Initialization method, device and equipment for input parameters of gas load prediction model

Abstract

The invention discloses a gas consumption load prediction model input parameter initialization method, device and equipment, and the method comprises the steps: obtaining the historical gas consumption data of a natural gas customer in a preset calculation time range, and carrying out the normalization processing of the obtained historical gas consumption data, and obtaining the normalized data; based on the industry to which the natural gas customers belong, a clustering algorithm, a gas consumption data volume scale and normalized data, performing clustering hierarchy division on the natural gas customers; aiming at each cluster of clustering hierarchical division, training the prediction model according to the normalized data of the natural gas customers of the cluster to obtain model input parameters of the prediction model aiming at the cluster; and obtaining model input parameters corresponding to the target cluster to which the to-be-predicted natural gas customer belongs, and initializing the prediction model by using the obtained model input parameters so as to predict the normalized to-be-analyzed data of the to-be-predicted natural gas customer, thereby obtaining the gas consumption load prediction data of the to-be-predicted natural gas customer, and improving the precision of the gas consumption load prediction data.

Description

Technical Field

[0001] The present invention relates to the technical field of natural gas new energy, and particularly relates to a method, device and equipment for initializing input parameters of a gas load prediction model. Background Art

[0002] With the increasingly deteriorating global environment, it has become extremely urgent to use clean energy to replace heavily polluting energy such as coal and oil. As a clean energy, natural gas has attracted more and more attention. For the demand side, using a prediction model to accurately predict gas load is beneficial to ensuring gas supply demand and is of great significance for ensuring the safe operation of the pipeline network and optimizing the dispatching of the pipeline network. Among them, the prediction accuracy of gas load is affected not only by the quality of gas consumption data of natural gas customers, the selected prediction model and influencing factors, but also by the input parameters of the prediction model during the prediction process. Among them, the model input parameters can be the composition ratio of the training set and the validation set, the maximum depth of the tree required by the prediction model, the learning rate, the sampling ratio, the random seed, etc.

[0003] Currently, when natural gas enterprises are predicting gas load, they generally initialize the prediction model based on the historical gas consumption data and gas consumption influencing factor data of natural gas customers collected, according to the model input parameters, to form a gas load prediction model, and input the collected data into the gas load prediction model for prediction. However, this method uses the same gas load prediction model for gas load prediction for different natural gas customers, that is, the gas load prediction model uses the same model input parameters for all natural gas customers, without considering the differences in the historical gas consumption data volume and gas consumption patterns of natural gas customers, resulting in low prediction accuracy of gas load for natural gas customers. Summary of the Invention

[0004] In order to improve the prediction accuracy of gas load for natural gas customers and enrich the technical route and increase the selection space, the embodiments of the present invention provide a method, device and equipment for initializing input parameters of a gas load prediction model.

[0005] In a first aspect, an embodiment of the present invention provides a method for initializing input parameters of a gas load prediction model, which may include:

[0006] Obtain the historical gas consumption data of natural gas customers within a pre-designed calculation time range, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data;

[0007] Based on the industry to which the natural gas customers belong, a pre-set clustering algorithm, the scale of user gas consumption data volume and the normalized data, perform hierarchical clustering on the natural gas customers;

[0008] For each cluster for which hierarchical clustering is performed, the prediction model is trained based on the normalized data of the natural gas customers in the cluster to obtain the model input parameters of the prediction model for the cluster.

[0009] Obtain the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, and initialize the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the predicted gas load data of the natural gas customer to be predicted.

[0010] Optionally, the method further includes:

[0011] For each cluster, a model input parameter sequence set is generated based on the model input parameters of the prediction model for the cluster.

[0012] Based on the model input parameter sequence sets corresponding to each cluster, a model input parameter library is constructed.

[0013] Optionally, the step of obtaining the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, initializing the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the predicted gas load data of the natural gas customer to be predicted may include:

[0014] Obtain the normalized data to be analyzed of the natural gas customer to be predicted, and determine the candidate clusters for hierarchical clustering to which the natural gas customer to be predicted belongs.

