Liquefied natural gas (LNG) export quantity short-term prediction method and system coupled with gas-liquid export characteristics

Abstract

The invention relates to an LNG (Liquefied Natural Gas) export quantity short-term prediction method and system coupled with gas-liquid export characteristics, and the method comprises the steps: carrying out the OVMD of historical data and basic meteorological information, obtaining an intrinsic mode component, and reconstructing the intrinsic mode component; inputting the reconstructed data into an output short-term hybrid prediction model which comprises a deep convolutional neural network module and a long-short-term memory neural network model, and extracting potential correlation characteristics between the gaseous output and the liquid output through the deep convolutional neural network model; extracting time dependence characteristics between the gaseous output and the meteorological information as well as between the liquid output and the meteorological information through a long-short-term memory neural network model; and carrying out feature fusion on the associated features and the time-dependent features to obtain a prediction result of the LNG output. According to the method, the accuracy of LNG output prediction is improved, an operator is helped to manage energy resources more effectively, supply chain operation is optimized, the operation cost is reduced, meanwhile, the stability of energy supply is ensured, and the overall economic benefits and social benefits are improved.

Description

Technical Field

[0001] The invention relates to a method and system for short-term prediction of LNG (liquefied natural gas) export volume coupled with gas-liquid export characteristics, belonging to the technical field of LNG export volume prediction. Background Art

[0002] With the continuous growth of global energy demand and the improvement of environmental awareness, liquefied natural gas, as a clean energy source, has become increasingly prominent. LNG has the characteristics of high energy density, convenient transportation, flexibility, safety and efficiency. With the advancement of LNG liquefaction technology and shipping technology, LNG has gradually become the main form of global cross-border trade in natural gas. After LNG arrives at the port receiving station, it can be gasified by the gasifier and then exported in gaseous form, or it can be exported in liquid form. Regardless of the method used, it is necessary to meet the needs of downstream industrial, commercial and household gas. The natural gas market is highly volatile and is affected by many factors, including geopolitical factors, changes in climate conditions, etc. Therefore, accurate prediction of LNG gas and liquid export volume is of great significance for enterprises to reasonably allocate resources and formulate export plans.

[0003] LNG export volume forecasting is essentially a time series forecasting problem. Time series forecasting is a method of predicting future values ​​or trends by analyzing and mining patterns and rules in time series data. Existing methods include methods based on traditional statistics, such as the autoregressive moving average (ARMA) model, seasonal autoregressive moving average (SARMA) model, grey forecasting model, and neural network-based methods, such as recurrent neural network (RNN), long short-term memory network (LSTM), etc. These models have achieved good forecasting results in many scenarios, but for the LNG export volume forecasting problem, these traditional forecasting models cannot fully capture the local correlation between gaseous and liquid export volumes during LNG export, so the forecasting effect is average and it is difficult to meet the current LNG receiving station's demand for high-precision forecasting of export volume. Summary of the invention

[0004] In response to the above problems, the purpose of the present invention is to provide a method and system for short-term prediction of LNG export volume that couples gas-liquid export characteristics, which improves the accuracy of LNG export volume prediction, helps operators to manage energy resources more effectively, optimize supply chain operations, and reduce operating costs, while ensuring the stability of energy supply and improving overall economic and social benefits.

[0005] To achieve the above-mentioned purpose, the present invention proposes the following technical scheme: a method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics, comprising the following steps: collecting historical data of gas export volume and liquid export volume and basic meteorological information of LNG receiving stations; performing optimal variational mode decomposition (OVMD) on the historical data and basic meteorological information to obtain intrinsic mode components, and reconstructing the intrinsic mode components; inputting the reconstructed data into a short-term hybrid prediction model for export volume, the short-term hybrid prediction model for export volume comprising a deep convolutional neural network module and a long short-term memory neural network model, extracting potential correlation features between gas export volume and liquid export volume through the deep convolutional neural network model; extracting time-dependent features between gas export volume and liquid export volume and the meteorological information through the long short-term memory neural network model; and performing feature fusion of the correlation features with the time-dependent features to obtain a prediction result of LNG export volume.

