A landfill leakage risk prediction method based on deep learning

Abstract

The present invention provides a landfill leakage risk prediction method based on deep learning, which belongs to the field of landfill leakage risk prediction, and comprises the following steps: obtaining leakage liquid data to be tested, performing standard processing on the leakage liquid data to be tested, obtaining standard data to be tested, inputting the standard data to be tested into a trained landfill leakage risk prediction model, and obtaining a landfill leakage risk prediction result; the present invention obtains a landfill leakage risk prediction model by using a long short-term memory recurrent neural network-gated recurrent unit neural network algorithm, so that the prediction model is not only applicable to long-term dependent sequence data, but also applicable to the processing of short-term information, thereby improving the practicability and reliability of model prediction.

Description

Technical Field

[0001] The present invention relates to the field of landfill leakage risk prediction, and in particular to a landfill leakage risk prediction method based on deep learning. Background Art

[0002] The research on landfill leakage risk has attracted much attention in recent years, especially in the use of deep learning technology to predict leakage risk and its possibility. The current research involves several aspects such as data collection and processing, application of deep learning models, feature extraction and prediction, and model optimization and verification.

[0003] Most of the current landfill leakage risk prediction methods based on deep learning use long short-term memory recurrent neural networks and gated recurrent unit schemes. However, long short-term memory recurrent neural networks are only applicable to long-term dependent sequence data, and gated recurrent units are only applicable to short-term information processing. In order to solve the above problems, the present invention proposes a landfill leakage risk prediction method based on deep learning using a long short-term memory recurrent neural network-gated recurrent unit scheme. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a landfill leakage risk prediction method based on deep learning, which is not only suitable for long-term dependent sequence data, but also for the processing of short-term information to improve the practicability and reliability of model prediction.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A landfill leakage risk prediction method based on deep learning, including:

[0007] Acquire the data of the leakage liquid to be tested, and perform standardization processing on the data of the leakage liquid to be tested to obtain standard data to be tested;

[0008] The standard data to be tested is input into a trained landfill leakage risk prediction model to obtain a landfill leakage risk prediction result; the landfill leakage risk prediction model construction step comprises:

[0009] Collect sample leachate data from target landfills;

[0010] Preprocessing the sample leachate data to obtain preprocessed data;

[0011] Randomly divide the preprocessed data into a training set and a test set;

[0012] Build a long short-term memory recurrent neural network-gated recurrent unit neural network structure to obtain the original training model;

[0013] Inputting the training set into the original training model for training to obtain a predicted value;

[0014] Calculating the predicted value and the loss data of the test set based on a loss function;

[0015] Determine whether the loss data converges. If so, obtain the landfill leakage risk prediction model. If not, use the optimizer to update the model parameters of the long short-term memory recursive neural network-gated recurrent unit neural network structure and return to the step of "inputting the training set into the original training model for training".

[0016] Preferably, a posterior distribution calculation is performed on the landfill leakage risk prediction model to obtain posterior distribution data, and the fitting of the landfill leakage risk prediction model is evaluated based on the posterior distribution data.

[0017] Preferably, the leachate data includes: leakage volume data and drainage volume data.

[0018] Preferably, the original training model structure is built in PyTorch.

[0019] Preferably, the loss function includes: mean square error, root mean square error, mean absolute error and mean absolute percentage error.

[0020] Preferably, the optimizer is an adaptive moment estimation algorithm.

[0021] Preferably, the preprocessing includes: missing value processing, outlier detection and feature standardization.

[0022] Preferably, the posterior distribution includes: coefficient of determination, mean square error, root mean square error, mean absolute error and mean absolute percentage error.

[0023] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0024] The present invention provides a landfill leakage risk prediction method based on deep learning, which belongs to the field of landfill leakage risk prediction, and comprises the following steps: obtaining leakage liquid data to be tested, performing standard processing on the leakage liquid data to be tested, obtaining standard data to be tested, inputting the standard data to be tested into a trained landfill leakage risk prediction model, and obtaining a landfill leakage risk prediction result; the present invention obtains a landfill leakage risk prediction model by using a long short-term memory recursive neural network-gated recurrent unit neural network algorithm, so that the prediction model is not only applicable to long-term dependent sequence data, but also applicable to the processing of short-term information, thereby improving the practicability and reliability of model prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] 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 will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0026] Figure 1 A flow chart of landfill leakage risk prediction based on deep learning provided by an embodiment of the present invention;