[0015] From the model input parameter library, obtain the optimal input parameter candidate sequence set corresponding to the candidate cluster.

[0016] Based on the optimal input parameter candidate sequence set, assign values to the prediction model to obtain a gas load prediction model.

[0017] Input the normalized data to be analyzed into the gas load prediction model to obtain the predicted gas load data of the natural gas customer to be predicted.

[0018] Optionally, the method further includes:

[0019] From the model input parameter library, obtain other input parameter candidate sequence sets corresponding to the candidate cluster.

[0020] For each group of other input parameter candidate sequence sets, assign values to the prediction model respectively, and input the normalized data to be analyzed into the assigned prediction model to obtain the sub-optimal predicted gas load data of the natural gas customer to be predicted.

[0021] Evaluate the gas load prediction data and the sub-optimal gas load prediction data according to the actual gas load data corresponding to the gas load prediction data of the natural gas customer to be predicted;

[0022] If the evaluation accuracy of the sub-optimal gas load prediction data is greater than that of the gas load prediction data, replace the optimal input parameter candidate sequence set with the input parameter candidate sequence set corresponding to the sub-optimal gas load prediction data.

[0023] Optionally, the normalization process for the obtained historical gas data includes:

[0024] Select the historical gas data with the largest value and the historical gas data with the smallest value from the obtained historical gas data;

[0025] Calculate the difference between the historical gas data with the largest value and the historical gas data with the smallest value to obtain the normalization difference;

[0026] For each historical gas data, calculate the difference between the historical gas data and the historical gas data with the smallest value, and calculate the quotient of the difference and the normalization difference to obtain the normalized data of the historical gas data.

[0027] Optionally, the hierarchical clustering of natural gas customers based on the industry to which the natural gas customers belong, a pre-set clustering algorithm, the scale of user gas data volume, and the normalized data includes:

[0028] For each normalized data, obtain the industry to which the natural gas customer corresponding to the normalized data belongs, query the classification result set. If the industry is in the classification result set, place the normalized data in the industry classification corresponding to the classification result set. If the industry is not in the classification result set, create a new industry classification in the classification result set and place the normalized data in the newly created industry classification;

[0029] For each industry classification in the classification result set, use a pre-set clustering algorithm to cluster the natural gas customers under the industry classification to obtain a clustering cluster set containing multiple clustering clusters for the industry classification;

[0030] For each clustering cluster in the clustering cluster set, according to a pre-set normalization classification threshold, perform threshold subdivision on the normalized data included in the clustering cluster to obtain multiple clustering cluster sub-classifications included in the clustering cluster.

[0031] Optionally, training the prediction model according to the normalized data of the natural gas customers in the cluster to obtain the model input parameters of the prediction model for the cluster includes:

[0032] Determine the prediction model of natural gas customers in the target cluster from the clusters obtained by hierarchical clustering;

[0033] Randomly assign values to the model input parameters of the determined prediction model to obtain multiple groups of different initial model input parameters;

[0034] For each group of initial model input parameters, respectively train the prediction model according to the normalized data included in the target cluster to adjust the initial model input parameters until the prediction model converges, and obtain each group of model input parameters corresponding to each group of initial model input parameters;

[0035] For each group of model input parameters, respectively assign values to the prediction model, and use the assigned prediction model to make predictions according to the normalized data included in the target cluster to obtain the prediction results corresponding to each group of model input parameters, and use the pre-set error algorithm to evaluate each prediction result respectively;

[0036] Based on the evaluation results, obtain the model input parameters of the prediction model for the target cluster.

[0037] In a second aspect, an embodiment of the present invention provides a device for initializing input parameters of a gas load prediction model, which may include:

[0038] A normalization module, configured to obtain historical gas consumption data of natural gas customers within a pre-designed calculation time range, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data;

[0039] A clustering and grading module, configured to perform hierarchical clustering on natural gas customers based on the industries to which the natural gas customers belong, a pre-set clustering algorithm, the scale of user gas consumption data volume, and the normalized data;

[0040] A model training module, configured to train a prediction model for each cluster obtained by hierarchical clustering according to the normalized data of the natural gas customers in the cluster, and obtain the model input parameters of the prediction model for the cluster;

[0041] A data prediction module, configured to obtain the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, initialize the prediction model by using the obtained model input parameters, and predict the normalized data to be analyzed of the natural gas customer to be predicted to obtain the gas load prediction data of the natural gas customer to be predicted.