[0006] Furthermore, the historical data of gas export volume and liquid export volume and basic meteorological information are preprocessed, and the preprocessing includes: removing outliers and noise through data cleaning technology; normalizing the cleaned data, and extracting key features that affect the LNG export volume from the normalized data.

[0007] Furthermore, the key features include maximum temperature, minimum temperature and average temperature.

[0008] Furthermore, the method for performing optimal variational mode decomposition (OVMD) is as follows: taking the superposition of each modal component equal to the original signal as a constraint condition, and deriving the model parameters: the total number of modal decompositions K , quadratic penalty factor α and Lagrange multiplier τ; through Fourier transform, the original sequence of LNG export volume is decomposed to obtain K narrowband intrinsic mode function components; by updating the model parameters, the decomposed K The narrowband intrinsic mode function components are updated until the optimal intrinsic mode function components that meet the constraints are obtained.

[0009] Furthermore, the center frequency method is used to determine the total number of modal decompositions K , the Lagrange multiplier τ is determined based on the minimization of the residual index. The formula of the residual index is as follows:

[0010] in, Indicates the length of the original signal of LNG export volume. is the residual indicator, Indicates kind of modality.

[0011] Short-term hybrid forecasting model for external output Furthermore, the intrinsic mode components generated by the optimal variational mode decomposition (OVMD) and the historical data of gaseous and liquid external transport are used as the first data set, and the intrinsic mode components and basic meteorological information are used as the second data set. The first data set and the second data set are input together into a short-term hybrid prediction model for external transport to train the model.

[0012] Furthermore, the short-term hybrid prediction model for external transmission volume includes an input layer, a gas-liquid external transmission feature extraction module and an output layer; the input layer is used to input the external transmission volume information of the historical time step of the pipeline and the meteorological information; the gas-liquid external transmission feature extraction module is used to extract the potential correlation characteristics between the gaseous and liquid external transmission volumes through a deep convolutional neural network; the time dependence relationship between the LNG external transmission volume and the meteorological information is extracted through a long short-term memory neural network, and then the potential nonlinear correlation relationship between the gaseous and liquid external transmission volumes and the long-term dependence of the external transmission volume over time are simultaneously explored; the output layer is used to output the prediction result of the LNG external transmission volume.

[0013] Furthermore, the input layer includes a feature adaptive weighting module, which is used to perform weighted transformation on the data input into the input layer, obtain the correlation between the input variables, output the weight of each input variable for the importance of the predicted external output, generate weighted input variables, and input the weighted input variables into the gas-liquid external output feature extraction module.

[0014] Furthermore, the activation function in the short-term hybrid prediction model of external output is a ReLU function, the short-term hybrid prediction model of external output uses an Adam optimizer for model training, a cross-validation method is used to evaluate the model performance, and the model parameters and structure are further optimized and adjusted according to the model validation results.

[0015] The present invention also discloses a short-term prediction system for LNG export volume coupled with gas-liquid export characteristics, including: a data acquisition module for collecting historical data of gaseous export volume and liquid export volume of LNG receiving stations and basic meteorological information; a data reconstruction module for performing optimal variational mode decomposition OVMD on the historical data and basic meteorological information to obtain intrinsic mode components and reconstruct the intrinsic mode components; a model training module for inputting the reconstructed data into a short-term hybrid prediction model for export volume, the short-term hybrid prediction model for export volume includes a deep convolutional neural network module and a long short-term memory neural network model, the deep convolutional neural network model is used to extract potential correlation features between gaseous export volume and liquid export volume; the long short-term memory neural network model is used to extract time-dependent features between gaseous export volume and liquid export volume and the meteorological information; a result output module is used to perform feature fusion of the correlation features with the time-dependent features to obtain the prediction results of LNG export volume. The technical solution of the present invention has at least the following technical effects or advantages: 1. In the prior art, due to the lack of LNG export volume prediction tools, field workers often determine the future LNG export volume based on manual experience and formulate corresponding scheduling plans, which are of poor guidance. The present invention couples the LNG gas-liquid export characteristics, comprehensively considers various factors affecting the LNG export volume, and significantly improves the accuracy of the prediction.

