Water environment operation and maintenance monitoring method and system based on data analysis

Abstract

The invention relates to the technical field of data processing, in particular to a water environment operation and maintenance monitoring method and system based on data analysis, and the method comprises the steps: collecting previous processing data and subsequent processing data for a water environment; performing timestamp alignment on the previous processing data and the subsequent processing data; matching the previous processing data at the previous set acquisition moment with the later processing data at the later set acquisition moment to obtain a data set corresponding to time staggering; inputting a data set by adopting a long-short-term memory network time sequence model; training, evaluating and optimizing the time sequence model; and predicting the water environment through the optimized time sequence model. According to the invention, the water environment operation and maintenance monitoring and prediction effects can be significantly improved, the subjectivity, real-time performance and data integration problems at present are effectively solved through comprehensive data operation and maintenance, efficient and stable operation of a water treatment system is ensured, and sustainable utilization of water resources is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a water environment operation and maintenance monitoring method and system based on data analysis. Background Art

[0002] At present, in the specific working conditions of water environment monitoring and management, the equipment involved mainly includes:

[0003] The grid is used to achieve the front-end treatment of water, intercept and remove solid impurities and floating objects in the water, prevent these impurities from entering the back-end treatment system, and ensure the smooth and efficient water treatment process; the pump station is responsible for lifting and conveying water, providing power for the water flow, and ensuring the flow and circulation of water in each link of the back-end treatment system, which includes sedimentation tanks, filters, disinfection equipment, etc.; the remote control module realizes remote monitoring and control of the water treatment system, often connecting various devices through the Internet of Things technology to provide data transmission and control interfaces; the power distribution equipment provides a stable power supply for various electrical equipment in the water treatment system to ensure the normal operation of the system. Through the effective operation and maintenance and comprehensive management of the above equipment, the treatment and improvement effect of the water environment can be significantly improved, and the sustainable use of water resources can be guaranteed.

[0004] However, in the current implementation process, only the collection of independent data from the front and back channels has been achieved, and the judgment and prediction of the effect of water environment operation and maintenance through various data are often achieved through manual methods, such as visualizing various data so that the staff can intuitively but relatively subjectively judge the results of water environment operation and maintenance based on the displayed results. This approach often has the following problems:

[0005] Human judgment is highly subjective, and different people have different levels of experience and knowledge, which leads to differences in judgment results, resulting in inconsistency and instability in operation and maintenance decisions. Manual analysis and judgment take time, making it difficult to respond to and handle emergencies in real time, delaying the discovery and resolution of problems, and may lead to more serious equipment failures and environmental pollution. The data of each device needs to be manually integrated and analyzed. The data volume is large and complex, and relying on manual analysis and judgment requires a lot of human resources, and the training and management costs are high. In addition, it is easy to miss important information or make errors. Inaccurate data analysis results will affect the quality and effectiveness of operation and maintenance decisions. Summary of the invention

[0006] The present invention provides a water environment operation and maintenance monitoring method and system based on data analysis, which can effectively solve the problems in the background technology.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] Water environment operation and maintenance monitoring methods based on data analysis include:

[0009] Collect pre-processing data and post-processing data for water environment;

[0010] Align the timestamps of front-end processing data and back-end processing data;

[0011] Matching the front-end processing data at the previously set acquisition time with the back-end processing data at the later set acquisition time to obtain a data set corresponding to the staggered time;

[0012] Using a long short-term memory network time series model, inputting the data set;

[0013] Implementing training, evaluation and optimization of the time series model;

[0014] The water environment is predicted by the optimized time series model.

[0015] Furthermore, the timestamps of the front-end processing data and the back-end processing data are aligned, including:

[0016] Determine whether the acquisition frequencies of the front-end processing data and the back-end processing data are consistent. If so, directly align the timestamps; if not, perform interpolation processing on the side with lower acquisition frequency, and the interpolation processing is used to fill in the missing values ​​so that the data amounts of the front-end processing data and the back-end processing data are consistent, and then align the timestamps.