[0042] In a third aspect, an embodiment of the present invention provides a storage medium, on which a program or instruction is stored, and when the program or instruction is run by a processor, it implements the method for initializing input parameters of a gas load prediction model as described in the first aspect.

[0043] Fourthly, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method for initializing input parameters of the gas load prediction model as described in the first aspect is implemented.

[0044] The beneficial effects of the above technical solutions provided by the embodiments of the present invention at least include:

[0045] An embodiment of the present invention provides a method, device, and equipment for initializing input parameters of a gas load prediction model. The method obtains the normalized data of natural gas customers, and based on the industry to which the natural gas customers belong, a pre-set clustering algorithm, the scale of user gas consumption data volume, and the normalized data, conducts a hierarchical clustering of natural gas customers. For each cluster obtained by the hierarchical clustering, the prediction model is trained according to the normalized data of the natural gas customers in this cluster to obtain the model input parameters of the prediction model for this cluster, so as to predict the gas load of the natural gas customers in this cluster. In this way, by training the prediction model for the natural gas customers in each cluster, obtaining the model input parameters, and obtaining a gas load prediction model for predicting the gas load of the natural gas customers classified at this level, the prediction accuracy of the gas load can be effectively improved.

[0046] Other features and advantages of the present invention will be described in the following specification, and some of them will become obvious from the specification, or be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in the written specification and the drawings.

[0047] The technical solutions of the present invention will be further described in detail below with reference to the drawings and embodiments. Description of the Drawings

[0048] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0049] Figure 1 is a flowchart of the method for initializing input parameters of the gas load prediction model provided by the embodiment of the present invention;

[0050] Figure 2 is a specific flowchart of step S102;

[0051] Figure 3 is a specific flowchart of step S103;

[0052] Figure 4 is a schematic structural diagram of the device for initializing input parameters of the gas load prediction model provided by the embodiment of the present invention;

[0053] Figure 5 This is a schematic structural diagram of the electronic device provided in the embodiment of the present invention. Detailed implementation manners

[0054] Hereinafter, exemplary embodiments of the present disclosure will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0055] The inventors found that in the prior art, by using the historical gas consumption data of each natural gas customer, training a prediction model, obtaining model input parameters, and when performing gas load prediction, using the model input parameters to initialize the gas load prediction model to predict the gas load of natural gas customers. However, for different natural gas customers, the gas consumption data and gas consumption patterns vary greatly. For example, for catering natural gas customers and domestic natural gas customers, the amount of gas consumption data and gas consumption patterns are quite different. Therefore, using the same gas load prediction model for gas load prediction results in a low prediction accuracy for the gas loads of different natural gas customers.

[0056] In the embodiment of the present invention, by considering the influence of the historical gas consumption data volume and the difference in gas consumption patterns of natural gas customers on the model input parameters, multi-dimensional classification of natural gas customers is performed, and a prediction model is trained based on the historical gas consumption data of the classified natural gas customers to obtain the model input parameters for this type of natural gas customer. Specifically, in this embodiment, considering the historical gas consumption data volume and the difference in gas consumption patterns of natural gas customers, classifying natural gas customers according to the industry to which the natural gas customers belong, gas consumption patterns, historical gas consumption data volume, etc., training a prediction model for each type of natural gas customer, and obtaining model input parameters can improve the prediction accuracy of the prediction model, thereby achieving the purpose of quickly and accurately predicting the gas load demand of natural gas customers.

[0057] Referring to Figure 1 As shown, an initialization method for model input parameters of a gas load prediction model is provided in the embodiment of the present invention, and the method may include the following steps:

[0058] Step S101: Obtain the historical gas consumption data of natural gas customers within a pre-designed calculation time range, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data.