[0016] 2. The present invention gives full play to the advantages of each model by constructing a hybrid model, reduces the highly nonlinear and non-stationary characteristics of the original data, thereby better extracting the potential nonlinear correlation between gas and liquid states, and improving the generalization and adaptability of the model.

[0017] 3. The solution in the present invention improves the stability and reliability of the prediction model through model optimization technology.

[0018] In summary, the method and system for short-term prediction of LNG export volume coupled with gas-liquid export characteristics provided by the present invention can effectively solve the problems existing in the prior art and can be widely used in short-term prediction of LNG export volume. Compared with the prior art, the prediction accuracy of the present invention is higher and the prediction deviation is smaller, which can provide a useful reference for the reasonable allocation and planning of LNG receiving station resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a statistical diagram of gas and liquid external transmission volume of an LNG receiving station in the past year according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the process of constructing an LNG export volume database in one embodiment of the present invention; Figure 3 It is a schematic diagram of the process of constructing a hybrid prediction model for coupling gas-liquid external transmission characteristics provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of a gas-liquid external transmission volume feature extraction module provided in one embodiment of the present invention; Figure 5 This is a comparison chart between the short-term prediction results of LNG export volume according to an embodiment of the present invention and the prediction results of LSTM and CNN-LSTM. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail through specific embodiments. However, it should be understood that the provision of specific embodiments is only for a better understanding of the present invention, and they should not be understood as limitations of the present invention. In the description of the present invention, it should be understood that the terms used are only for the purpose of description and cannot be understood as indicating or implying relative importance.

[0021] LNG export volume is divided into two parts: gaseous export volume and liquid export volume. At present, for the problem of LNG export volume prediction on site, most of them use recurrent neural network (RNN) and long short-term memory neural network (LSTM) models to predict liquid export volume and gaseous export volume separately. Compared with gaseous export volume, LNG liquid export volume has smaller fluctuations and stronger regularity. Therefore, the prediction effect of using traditional RNN or LSTM neural network models is better, and the deviation is within the acceptable range on site. However, for the problem of gaseous export volume prediction, due to its large fluctuations, the prediction accuracy is greatly reduced, the deviation is large, and it cannot provide an effective reference for the reasonable allocation of receiving station resources.

[0022] In order to solve the problem of low accuracy in gaseous export volume prediction and further improve the accuracy of liquid export volume prediction, the present invention proposes a short-term prediction method and system for LNG export volume that couples gas-liquid export features, integrates the coupled correlation features of LNG gaseous and liquid export volumes into the prediction model, and proposes a short-term hybrid prediction model for export volume. The short-term hybrid prediction model for export volume includes a deep convolutional neural network module and a long short-term memory neural network model. The proposed model is compared with the traditional prediction results, and an ablation experiment is performed on the model proposed by the present invention to verify the importance of each module of the proposed model to the prediction results. The scheme of the present invention can greatly improve the accuracy of short-term prediction of LNG export volume, provide effective data support for enterprises to manage LNG resources more efficiently and make operational plans in advance, help operators to manage energy resources more effectively, optimize supply chain operations, reduce operating costs, and ensure the stability of energy supply, and improve overall economic and social benefits. The scheme of the present invention is described in detail below through embodiments in conjunction with the accompanying drawings.

[0023] Embodiment 1 This embodiment discloses a short-term prediction method for LNG export volume coupled with gas-liquid export characteristics, which uses the LNG export volume data of a domestic LNG receiving station in a certain year as the test object and performs case analysis. The statistical data of the gas-liquid export volume of the LNG receiving station in the past year are as follows: Figure 1 As shown, the following steps are included: S1 collects historical data on gaseous and liquid outflows of LNG receiving stations and basic meteorological information.