[0027] Figure 2 The leakage prediction value and the actual value fitting effect diagram provided by the embodiment of the present invention are as follows: Figure 2 (a) is a comparison chart of the predicted and true values ​​of the leakage volume of the LSTM model training set, validation set, and test set. Figure 2 (b) is a comparison chart of the predicted and true values ​​of the leakage volume of the GRU model training set, validation set, and test set. Figure 2 (c) is a comparison chart of the predicted and true values ​​of the leakage volume of the LSTM-GRU model training set, validation set, and test set;

[0028] Figure 3 This is a fitting effect diagram of the predicted value and the actual value of the drainage volume provided by the embodiment of the present invention. Figure 3 (a) is a comparison chart of the predicted and true values ​​of the drainage volume of the LSTM model training set, validation set, and test set. Figure 3 (b) is a comparison chart of the predicted and true values ​​of the conduction volume of the GRU model training set, validation set, and test set. Figure 3 (c) is a comparison chart of the predicted and true values ​​of the drainage volume of the LSTM-GRU model training set, validation set, and test set;

[0029] Figure 4 The relationship diagram between the predicted value and the true value of the leakage amount during the training period, the test period and the verification period provided by the embodiment of the present invention is as follows: Figure 4 (a) is a comparison chart of the predicted value and the true value of the leakage during the LSTM model training period. Figure 4 (b) is a comparison chart of the predicted value and the actual value of the leakage during the LSTM model test period. Figure 4 (c) is a comparison chart of the predicted value and the actual value of the leakage during the LSTM model verification period. Figure 4 (d) is a comparison chart of the predicted value and the true value of the leakage during the GRU model training period. Figure 4 (e) is a comparison chart of the predicted value and the actual value of the leakage during the GRU model test period. Figure 4 (f) is a comparison chart of the predicted value and the true value of the leakage during the GRU model validation period. Figure 4 (g) is a comparison chart of the predicted value and the actual value of the leakage during the LSTM-GRU model training period. Figure 4(h) is a comparison chart of the predicted value and the actual value of the leakage during the LSTM-GRU model test period. Figure 4 (i) is a comparison chart of the predicted value and the actual value of the leakage during the validation period of the LSTM-GRU model;

[0030] Figure 5 The relationship diagram between the predicted value and the true value of the drainage volume during the training period, the test period and the verification period provided by the embodiment of the present invention is as follows: Figure 5 (a) is a comparison chart of the predicted value and the true value of the drainage volume during the LSTM model training period. Figure 5 (b) is a comparison chart of the predicted value and the actual value of the drainage volume during the LSTM model test period. Figure 5 (c) is a comparison chart of the predicted value and the true value of the drainage volume during the LSTM model verification period. Figure 5 (d) is a comparison chart of the predicted value and the actual value of the conduction volume during the GRU model training period. Figure 5 (e) is a comparison chart of the predicted value and the actual value of the conduction volume during the GRU model test period. Figure 5 (f) is a comparison chart between the predicted value and the true value of the conduction volume during the GRU model verification period. Figure 5 (g) is a comparison chart between the predicted value and the actual value of the conduction volume during the LSTM-GRU model training period. Figure 5 (h) is a comparison chart between the predicted value and the actual value of the conduction volume during the test period of the LSTM-GRU model. Figure 5 (i) is a comparison chart of the predicted value and the actual value of the drainage volume during the LSTM-GRU model verification period;

[0031] Figure 6 A flowchart of the prediction model framework provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] 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.

[0033] The purpose of the present invention is to provide a landfill leakage risk prediction method based on deep learning, which is not only applicable to long-term dependent sequence data, but also to the processing of short-term information to improve the practicality and reliability of model prediction.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] Figure 1 The flow chart of landfill leakage risk prediction based on deep learning is as follows: Figure 1 As shown, the present invention provides a landfill leakage risk prediction method based on deep learning, comprising:

[0036] Step 100, obtaining data of the leakage liquid to be tested, and performing standardization processing on the data of the leakage liquid to be tested to obtain standard data to be tested;

[0037] Step 200, inputting the standard data to be tested into the trained landfill leakage risk prediction model to obtain the landfill leakage risk prediction result; the landfill leakage risk prediction model construction step includes:

[0038] Collect sample leachate data from target landfills;

[0039] Preprocessing the sample leachate data to obtain preprocessed data;

[0040] Randomly divide the preprocessed data into a training set and a test set;

[0041] Build a long short-term memory recurrent neural network-gated recurrent unit neural network structure to obtain the original training model;

[0042] Inputting the training set into the original training model for training to obtain a predicted value;

[0043] Calculating the predicted value and the loss data of the test set based on a loss function;

[0044] Determine whether the loss data converges. If so, obtain the landfill leakage risk prediction model. If not, use the optimizer to update the model parameters of the long short-term memory recursive neural network-gated recurrent unit neural network structure and return to the step of "inputting the training set into the original training model for training".