[0059] In this step, by obtaining the historical gas consumption data of natural gas customers and performing normalization processing, the influence of different dimensions and units of historical gas consumption data is eliminated, and data preparation is performed for subsequent clustering analysis of natural gas customers.

[0060] In an embodiment of the present invention, after obtaining the historical gas consumption data of natural gas customers within a preset period and before normalizing the obtained historical gas consumption data, the method may further clean and process the data, specifically including: collecting and processing the historical gas consumption data of each natural gas customer into monthly gas consumption data, and the above-mentioned preset calculation time range is in years; cleaning and processing the monthly gas consumption data.

[0061] It should be noted that in this embodiment, cleaning and processing the data is a data preprocessing process. The above-mentioned monthly gas consumption data is the data obtained by accumulating the historical gas consumption data in different dimensions such as hourly, daily, and weekly. For example, the daily historical gas consumption data of each natural gas user to be analyzed within the calculation time range is obtained, and the daily historical gas consumption data of each natural gas customer is accumulated to obtain the monthly gas consumption data. As an optional embodiment, the monthly gas consumption data is stored in a database.

[0062] In this step, by cleaning and processing the monthly gas consumption data, high-quality historical gas consumption data of natural gas customers can be obtained. In specific implementation, when cleaning the monthly gas consumption data, it can be based on the principle of whether there is continuous data within the calculation time range to clean the monthly gas consumption data of natural gas customers. As an optional embodiment, cleaning the monthly gas consumption data may include: for each natural gas customer, determining whether the monthly gas consumption data of the natural gas customer is continuous. If it is not continuous, the monthly gas consumption data of the natural gas customer is excluded.

[0063] In this embodiment, natural gas customers with discontinuous data are screened out, and the historical gas consumption data of the natural gas customers is excluded. As an optional embodiment, for natural gas customers with a data value of 0 but continuous data, the corresponding historical gas consumption data is not excluded.

[0064] In this embodiment, processing the monthly gas consumption data may include: for the monthly gas consumption data after cleaning, processing the missing values, abnormal values, and duplicate values existing in the monthly gas consumption data according to a preset processing strategy to obtain monthly gas consumption data that meets the quality requirements.

[0065] In this embodiment, the monthly gas consumption data of the natural gas customers obtained after cleaning is processed. That is, for the remaining natural gas customers obtained after cleaning, if there are low-quality data such as missing values, abnormal values, and duplicate values in the corresponding monthly gas consumption data or historical gas consumption data, they are processed according to a preset processing strategy to obtain historical gas consumption data that meets the quality requirements.

[0066] In this embodiment, as an optional embodiment, the monthly gas consumption data of natural gas customers is represented by the following formula:

[0067] x = {x1, x2,......., x n}

[0068] where x is the monthly gas consumption dataset of natural gas customers within the pre-designed budget time range, and x n is the monthly gas consumption data of the natural gas customer in the nth month, and n is the number of months included in the pre-designed budget time range. For example, if the pre-designed budget time range is 5 years, then n is 60.

[0069] In this embodiment, as an alternative embodiment, the normalization process of the obtained historical gas consumption data includes: First, select the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value from the obtained historical gas consumption data; Then, calculate the difference between the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value to obtain the normalization difference; Finally, for each historical gas consumption data, calculate the difference between the historical gas consumption data and the historical gas consumption data with the smallest value, and calculate the quotient of this difference and the normalization difference to obtain the normalized data of this historical gas consumption data.

[0070] In this embodiment, as an alternative embodiment, the minimum-maximum normalization (Min-Max Scaling) method is used to linearly map the historical gas consumption data to the range of [0, 1]. In this way, through data normalization, the dimensional and unit differences between different indicators can be eliminated, making different indicators comparable, thereby improving the accuracy of data analysis, the convergence and stability of the prediction model, reducing the deviation of feature weights, improving the interpretability of the prediction model, and better supporting decision-making and predictive analysis.