[0017] Furthermore, the staggered time corresponding to the staggered time is determined by the following steps:

[0018] Obtain the front-end processing data and back-end processing data within the set time;

[0019] The cross correlation between the front-end processing data and the back-end processing data is calculated, and the cross correlation function is:

[0020]

[0021] Among them, a t is the value of the front-end processing data at time t, k is the time lag step number, b t+k is the value of the post-processing data after a lag of k steps at time t, and are the average values ​​of the front-end processing data and the back-end processing data, respectively, and N is the number of data points;

[0022] The staggered time is calculated by using the lag time step number k that maximizes the value of the cross-correlation function.

[0023] Furthermore, the long short-term memory network time series model includes:

[0024] An input layer, receiving the dataset as input;

[0025] The LSTM layer consists of several LSTM units, each of which contains a memory unit and three gating mechanisms to control the flow of information;

[0026] Several fully connected layers receive the output from the LSTM layer and perform processing and feature conversion to map the high-dimensional features to the required dimensions;

[0027] The output layer outputs the prediction results based on the feature transformation results.

[0028] Furthermore, the number of the LSTM layers is a linear function of the number of the fully connected layers, and the number of the fully connected layers is less than the number of the LSTM layers.

[0029] Furthermore, the relationship between the number of the LSTM layers and the number of the fully connected layers is:

[0030] N LSTM =α×N FC ;

[0031] Among them, N LSTM is the number of LSTM layers, N FC is the number of the fully connected layers; α is a positive integer and is positively correlated with the difference in acquisition frequency between the front-end processing data and the back-end processing data.

[0032] Furthermore, a peeking connection is added between the memory unit of the long short-term memory network time series model and the three gating mechanisms.

[0033] Furthermore, the training, evaluation and optimization of the time series model are implemented, including:

[0034] Converting the data set and dividing the data set into a training set, a validation set and a test set;

[0035] Compile the time series model, select the loss function, optimizer and evaluation metric;

[0036] Setting hyperparameters of the time series model, iteratively training the time series model using the training set, and evaluating the time series model using the validation set during the iterative training, and adjusting the time series model according to the evaluation result of the validation set;

[0037] Use the test set to evaluate the performance of the time series model and calculate the evaluation index;

[0038] According to the evaluation results of the test set, the time series model is optimized, and the optimized time series model is deployed in the water environment monitoring system.

[0039] Further, converting the data set and dividing the data set into a training set, a validation set and a test set includes:

[0040] Defining a time window, and creating a data pair based on the time window, wherein the data pair includes feature data and label data;

[0041] Dividing the data set into a training set, a validation set and a test set;

[0042] The shape of the feature data is adjusted to match the input of the time series model.

[0043] A water environment operation and maintenance monitoring system based on data analysis, the system comprising:

[0044] Water environment data collection module, which collects pre-processing data and post-processing data for water environment;

[0045] The timestamp alignment module aligns the timestamps of the front-end processing data and the back-end processing data;

[0046] The water environment data matching module matches the front-end processing data at the previously set collection time with the back-end processing data at the later set collection time to obtain the corresponding data set at different times;

[0047] A long short-term memory network model module adopts a long short-term memory network time series model and inputs the data set;

[0048] A time series model training module, which implements the training, evaluation and optimization of the time series model;

[0049] The time series model prediction module predicts the water environment through the optimized time series model.

[0050] The technical solution of the present invention can achieve the following technical effects:

[0051] The present invention can significantly improve the effects of water environment operation and maintenance monitoring and prediction. Through comprehensive data operation and maintenance, it can effectively solve the current subjectivity, real-time and data integration problems, ensure the efficient and stable operation of the water treatment system, and realize the sustainable use of water resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 It is a flow chart of the water environment operation and maintenance monitoring method based on data analysis;

[0054] Figure 2 It is a structural diagram of the long short-term memory network time series model;

[0055] Figure 3 A flowchart for converting a data set and dividing the data set into a training set, a validation set, and a test set;

[0056] Figure 4 This is a structural diagram of the water environment operation and maintenance monitoring system based on data analysis. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0059] Embodiment 1

[0060] like Figure 1 As shown in the figure, the water environment operation and maintenance monitoring method based on data analysis includes:

[0061] A1: Collect the pre-processing data and post-processing data for water environment;

[0062] A2: Align the timestamps of the front-end processing data and the back-end processing data;

[0063] A3: Matching the front-end processing data of the collection time set earlier and the back-end processing data of the collection time set later to obtain the data set corresponding to the staggered time;

[0064] A4: Use the long short-term memory network time series model and input the data set;

[0065] A5: Implement training, evaluation and optimization of time series models;

[0066] A6: Predict the water environment through the optimized time series model.