[0024] like Figure 2 As shown, according to the historical export volume data of the on-site LNG receiving station, the export volume database is pre-built, and the historical data of the gaseous export volume and the liquid export volume and the basic meteorological information are pre-processed. The pre-processing includes: removing outliers and ensuring data quality through data cleaning technology; normalizing the cleaned data to eliminate the dimensional influence between data indicators, accelerate the gradient descent process, and improve the model accuracy. The key features that affect the LNG export volume are extracted from the normalized data. The key features can be the temperature, air pressure and other features in the meteorological conditions. In this embodiment, the highest temperature, the lowest temperature and the average temperature are preferably used as the key features. However, the selection of key features is not limited to this, and any characteristic parameter that plays a key role in the gaseous export volume and the liquid export volume can be used as a key feature.

[0025] S2 performs optimal variational mode decomposition (OVMD) on historical data and basic meteorological information to form intrinsic mode components and reconstruct the intrinsic mode components.

[0026] The method for performing optimal variational mode decomposition (OVMD) is as follows: Variational mode decomposition (VMD) is a signal decomposition algorithm that can effectively process non-stationary and nonlinear signals. By iteratively searching for variational models, complex signals can be decomposed into K It can generate modal components with limited bandwidth, solve the problems of endpoint effect and modal component aliasing in empirical mode decomposition, and is often used in fields such as signal processing and machine learning.

[0027] The core of variational mode decomposition (VMD) is to solve the variational problem and decompose the original signal into K modal components with finite bandwidths and different center frequencies, and at the same time minimize the sum of the estimated bandwidths of each modal component. Therefore, the constraint condition is: the superposition of each modal component is equal to the original signal as the constraint condition: This constraint can be expressed as follows:

[0028]

[0029] in, Indicates Modes; represents the set of center frequencies of all modes; represents the number of modal components; is the Dirac distribution; represents the original signal, t It's time. j Is an imaginary unit.

[0030] In order to solve the constraint problem, the model parameters are introduced: the quadratic penalty factor α and the Lagrange multiplier τ. After the model parameters are introduced, the calculation formula of the constraint problem is:

[0031] Where L is the Lagrangian function, Represents the IMF components of each intrinsic mode function after decomposition; Indicates the center frequency of each component; is the quadratic penalty factor; is the Lagrange multiplier; Indicates that the function is in time The partial derivatives on ; Represents the convolution operator.

[0032] Through Fourier transform, we can get the decomposition of the original sequence of LNG export volume. K narrowband intrinsic mode function components; by updating the model parameters, the decomposed K The narrowband intrinsic mode function components are updated until the optimal intrinsic mode function components that meet the constraints are obtained.

[0033] The alternating direction multiplier method is used to update alternately, transforming the above variational problem into an optional model. and The update formula is as follows:

[0034]

[0035] in: , , , Respectively , , , The Fourier transform of the original sequence of LNG export volume is obtained by Fourier transform. narrowband IMF components.

[0036] In the process of VMD decomposing the original signal of LNG output, the total number of modal decompositions , quadratic penalty factor and the Lagrange multiplier The determination of will directly affect the final effect of decomposition. Therefore, the center frequency method is used to determine the total number of modal decompositions. K , the Lagrange multiplier τ is determined based on the minimization of the residual index. The formula of the residual index is as follows:

[0037] in, Indicates the length of the original signal of LNG export volume. is the residual indicator, Indicates The VMD method after parameter optimization is called optimal variational mode decomposition (OVMD). Using the OVMD method, the original sequence data of LNG export volume can be decomposed into IMF components.

[0038] Since there is a correlation between the LNG gaseous export volume and the liquid export volume, and the export volume is also affected by the meteorological environment, this embodiment uses the liquid export volume, the maximum temperature, the minimum temperature, and the average temperature as characteristic variables of the model for predicting the gaseous export volume; similarly, the gaseous export volume, the maximum temperature, the minimum temperature, and the average temperature are used as characteristic variables of the model for predicting the liquid export volume.

[0039] Taking the reconstruction of the gaseous outflow prediction model data set as an example, the method for reconstructing the intrinsic mode components is as follows: the intrinsic mode components generated by optimal variational mode decomposition (OVMD) and the historical data of gaseous outflow and liquid outflow are used as the first data set, and the intrinsic mode components and basic meteorological information are used as the second data set. The first data set and the second data set are input into the short-term hybrid prediction model of outflow to train the model.