[0071] In this embodiment, as an alternative embodiment, taking the historical gas consumption data as the monthly gas consumption data as an example, the monthly gas consumption data is normalized using the following formula:

[0072] X i = (x i - min(∑x n )) / (max(∑x n ) - min(∑x n ))

[0073] where X i is the monthly normalized gas consumption corresponding to the monthly gas consumption data of the natural gas customer in the ith month, x i is the monthly gas consumption data of the natural gas customer in the ith month, max(∑x n ) is the maximum monthly gas consumption data among all monthly gas consumption data of each natural gas customer within the pre-designed budget time range, and min(∑x n ) is the minimum monthly gas consumption data among all monthly gas consumption data of each natural gas customer within the pre-designed budget time range.

[0074] Step S102: Based on the industry to which the natural gas customers belong, a preset clustering algorithm, the scale of the user gas consumption data volume, and the normalized data, perform a hierarchical classification of the natural gas customers.

[0075] This step classifies the natural gas customers. For example, first classify according to the industry where the natural gas customers are located, then use the clustering algorithm to cluster the industry classification results, and finally, according to the scale of the gas consumption data volume, subdivide the clustering results to finally obtain a multi-level clustering result of the natural gas customers.

[0076] Figure 2 It is a schematic flowchart of step S102 in a method for initializing model input parameters provided by an embodiment of the present invention. Refer to Figure 2 As shown, as an optional embodiment, based on the industry to which the natural gas customers belong, a preset clustering algorithm, and the normalized data, performing a hierarchical classification of the natural gas customers may include the following steps:

[0077] Step S201: For each normalized data, obtain the industry to which the natural gas customer corresponding to the normalized data belongs, query the classification result set. If the industry is in the classification result set, place the normalized data in the industry classification corresponding to the classification result set. If the industry is not in the classification result set, create a new industry classification in the classification result set and place the normalized data in the newly created industry classification.

[0078] In this embodiment, considering that for natural gas customers within the same industry, there are great similarities in gas consumption patterns. Therefore, when analyzing the gas consumption data of natural gas customers, it can be analyzed separately according to different industries, which can improve the accuracy of the model input parameters obtained by training the prediction model. That is, classify the natural gas customers according to the industry where the natural gas customers are located to obtain a classification result set D1.

[0079] In this embodiment, as an optional embodiment, the industry to which the natural gas customer belongs can be determined by querying the occupation of the natural gas customer and the attributes of the natural gas customer, such as whether it is a resident or a company, enterprise, etc. information. By pre-determining the industry, that is, the classification result set, the natural gas customer is attributed to an industry classification in the classification result set.

[0080] Step S202: For each industry classification in the classification result set, use the preset clustering algorithm to cluster the natural gas customers under the industry classification to obtain a cluster set containing multiple clusters for the industry classification.

[0081] In this embodiment, as an alternative embodiment, based on the industry classification results, a machine learning clustering algorithm is used to cluster natural gas customers in the classification result set. Among them, the clustering algorithm includes, but is not limited to, one or any combination of K-means clustering algorithm, hierarchical clustering algorithm, density clustering, and mean shift clustering algorithm, etc. Using the clustering algorithm, clustering is performed by industry classification to obtain a cluster set D2 corresponding to the industry classification and containing multiple clusters.

[0082] In this embodiment, taking the K-means clustering algorithm as an example, the K-means clustering algorithm is a distance-based clustering algorithm that divides the objects (natural gas customers) in the industry classification into K clusters. Through an iterative method, each object included in the industry classification is assigned to the nearest cluster, and the center point of the cluster is updated until the convergence condition is reached to obtain the final cluster set D2. That is, according to the clustering algorithm, for each industry classification in the classification result set, the historical gas consumption data of each natural gas customer in the industry classification is substituted into the clustering algorithm for calculation to obtain the cluster set D2 of the industry classification. For example, the clusters include the consumption characteristics and user characteristics of natural gas customers. Among them, the consumption characteristics of cluster 1 include, but are not limited to: relatively stable total consumption, higher gas consumption in winter than in summer, significant impact of the Spring Festival holiday; the user characteristics include, but are not limited to: the market in the operation area is relatively mature, the proportion of industrial and public welfare commercial users' demand is high, and industrial users mainly use industrial boilers for gas; the consumption characteristics of the cluster set D2 include: the total consumption shows a downward trend, there is no obvious seasonal fluctuation law, and the monthly consumption fluctuates violently; the user characteristics include: obvious dual-gas source user characteristics; the consumption characteristics of the cluster set D3 include: the consumption shows a steady growth trend, the "high in winter and low in summer" characteristic is significant, and the peak-valley difference shrinks year by year; the user characteristics include: the scale of end-users increases year by year, and the new users are mainly industrial and other non-heating coal-to-gas demands, and the proportion of heating gas consumption decreases.