[0067] In this embodiment, the use of long short-term memory network can capture the long-term and short-term dependencies in time series data, which is particularly important for water treatment systems, because the effect of the previous treatment is often reflected in the subsequent processed data after a period of time. Through the use of this model, the accuracy and reliability of the prediction can be improved, and the dynamic behavior of the system can be better understood and predicted. The long short-term memory network can automatically extract important features from the input data without the need for complex manual feature engineering, reducing the complexity and error of human operations and improving the adaptability and generalization ability of the model.

[0068] In actual scenarios, there may be complex nonlinear relationships between the variables in the water treatment system. The time series model can effectively handle these nonlinear relationships. Through continuous time series data input, the prediction results can be updated in real time, so as to timely discover and warn of potential problems, ensure the stable operation of the system, and reduce the impact of emergencies on the system.

[0069] Since the impact of the front-end treatment on the subsequent treatment in the water treatment process is not instantaneous, there is a certain time delay. The present invention can capture this time delay effect through staggered correspondence, thereby more accurately reflecting the actual operating state of the system, improving the model's understanding and prediction capabilities of time dynamics, and reducing the noise and errors caused by data asynchrony through staggered correspondence, ensuring the accuracy and consistency of input data. In actual systems, the impact of the front-end treatment effect on the subsequent treatment is gradually apparent, and staggered correspondence can better reflect this gradual impact, making the model closer to the actual operating conditions and the prediction results more realistic.

[0070] The present invention can significantly improve the effects of water environment operation and maintenance monitoring and prediction, solve the current subjectivity, real-time and data integration problems, ensure the efficient and stable operation of the water treatment system, and realize the sustainable use of water resources.

[0071] Furthermore, the timestamps of the front-end processing data and the back-end processing data are aligned, including:

[0072] Determine whether the acquisition frequencies of the front-end processing data and the back-end processing data are consistent. If so, align the timestamps directly; if not, perform interpolation processing on the side with lower acquisition frequency. Interpolation processing is used to fill in missing values ​​to make the data volume of the front-end processing data and the back-end processing data consistent, and then align the timestamps.

[0073] Specifically, the acquisition frequencies of the front-end processing data and the back-end processing data are consistent, and no additional processing is required; if the acquisition frequencies are inconsistent, the party with the lower acquisition frequency needs to be interpolated. The purpose of interpolation is to fill in the missing values ​​of the data set with a lower acquisition frequency so that its data points are consistent with the data points of the data set with a higher acquisition frequency. Interpolation methods such as linear interpolation, spline interpolation, and polynomial interpolation can be used for interpolation. By determining whether the acquisition frequencies are consistent and interpolating the party with the lower acquisition frequency, the problem of inconsistent data caused by different acquisition frequencies can be effectively solved. At the same time, interpolation can fill in missing values ​​and reduce data loss caused by inconsistent acquisition frequencies, thereby improving data integrity.

[0074] As a preferred embodiment of the above, the staggered time corresponding to the staggered time is determined by the following steps:

[0075] B1: Obtain the front-end processing data and back-end processing data within the set time. In the specific implementation process, the sampling time interval needs to be set, and the length of the set time is determined according to the sampling time interval to ensure sufficient data volume;

[0076] B2: Calculate the cross-correlation between the front-end processing data and the back-end processing data. The cross-correlation function is:

[0077]

[0078] Among them, a t is the value of the previous processing data at time t, k is the time lag step number, b t+k is the value of the post-processing data after a lag of k steps at time t, and are the average values ​​of the front-end processing data and the back-end processing data, respectively, and N is the number of data points; wherein the time lag step number k refers to the step difference between the front-end processing data and the back-end processing data. For example, a lag of 3 steps means that the back-end processing data lags behind the front-end processing data by 3 sampling points, i.e., 3 sampling time intervals;

[0079] B3: Use the lag time step number k that maximizes the value of the cross-correlation function to calculate the time delay, which is equal to the lag time step number k multiplied by the sampling time interval of each time step.

