obtain the optimal fitness of this iteration; Step 7: If the best fitness of this round of iteration is greater than the global best fitness, the best fitness of this round of iteration is used as the global best fitness, and the position of the sparrow corresponding to the best fitness of this round of iteration is used as the best position; Step eight, determining whether the current round of iteration has reached the stopping condition, if not, returning to step four; otherwise, stopping the iteration, and determining the hyperparameters of the long short-term memory network model according to the optimal position.

8. An environmental assessment and monitoring device based on big data analysis, characterized in that: The device comprises: The first data acquisition module is used to acquire environmental monitoring data of the target greenhouse within a preset period of time; the environmental monitoring data includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity; each type of environmental monitoring data is obtained by simultaneously monitoring multiple sensors of the same type; A data preprocessing module, used for cleaning and fusing the environmental monitoring data to obtain a first feature matrix; A second data acquisition module, used to acquire the operating state of the control equipment of the target greenhouse within the preset time period, and obtain a second characteristic matrix according to the operating state; A data merging module, used for merging the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix; A prediction module is used to use the third feature matrix as an input of a pre-trained long short-term memory network model to obtain a prediction result of the target greenhouse at a target time point, and adjust the operating parameters of the control device according to the prediction result; the prediction result includes air temperature, air humidity, soil temperature, soil humidity, carbon dioxide concentration and light intensity.

Images