[0083] Step S203: For each cluster in the cluster set, according to the preset normalization classification threshold, the normalized data included in the cluster is threshold-subdivided to obtain multiple cluster sub-classifications included in the cluster.

[0084] In this step, the scale of the historical gas consumption data of natural gas customers will affect the accuracy of the prediction model results. Therefore, the model input parameters are distinguished according to the scale of the historical gas consumption data volume. As an alternative embodiment, the normalization classification threshold is the number threshold, and one or more normalization classification thresholds can be set according to historical experience or actual requirements. The number of historical gas consumption data of natural gas customers included in each cluster is compared with the number threshold, so as to classify natural gas customers into multiple categories, and a clustering sub-result set D3 including multiple clustering sub-categories of this cluster is obtained. As an alternative embodiment, according to historical experience, the number threshold n is set to the number of data for 5 years. For example, if n = 60 is set, then when performing clustering sub-classification, the number of historical gas consumption data of natural gas customers for 5 years is compared with 60.

[0085] In this embodiment, taking the normalization classification threshold as one example, for each cluster, according to the natural gas users in this cluster, the number of historical gas consumption data of this natural gas user is obtained and compared with the number threshold. If it is greater than or equal to the number threshold, it is placed in the first sub-classification of the cluster; if it is less than the number threshold, it is placed in the second sub-classification of the cluster.

[0086] In this embodiment, as another alternative embodiment, multiple normalization classification thresholds can also be set. For example, the normalization classification thresholds include the first number threshold and the second number threshold. The first number threshold n1 = 40 and the second number threshold n2 = 70 are set. Thus, for each cluster, natural gas customers can be divided into three clustering sub-categories: the number of historical gas consumption data less than 40 is the first sub-classification of the cluster; greater than or equal to 40 and less than or equal to 70 is the second sub-classification of the cluster; greater than 70 is the third sub-classification of the cluster. This embodiment does not make any limitations in this regard.

[0087] Step S103: For each cluster subjected to clustering hierarchy division, according to the normalized data of the natural gas customers in this cluster, train the prediction model to obtain the model input parameters of the prediction model for this cluster.

[0088] In this embodiment, each cluster subjected to clustering hierarchy division corresponds to a clustering sub-classification. For each clustering sub-classification, the normalized data of each natural gas customer included in this clustering sub-classification is input into the prediction model, so as to train the model input parameters of the prediction model according to the historical gas consumption data of the natural gas customers included in this clustering sub-classification, and finally obtain multiple model input parameters. As an alternative embodiment, multiple model input parameters corresponding to one cluster form a model input parameter sequence.

[0089] In this embodiment, based on the clustering sub-result set D3 of natural gas customers, the optimal model input parameter sequences of the natural gas customers included in each clustering sub-classification are calculated respectively to form a model input parameter library.

[0090] Figure 3 This is a schematic flowchart of step S103 in a method for initializing model input parameters provided by an embodiment of the present invention. Refer to Figure 3 As shown, in this embodiment, as an optional embodiment, based on the normalized data of the clustered natural gas customers, training the prediction model to obtain the model input parameters of the prediction model for this cluster may include the following steps:

[0091] Step S301, determine the prediction model of the natural gas customers in the target cluster from the clusters obtained by hierarchical clustering.

[0092] In this embodiment, as an optional embodiment, each cluster may share the same prediction model, or each cluster may correspond to one or more prediction models. Among them, the prediction models include, but are not limited to: ridge regression model, Lasso model, random forest model, Xgboost model, neural network model, etc.