[0080] In this preferred solution, the time dependency between the front-end processing data and the back-end processing data can be accurately captured through cross-correlation analysis, and the optimal staggered time can be determined to make data alignment more accurate. This method of confirming the staggered time can reduce the noise and errors caused by data asynchrony and improve the quality of data alignment.

[0081] In the implementation process, when the value of the cross-correlation function reaches the maximum, the optimal staggered time is calculated by the lag time step k for the following reasons:

[0082] (1) Strongest correlation:

[0083] The maximum cross-correlation value indicates that the correlation between the two time series is the strongest at the lag time step k, which means that the change pattern of the front-end processing data can best explain the change of the back-end processing data;

[0084] (2) Capturing actual time delay:

[0085] In the water treatment process, the effect of the previous treatment often takes a certain amount of time to be reflected in the data of the subsequent treatment. The lag time step k corresponding to the maximum cross-correlation can reflect this actual time delay as much as possible.

[0086] (3) Improve forecast accuracy:

[0087] Using the lag time corresponding to the maximum cross-correlation to align data can reduce the noise and errors caused by data asynchrony, improve data quality, and thus enhance the accuracy and reliability of the input data of the prediction model.

[0088] As a preferred embodiment of the above, Figure 2 As shown, the long short-term memory network time series model includes:

[0089] The input layer accepts a data set as input; the input layer accepts a time series data set after the front-end processing data and the back-end processing data are aligned as input.

[0090] The LSTM layer is composed of several LSTM units, each of which contains a memory unit and three gating mechanisms to control the flow of information. The LSTM layer is composed of several LSTM units and is used to process time series data and capture the time dependency and sequence information in the data. The memory unit is used to store the long-term dependency information of the time series. The three gating mechanisms include the input gate, the forget gate, and the output gate, which together control the flow of information in the memory unit. The input gate controls the proportion of new information entering the memory unit, the forget gate determines which information in the memory unit needs to be forgotten, and the output gate controls which information in the memory unit is output. Through these gating mechanisms, LSTM can effectively capture and retain important information in time series data and reduce the gradient disappearance and explosion problems.

[0091] Several fully connected layers receive the output from the LSTM layer, and perform processing and feature conversion to map the high-dimensional features to the required dimensions. The fully connected layer is usually placed after the LSTM layer, responsible for receiving the output of the LSTM layer and performing further processing and feature conversion. Through the processing of several fully connected layers, the model can better extract and convert the high-dimensional features output by the LSTM layer, thereby improving the prediction performance and accuracy of the model.

[0092] The output layer outputs the prediction results based on the feature transformation results. Nonlinear activation functions can be applied in the fully connected layer and the output layer to introduce nonlinear characteristics, such as ReLU, Leaky ReLU, tanh, etc., to enhance the nonlinear representation ability of the model and improve the model's fitting ability.

[0093] In order to make the performance of the long short-term memory network time series model in the above embodiment better, the number of LSTM layers is a linear function of the number of fully connected layers, and the number of fully connected layers is less than the number of LSTM layers.

[0094] Through this optimization, the model complexity and computational efficiency can be effectively balanced. In water environment monitoring, the data from front-end and back-end processing contain a large number of time-dependent features. The LSTM layer extracts these complex features through a deeper network structure to understand the trends and laws of water quality changes. The number of fully connected layers is smaller than that of the LSTM layer, which reduces the computational complexity and training time of the model, improves the efficiency of real-time monitoring and prediction, and ensures that the system can respond quickly.