[0040] S3 inputs the reconstructed data into the short-term hybrid prediction model of external output. The short-term hybrid prediction model of external output includes a deep convolutional neural network module and a long short-term memory neural network module. The deep convolutional neural network model is used to extract the correlation characteristics between gaseous external output and liquid external output; the long short-term neural memory network model is used to extract the time dependence characteristics between gaseous external output and liquid external output and meteorological information.

[0041] like Figure 3 As shown, the short-term hybrid forecasting model of external transport volume includes an input layer, a gas-liquid external transport feature extraction module and an output layer; The input layer is used to input the external transmission volume information and meteorological information of the historical time step of the pipeline; the gas-liquid external transmission feature extraction module is set as a multi-deep learning model coupling module, that is, a CNN and LSTM model coupling module, which is used to extract the correlation characteristics between the gas and liquid external transmission volumes through a deep convolutional neural network; the time dependency between the LNG external transmission volume and the meteorological characteristics is extracted through a long short-term memory neural network, and then the potential nonlinear correlation between the gas and liquid external transmission volumes and the long-term dependency of the external transmission volume over time are simultaneously explored; The output layer is used to output the prediction results of LNG export volume.

[0042] The input layer includes a feature adaptive weighting module, which is used to perform weighted transformation on the data input into the input layer, obtain the correlation between the input variables, output the weight of each input variable for the importance of the predicted external output, generate weighted input variables, and input the weighted input variables into the gas-liquid external output feature extraction module.

[0043] Specifically, the convolutional neural network (CNN) is a feedforward neural network that contains convolution calculations and has a deep structure. It is one of the representative algorithms in the field of deep learning. Its basic structure consists of an input layer, a convolutional layer, a pooling layer (also called a sampling layer), a fully connected layer, and an output layer. Generally, several convolutional layers and pooling layers are selected and set alternately, that is, a pooling layer is connected after a convolutional layer, and a convolutional layer is connected after the pooling layer, and so on. In the forward transmission process of information, each kernel in the convolutional layer is convolved on the input vector of the convolutional layer to form a feature vector. The convolution of CNN is as follows:

[0044] in, express Tier The output of each channel; represents the channel of the convolutional neural network, That is Tier channels; is the weight matrix; Represents the bias matrix. Similar to general neural networks, and It will be continuously iterated, optimized and updated with back propagation.

[0045] Through the CNN model, the potential correlation features between the LNG gaseous and liquid external transmission volumes can be extracted so as to be integrated with the features extracted by the next model.

[0046] The LSTM neural network model consists of an input layer, a hidden layer, and an output layer. The hidden layer has a gate structure and a memory unit. The LSTM unit can be )、Input Gate( ) and the output gate ( ) selectively add or forget information. The Forget Gate determines which past information is no longer important for predicting future output and can be "forgotten"; the Input Gate determines which information input at the current time point should be "remembered"; and the Output Gate determines which information should be used to generate the final prediction output.

[0047] The mathematical description of the moment memory unit update mechanism and the gate structure control mechanism is as follows:

[0048]

[0049]

[0050]

[0051]

[0052] in, is the input gate; It is the Gate of Oblivion; yes The memory unit of the moment; is the output gate; is the input vector; is the output vector; is the weight vector; is the bias vector; is the sigma activation function.

[0053] like Figure 4 As shown in the figure, the correlation features of LNG gas and liquid export volume and the time dependency between LNG export volume and meteorological conditions are obtained through CNN and LSTM respectively. Next, the two types of features extracted are spliced ​​and passed through the fully connected layer to achieve feature fusion and interaction, so that the network can learn the correlation between different features and improve the network's ability to represent data. Subsequently, the data after feature extraction and conversion is mapped to the final output space to achieve short-term prediction of LNG export volume.

[0054] The method for training the short-term hybrid prediction model for external output is to construct an initial short-term hybrid prediction model for external output through deep learning algorithms, namely CNN and LSTM, input the first data set and the second data set into the initial short-term hybrid prediction model for external output, train and verify the model, and obtain a trained short-term hybrid prediction model for external output.