[0045] Furthermore, a posterior distribution calculation is performed on the landfill leakage risk prediction model to obtain posterior distribution data, and the fitting of the landfill leakage risk prediction model is evaluated based on the posterior distribution data.

[0046] Specifically, the leachate data includes: leakage volume data and drainage volume data.

[0047] Optionally, the original training model structure is built in PyTorch.

[0048] Optionally, the loss function includes: mean square error, root mean square error, mean absolute error and mean absolute percentage error.

[0049] Specifically, the optimizer is an adaptive moment estimation algorithm.

[0050] Specifically, the preprocessing includes: missing value processing, outlier detection and feature standardization.

[0051] Optionally, the posterior distribution includes: coefficient of determination, mean square error, root mean square error, mean absolute error and mean absolute percentage error.

[0052] Furthermore, LSTM and GRU are two commonly used recurrent neural network structures that perform well in time series data modeling. LSTM is outstanding in its ability to capture long-term dependencies, while GRU can show similar performance with fewer parameters. LSTM-GRU combines the advantages of long short-term memory networks (LSTM) and gated recurrent units (GRU) to overcome each other's limitations and achieve more accurate and robust leakage risk prediction to improve the prediction performance and robustness of the model. The composite model first uses the LSTM network to learn the long-term dependencies in the landfill time series data and capture the complex patterns and trends therein. Then, by introducing the GRU network, the model's grasp of short-term sequence information is strengthened, and the model's ability to capture local features of the data is improved. This combination effectively balances the use of long-term and short-term information and improves the comprehensiveness and robustness of the prediction model.

[0053] Specifically, the overall framework of the LSTM-GRU prediction model includes five functional modules: input layer, hidden layer, output layer, network training and network prediction. The input layer is responsible for the preliminary processing of the real-time monitoring data of leachate to meet the network input requirements. The hidden layer uses a network constructed by LSTM-GRU units, and the output layer provides prediction results. The network training uses stochastic gradient descent (SGD) as the optimizer, and the network prediction uses an iterative method.

[0054] Specifically, this embodiment analyzes 3643 sets of simulation data, of which 2914 sets of measured data are used as training data sets for the LSTM model, the GRU model, and the LSTM-GRU model, and the remaining 728 sets of measured data are used as test and verification data sets for the three models.

[0055] Furthermore, a neural network of LSTM-GRU combination was built in PyTorch. In addition, during the training process, the difference between the model prediction results and the actual labels was evaluated for multi-classification problems by using the CrossEntropyLoss loss function, and the Adam optimizer was selected for parameter optimization, where the learning rate lr parameter affects the step size and convergence speed of the model parameter update and is set to 0.001.

[0056] refer to Figure 2 and Figure 3Of the 3643 sets of simulation data used in this embodiment, 2914 sets are used for training, 364 sets are used for testing, and 364 sets are used for verification. The features cover leakage and drainage, as well as influencing factors such as landfill environment and landfill conditions. Three neural network models were built: LSTM, GRU, and LSTM-GRU, which are used to predict leakage and drainage respectively. LSTM is suitable for sequence data with long-term dependence and memory, GRU focuses more on processing information in the short term, and LSTM-GRU combines the advantages of both. The results of the fitting effect of the predicted values ​​of leakage and drainage of the three neural network prediction models with the true values ​​are shown in the figure. Figure 2 and Figure 3 The chart reflects the fit between the model prediction value and the true value.

[0057] Furthermore, the model training process adopts steps such as data preprocessing, feature scaling and model optimization. In order to ensure the adaptability of the model to the data, cross-validation technology is used to ensure the randomness and representativeness of the training set and the validation set. Data preprocessing includes missing value processing, outlier detection and feature standardization to ensure the accuracy and consistency of the data. In the model training, different hyperparameter combinations are tried, and the best performing model parameters are selected through the effect evaluation of the validation set.

[0058] Specifically, after comprehensively considering the performance of the LSTM, GRU and LSTM-GRU models in the prediction of leakage and drainage, the differences in the generalization ability, the ability to handle abnormal data, and the ability to capture long-term and short-term features of each model were further explored. The adaptability and robustness of the model were analyzed more deeply for different scenarios and data characteristics. The importance of data preprocessing and model optimization was emphasized, thereby improving the performance and robustness of the model.

[0059] Furthermore, the performance indicators of the three different models (LSTM, GRU, LSTM-GRU) during the training and validation periods include the coefficient of determination (R2), mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE).