[0095] During the implementation process, by setting the number of LSTM layers as a linear function of the number of fully connected layers, the model can effectively perform feature conversion and dimensionality reduction while extracting time series features. The use of linear relationships makes the network structure design intuitive, reduces the complexity of hyperparameter tuning, and improves work efficiency. By setting a reasonable ratio, the number of layers can be quickly determined, reducing experimental and adjustment time, and quickly deploying effective monitoring and prediction models.

[0096] As a preference for the above embodiment, the relationship between the number of LSTM layers and the number of fully connected layers is:

[0097] N LSTM =α×N FC ;

[0098] Among them, N LSTM is the number of LSTM layers, N FC is the number of fully connected layers; α is a positive integer and is positively correlated with the difference in acquisition frequency between the front-end processing data and the back-end processing data.

[0099] When the acquisition frequency of the front-end processing data and the back-end processing data is very different, the time delay and dependency between the data will be more complex. In this case, increasing the number of LSTM layers can help better model these complex dependencies, thereby improving prediction accuracy. α is positively correlated with the acquisition frequency difference, so that the number of LSTM layers can be dynamically adjusted according to the time-dependent characteristics of the data, so that the model can maintain good performance in different monitoring scenarios and improve generalization and adaptability.

[0100] As a preferred embodiment of the above, a peek connection is added between the memory unit and the three gating mechanisms of the long short-term memory network time series model. The peek connection is a mechanism to enhance the long short-term memory network, so that each gating mechanism can directly access the state of the memory unit. In the standard long short-term memory network, the calculation of each gate depends only on the current input and the hidden state of the previous moment. After the introduction of the peek connection, the forget gate, input gate and output gate are allowed to directly use the memory unit state of the previous moment in addition to relying on the current input and the hidden state of the previous moment during calculation, which can better capture and utilize the long-term and short-term dependencies in the time series data, allowing information to flow more flexibly between the memory unit and the gating mechanism, thereby improving the expressive power of the model. It should be noted that the introduction of the peek connection will increase the number of parameters of the model, thereby increasing the risk of overfitting, which can be controlled in combination with regularization methods, such as Dropout regularization, elastic network regularization and batch normalization.

[0101] Furthermore, the training, evaluation and optimization of time series models are implemented, including:

[0102] C1: transform the data set and divide the data set into training set, validation set and test set;

[0103] Specifically, time series data is essentially a series of data points that change over time. In order to apply time series models for prediction, time series data can be converted into the format of supervised learning problems, that is, data pairs of input features and target variables; after converting the data, the data set can be divided into training set, validation set and test set, for example, 70% training set, 15% validation set, and 15% test set; when dividing the data set, ensure the randomness of the data to avoid deviations in the division process, and the obtained training set, validation set and test set should maintain the same data distribution to ensure the reliability of the evaluation results.

[0104] C2: Compile the time series model, select the loss function, optimizer and evaluation metric;

[0105] Before training the model, you can select appropriate loss functions, optimizers, and evaluation indicators to compile the model; the loss function is a function that measures the gap between the model's predicted value and the actual value, including mean square error, mean absolute error, cross entropy loss, etc.; the optimizer is used to update the model's weight parameters to minimize the loss function, including gradient descent, stochastic gradient descent, Adam optimizer, etc. Different optimizers have different update rules and hyperparameters; the evaluation indicators set when compiling the model are not used to optimize the model, but they can provide useful information during training and evaluation. Mean square error, root mean square error, mean absolute error, etc. can be used as evaluation indicators.

[0106] C3: Set the hyperparameters of the time series model, use the training set to iteratively train the time series model, and use the validation set to evaluate the time series model during the iterative training process. Adjust the time series model based on the evaluation results of the validation set.