[0055] The activation function in the short-term hybrid prediction model of external output is the ReLU function. The short-term hybrid prediction model of external output uses the Adam optimizer for model training, and the cross-validation method is used to evaluate the model performance to prevent overfitting and improve the generalization ability of the model. The model parameters and structure are further optimized and adjusted according to the model verification results to optimize the prediction effect.

[0056] S4 fuses the correlation features with the time-dependent features to obtain the prediction results of the LNG export volume.

[0057] The actual LNG export volume data for a period of time before the forecast date is obtained, and the local temperature and other meteorological conditions during this period are obtained. According to the above data processing and reconstruction process, the original data is processed, and the processed data is input into the short-term hybrid forecasting model of export volume to obtain the forecast result of LNG export volume. The forecast result is as follows: Figure 5 shown.

[0058] like Figure 5 As shown in the figure, according to the line graph and scatter plot results, in the gaseous outflow prediction scenario, when the outflow is larger than 10,000 m³ / d, the prediction effect of the model is reduced, which can be easily observed in the results of the scatter plot. From the comparison results, it can be seen that the model proposed in the present invention has the best prediction effect, and the predicted value is more consistent with the true value. The prediction effect of the CNN-LSTM model is second, and the prediction effect of the LSTM model is the worst.

[0059] In order to further quantitatively illustrate the superiority of the prediction performance of the present invention, the coefficient of determination (R 2 ), root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) as measurement indicators, and compared with the existing LNG export volume prediction technology. The comparison results are shown in the following table: Table 1 Comparison of prediction results with existing technologies

[0060] It can be seen that compared with traditional time series prediction methods, such as LSTM, CNN-LSTM and other methods, the prediction model proposed in the present invention has achieved the best prediction performance. Compared with a single LSTM model, the prediction effect of this technology has been greatly improved. And according to the prediction results, the prediction performance of the CNN-LSTM model is better than that of the LSTM model, which verifies that the introduction of convolutional neural networks to extract features between gas and liquid transport volumes helps improve prediction accuracy. However, the prediction performance of CNN-LSTM is worse than that of the model proposed in the present invention, which further illustrates that the introduction of optimal variational mode decomposition (OVMD) can improve the prediction accuracy of the model.

[0061] Embodiment 2 Based on the same inventive concept, this embodiment discloses a short-term prediction system for LNG export volume coupled with gas-liquid export characteristics, including: Data collection module, used to collect historical data of gaseous and liquid outflows of LNG receiving stations and basic meteorological information; Data reconstruction module, used to perform optimal variational mode decomposition (OVMD) on historical data and basic meteorological information to form intrinsic mode components and reconstruct the intrinsic mode components; The model training module is used to input the reconstructed data into the short-term hybrid forecasting model of the external output. The short-term hybrid forecasting model of the external output includes a deep convolutional neural network module and a long short-term memory neural network model. The deep convolutional neural network model is used to extract the correlation characteristics between the gas external output and the liquid external output; the long short-term memory neural network model is used to extract the time dependence characteristics of the gas external output and the liquid external output and meteorological information; The result output module is used to fuse the correlation features with the time-dependent features to obtain the prediction results of the LNG export volume.

[0062] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0063] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0064] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not deviate from the spirit and scope of the present invention should be included in the protection scope of the claims of the present invention. The above content is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A short-term prediction method for LNG export volume coupled with gas-liquid export characteristics, characterized in that: The following steps are involved: Collect historical data on gaseous and liquid outflows of LNG receiving stations and basic meteorological information; Performing optimal variational mode decomposition (OVMD) on the historical data and basic meteorological information to obtain intrinsic mode components, and reconstructing the intrinsic mode components; The reconstructed data is input into a short-term hybrid prediction model for external output, wherein the short-term hybrid prediction model for external output includes a deep convolutional neural network module and a long short-term memory neural network model, and the potential correlation characteristics between the gas external output and the liquid external output are extracted through the deep convolutional neural network model; and the time dependence characteristics between the gas external output and the liquid external output and the meteorological information are extracted through the long short-term memory neural network model; The correlation feature is fused with the time-dependent feature to obtain a prediction result of the LNG export volume.

2. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 1, characterized in that: The historical data of gas export volume and liquid export volume and basic meteorological information are preprocessed, and the preprocessing includes: removing outliers and noise through data cleaning technology; normalizing the cleaned data, and extracting key features that affect the LNG export volume from the normalized data.

3. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 2, characterized in that: The key characteristics include maximum temperature, minimum temperature and average temperature.

4. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 1, characterized in that: The method for performing optimal variational mode decomposition OVMD is: The superposition of each modal component is equal to the original signal as a constraint condition, which leads to the model parameters: the total number of modal decompositions K , quadratic penalty factor α and Lagrange multiplier τ; Through Fourier transform, we can get the decomposition of the original sequence of LNG export volume. K Narrowband intrinsic mode function components; By updating the model parameters, the decomposed K The narrowband intrinsic mode function components are updated until the optimal intrinsic mode function components that meet the constraints are obtained.

5. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 4, characterized in that: The total number of modal decompositions is determined using the center frequency method K , the Lagrange multiplier τ is determined based on the minimization of the residual index. The formula of the residual index is as follows: in, Indicates the length of the original signal of LNG export volume. is the residual indicator, Indicates kind of modality.

6. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 1, characterized in that: The intrinsic mode components generated by the optimal variational mode decomposition OVMD and the historical data of gaseous and liquid external transport are used as the first data set, and the intrinsic mode components and basic meteorological information are used as the second data set. The first data set and the second data set are input together into a short-term hybrid prediction model for external transport to train the model.

7. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 1, characterized in that: The short-term hybrid prediction model for external transmission volume includes an input layer, a gas-liquid external transmission feature extraction module and an output layer; the input layer is used to input the external transmission volume information of the historical time step of the pipeline and the meteorological information; the gas-liquid external transmission feature extraction module is used to extract the potential correlation characteristics between the gaseous and liquid external transmission volumes through a deep convolutional neural network; the time dependence relationship between the LNG external transmission volume and the meteorological information is extracted through a long short-term memory neural network, and then the potential nonlinear correlation relationship between the gaseous and liquid external transmission volumes and the long-term dependence of the external transmission volume over time are simultaneously mined; the output layer is used to output the prediction result of the LNG external transmission volume.

8. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 7, characterized in that: The input layer includes a feature adaptive weighting module, which is used to perform weighted transformation on the data input into the input layer, obtain the correlation between the input variables, output the weight of each input variable for the importance of the predicted external output, generate weighted input variables, and input the weighted input variables into the gas-liquid external output feature extraction module.

9. The method for short-term prediction of LNG export volume coupled with gas-liquid export characteristics according to claim 8, characterized in that: The activation function in the short-term hybrid prediction model for external output is the ReLU function. The short-term hybrid prediction model for external output uses the Adam optimizer for model training, uses a cross-validation method to evaluate the model performance, and further optimizes and adjusts the model parameters and structure based on the model validation results.

10. A short-term prediction system for LNG export volume coupled with gas-liquid export characteristics, characterized in that: include: Data collection module, used to collect historical data of gaseous and liquid outflows of LNG receiving stations and basic meteorological information; A data reconstruction module, used for performing optimal variational mode decomposition (OVMD) on the historical data and basic meteorological information to obtain intrinsic mode components, and reconstructing the intrinsic mode components; A model training module is used to input the reconstructed data into a short-term hybrid prediction model for external output, wherein the short-term hybrid prediction model for external output includes a deep convolutional neural network module and a long short-term memory neural network model, and the potential correlation characteristics between the gas external output and the liquid external output are extracted through the deep convolutional neural network model; and the time dependence characteristics between the gas external output and the liquid external output and the meteorological information are extracted through the long short-term memory neural network model; The result output module is used to fuse the associated features with the time-dependent features to obtain the prediction result of the LNG export volume.

Images