[0060] refer to Figure 4 and Figure 5 The relationship between the predicted value and the true value of the leakage and drainage volume during the training period, test period and verification period is shown in the figure below: Figure 4 and Figure 5As shown. First, during the training and validation periods, the LSTM-GRU model performed best in all indicators. Its R2 value was the highest and its error indicators (MSE, RMSE, MAE) were the lowest, and its MAPE was also the smallest among the three models. This shows that the LSTM-GRU model has better prediction performance and accuracy in predicting the amount of landfill seepage. Secondly, compared with the separate LSTM and GRU models, the GRU model is slightly better than the LSTM model. The GRU model has a higher R2 value than the LSTM model during the training and validation periods, and the error indicators (MSE, RMSE, MAE) are also smaller, and the performance is more robust. Overall, the LSTM-GRU model showed the best prediction ability, followed by the GRU model, while the LSTM model performed slightly worse. This comparison can help select the most suitable model for landfill seepage prediction. In summary, by comparing with the LSTM and GRU models, it is found that the composite model has more advantages; it can be concluded that the LSTM-GRU network is the best choice for predicting sequence data such as leakage and drainage.

[0061] Specifically, Figure 2 The relevant explanations are as follows: the x-axis corresponds to the date; the y-axis corresponds to the leakage amount; PredictedDischarge is the predicted value of leakage; True Discharge is the true value of leakage.

[0062] Specifically, Figure 3 The relevant explanations are as follows: the x-axis corresponds to the date; the y-axis corresponds to the discharge volume; PredictedDischarge is the predicted value of the discharge volume; True Discharge is the actual value of the discharge volume.

[0063] Specifically, Figure 4 The relevant explanations are as follows: the x-axis corresponds to the measured value; the y-axis corresponds to the leakage prediction value; TrainingPredictions are the training prediction values; TestPredictions are the test prediction values; ValidationPredictions are the validation prediction values.

[0064] Specifically, Figure 5 The relevant explanations are as follows: the x-axis corresponds to the measured value; the y-axis corresponds to the predicted value of the displacement; TrainingPredictions are the training prediction values; Test Predictions are the test prediction values; Validation Predictions are the validation prediction values.

[0065] The beneficial effects of the present invention are as follows:

[0066] The present invention utilizes a long short-term memory recursive neural network-gated recurrent unit neural network algorithm to obtain a landfill leakage risk prediction model, so that the prediction model is not only applicable to long-term dependent sequence data, but also to short-term information processing, thereby improving the practicability and reliability of model prediction.

[0067] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0068] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A landfill leakage risk prediction method based on deep learning, characterized in that: include: Acquire the data of the leakage liquid to be tested, and perform standardization processing on the data of the leakage liquid to be tested to obtain standard data to be tested; Inputting the standard data to be tested into a trained landfill leakage risk prediction model to obtain a landfill leakage risk prediction result; The landfill leakage risk prediction model construction steps include: Collect sample leachate data from target landfills; Preprocessing the sample leachate data to obtain preprocessed data; Randomly divide the preprocessed data into a training set and a test set; Build a long short-term memory recurrent neural network-gated recurrent unit neural network structure to obtain the original training model; Inputting the training set into the original training model for training to obtain a predicted value; Calculating the predicted value and the loss data of the test set based on a loss function; Determine whether the loss data converges, if so, obtain the landfill leakage risk prediction model, if not, use the optimizer to update the model parameters of the long short-term memory recurrent neural network-gated recurrent unit neural network structure and return to the step of "inputting the training set into the original training model for training"; Also includes: Performing posterior distribution calculation on the landfill leakage risk prediction model to obtain posterior distribution data, and evaluating the fitting of the landfill leakage risk prediction model according to the posterior distribution data; The original training model structure is built in PyTorch.

2. A landfill leakage risk prediction method based on deep learning according to claim 1, characterized in that: The leachate data includes: leakage volume data and drainage volume data.

3. The landfill leakage risk prediction method based on deep learning according to claim 1 is characterized in that: The loss functions include mean square error, root mean square error, mean absolute error and mean absolute percentage error.

4. The landfill leakage risk prediction method based on deep learning according to claim 1 is characterized in that: The optimizer is an adaptive moment estimation algorithm.

5. The landfill leakage risk prediction method based on deep learning according to claim 1 is characterized in that: The preprocessing includes: missing value processing, outlier detection and feature standardization.

6. The landfill leakage risk prediction method based on deep learning according to claim 1 is characterized in that: The posterior distribution includes: coefficient of determination, mean square error, root mean square error, mean absolute error and mean absolute percentage error.

Images