[0107] On the basis of the above embodiment, the hyperparameters of the model can be set, including learning rate, batch size, number of training rounds, etc., and then the model can be iteratively trained using the training set, and the validation set can be used for evaluation during the training process; the specific process is to divide the training data into multiple small batches, each small batch contains a certain number of samples, forward propagate the data of each small batch, propagate the input data through the network, calculate the output of each layer until the final prediction result is obtained, use the loss function to calculate the difference between the prediction result of each small batch and the true value, and then back propagate the data of each small batch to calculate the gradient of the loss relative to each weight , these gradients are used to update the weights. The gradients can be calculated by the chain rule. The gradient of the loss relative to each parameter is calculated layer by layer. The calculated gradients can be used to update the weights. The weights of the model are updated through the optimizer. The above steps are repeated for all small batches until the entire training data set is traversed. After each training cycle, the performance of the model is evaluated on the validation set, and the hyperparameters are adjusted according to the performance of the validation set to optimize the model performance. This helps monitor the training progress of the model and detect overfitting. In order to better prevent overfitting and improve the generalization ability of the model, the model can be iteratively trained and the above steps are repeated until the loss function converges or the preset number of training rounds is reached.

[0108] C4: Use the test set to evaluate the performance of the time series model and calculate the evaluation indicators;

[0109] C5: Based on the evaluation results of the test set, the time series model is optimized and deployed into the water environment monitoring system.

[0110] After all adjustments are completed in this step, the model is retrained. The final trained model can be used to evaluate on the test set and calculate the evaluation indicators. If the performance on the test set is not ideal, the model may need to be further adjusted. According to the evaluation results of the test set, the model structure, hyperparameters, or other modifications can be adjusted to improve model performance. After the adjustment, the model needs to be retrained and evaluated using the updated test set. The test set data is used to evaluate the performance of the model on unseen data to ensure the generalization ability of the model. At the same time, the results of the test set evaluation will provide the performance of the model in actual applications.

[0111] As a preferred embodiment of this embodiment, Figure 3 As shown, the data set is transformed and divided into a training set, a validation set, and a test set, including:

[0112] D1: Define a time window and create data pairs based on the time window. The data pairs include feature data and label data.

[0113] D2: Divide the data set into training set, validation set and test set;

[0114] D3: Adjust the shape of feature data to match the input of the time series model.

[0115] This step converts the time series data into a format suitable for supervised learning. The time window can be defined to determine how many past time steps of data are used to predict one or more future time steps. This step determines how much past data the model uses for prediction. The appropriate window length can effectively capture the temporal dependency of the data. Based on the time window, input features and corresponding labels are generated. The input features are the data within the time window, and the labels are the prediction targets after the time window, so that the time series data can be used as training data for supervised learning algorithms. The data set is divided into training set, validation set, and test set to evaluate the performance of the model and train and tune the model. In addition, the input data of the model needs to have a specific dimension. The shape of the input feature data can be adjusted to meet the requirements of the model input to ensure that the data can be correctly processed by the model. Since different data sets need to have consistent formats and shapes to ensure that the model can be correctly trained and evaluated, but the adjustment of the data shape may cause the data of the training set and test set to no longer represent the original distribution, thereby affecting the fairness of the model evaluation, the shape of the data can be adjusted after the data is divided.

[0116] Embodiment 2

[0117] like Figure 4 As shown, the water environment operation and maintenance monitoring system based on data analysis includes:

[0118] Water environment data collection module, which collects pre-processing data and post-processing data for water environment;

[0119] The timestamp alignment module aligns the timestamps of the front-end processing data and the back-end processing data;

[0120] The water environment data matching module matches the front-end processing data at the previously set collection time with the back-end processing data at the later set collection time to obtain the corresponding data set at different times;

[0121] The long short-term memory network model module uses the long short-term memory network time series model and inputs a data set;

[0122] Time series model training module, which implements the training, evaluation and optimization of time series models;

[0123] The time series model prediction module predicts the water environment through the optimized time series model. The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by technicians in this industry that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of 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 water environment operation and maintenance monitoring method based on data analysis, characterized in that: include: Collect pre-processing data and post-processing data for water environment; Align the timestamps of front-end processing data and back-end processing data; Matching the front-end processing data at the previously set acquisition time with the back-end processing data at the later set acquisition time to obtain a data set corresponding to the staggered time; Using a long short-term memory network time series model, inputting the data set; Implementing training, evaluation and optimization of the time series model; The water environment is predicted by the optimized time series model.

2. The water environment operation and maintenance monitoring method based on data analysis according to claim 1 is characterized in that: Performing time stamp alignment on the front-end processing data and the back-end processing data includes: Determine whether the acquisition frequencies of the front-end processing data and the back-end processing data are consistent. If so, directly align the timestamps; if not, perform interpolation processing on the side with lower acquisition frequency, and the interpolation processing is used to fill in the missing values ​​so that the data amounts of the front-end processing data and the back-end processing data are consistent, and then align the timestamps.

3. The water environment operation and maintenance monitoring method based on data analysis according to claim 2 is characterized in that: The staggered time corresponding to the staggered time is determined by the following steps: Obtain the front-end processing data and back-end processing data within the set time; The cross correlation between the front-end processing data and the back-end processing data is calculated, and the cross correlation function is: Among them, a t is the value of the front-end processing data at time t, k is the time lag step number, b t+k is the value of the post-processing data after a lag of k steps at time t, and are the average values ​​of the front-end processing data and the back-end processing data, respectively, and N is the number of data points; The staggered time is calculated by using the lag time step number k that maximizes the value of the cross-correlation function.

4. The water environment operation and maintenance monitoring method based on data analysis according to any one of claims 1 to 3, characterized in that: The long short-term memory network time series model includes: An input layer, receiving the dataset as input; The LSTM layer consists of several LSTM units, each of which contains a memory unit and three gating mechanisms to control the flow of information; Several fully connected layers receive the output from the LSTM layer and perform processing and feature conversion to map the high-dimensional features to the required dimensions; The output layer outputs the prediction results based on the feature transformation results.

5. The water environment operation and maintenance monitoring method based on data analysis according to claim 4 is characterized in that: The number of the LSTM layers is a linear function of the number of the fully connected layers, and the number of the fully connected layers is less than the number of the LSTM layers.

6. The water environment operation and maintenance monitoring method based on data analysis according to claim 5 is characterized in that: The relationship between the number of the LSTM layers and the number of the fully connected layers is: N LSTM =α×N FC ; Among them, N LSTM is the number of LSTM layers, N FC is the number of the fully connected layers; α is a positive integer and is positively correlated with the difference in acquisition frequency between the front-end processing data and the back-end processing data.

7. The water environment operation and maintenance monitoring method based on data analysis according to claim 4 is characterized in that: Adding a peek connection between the memory unit and the three gating mechanisms of the long short-term memory network time series model.

8. The water environment operation and maintenance monitoring method based on data analysis according to claim 1 is characterized in that: Implement the training, evaluation and optimization of the time series model, including: Converting the data set and dividing the data set into a training set, a validation set and a test set; Compile the time series model, select the loss function, optimizer and evaluation metric; Setting hyperparameters of the time series model, iteratively training the time series model using the training set, and evaluating the time series model using the validation set during the iterative training, and adjusting the time series model according to the evaluation result of the validation set; Use the test set to evaluate the performance of the time series model and calculate the evaluation index; According to the evaluation results of the test set, the time series model is optimized, and the optimized time series model is deployed in the water environment monitoring system.

9. The water environment operation and maintenance monitoring method based on data analysis according to claim 8 is characterized in that: Converting the data set and dividing the data set into a training set, a validation set, and a test set, including: Defining a time window, and creating a data pair based on the time window, wherein the data pair includes feature data and label data; Dividing the data set into a training set, a validation set and a test set; The shape of the feature data is adjusted to match the input of the time series model.

10. The water environment operation and maintenance monitoring system based on data analysis is characterized by: The system comprises: Water environment data collection module, which collects pre-processing data and post-processing data for water environment; The timestamp alignment module aligns the timestamps of the front-end processing data and the back-end processing data; The water environment data matching module matches the front-end processing data at the previously set collection time with the back-end processing data at the later set collection time to obtain the corresponding data set at different times; A long short-term memory network model module adopts a long short-term memory network time series model and inputs the data set; A time series model training module, which implements the training, evaluation and optimization of the time series model; The time series model prediction module predicts the water environment through the optimized time series model.

Images