[0093] Step S302, randomly assign values to the model input parameters of the determined prediction model to obtain multiple groups of different initial model input parameters.

[0094] In this embodiment, as an optional embodiment, according to historical experience, initialize the model input parameters of the prediction model to obtain multiple groups of different initial model input parameters. Among them, the initial model input parameters include, but are not limited to: the composition ratio of the training set and the validation set, the maximum depth of the tree required by the prediction model, the learning rate, the sampling ratio, the random seed, etc. Different prediction models have different numbers of initial model input parameters included. By assigning different initial values to the model input parameters, multiple groups of different initial model input parameters of the same prediction model are obtained, and each prediction model corresponds to multiple groups of initial model input parameters.

[0095] Step S303, for each group of initial model input parameters, respectively train the prediction model based on the normalized data included in the target cluster to adjust the initial model input parameters until the prediction model converges, and obtain each group of model input parameters corresponding to each group of initial model input parameters.

[0096] In this embodiment, taking a set of initial model input parameters as an example, the prediction model is assigned according to the set of initial model input parameters. Among the normalized data included in the target cluster, the normalized data before a certain historical period is used as the input of the assigned prediction model to predict this historical period, and normalized prediction data is obtained. Based on the normalized data and the normalized prediction data of this historical period, the initial model input parameters of the prediction model are adjusted using the backpropagation algorithm until the prediction model converges. The final set of adjusted initial model input parameters is a corresponding set of model input parameters. For other sets of initial model input parameters, they are similar to this set of initial model input parameters. Therefore, for the target cluster, for each prediction model, multiple sets of model input parameters can be obtained.

[0097] Step S304: For each set of model input parameters, the prediction model is assigned respectively. According to the normalized data included in the target cluster, the assigned prediction model is used for prediction to obtain the prediction results corresponding to each set of model input parameters, and each prediction result is evaluated using a pre-set error algorithm.

[0098] In this embodiment, for each prediction model, each set of model input parameters is substituted into the prediction model for prediction to obtain the prediction results. As an alternative embodiment, the error algorithms include but are not limited to: coefficient of determination (R2, R-Square) algorithm, mean square error (MSE, Mean Square Error) algorithm, mean absolute percentage error (MAPE, Mean Absolute Percentage Error) algorithm, root mean square error (RMSE, Root Mean Square Error) algorithm, mean absolute error (MAE, Mean Absolute Error) algorithm, etc. The error algorithms are used to evaluate each prediction result. Taking the mean square error algorithm as an example, by calculating the mean square error of the prediction result, if the calculated mean square error is smaller, it indicates that the prediction result is better, and the model input parameters corresponding to this prediction result are more optimal.

[0099] Step S305: Based on the evaluation results, the model input parameters of the prediction model for the target cluster are obtained.

[0100] In this embodiment, as an alternative embodiment, the model input parameters corresponding to the optimal evaluation result are selected from the evaluation results as the model input parameters of the prediction model for the target cluster. As another alternative embodiment, the model input parameters corresponding to the top n prediction results can also be selected to obtain a sequence set of model input parameters of the prediction model for the target cluster. For example, from the prediction results for evaluation, 5 sets of model input parameters with the best effects are selected as the final sequence set of model input parameters of the prediction model corresponding to the natural gas customers under this target cluster.

[0101] Step S104: Obtain the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, and initialize the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the predicted gas load data of the natural gas customer to be predicted.

[0102] In this embodiment, after forming the model input parameter library, the model input parameter library can be used to predict the gas load of natural gas customers, that is, by classifying the natural gas customers, obtaining the model input parameters of the classified natural gas customers from the model input parameter library, and substituting them into the prediction model for prediction.

[0103] In this embodiment, as an optional embodiment, obtaining the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, and initializing the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the predicted gas load data of the natural gas customer to be predicted, includes:

[0104] A11. Obtain the normalized data to be analyzed of the natural gas customer to be predicted, and determine the candidate clusters for hierarchical clustering division to which the natural gas customer to be predicted belongs.

[0105] In this embodiment, when it is necessary to predict the gas load of a natural gas customer, obtain the historical gas consumption data of the natural gas customer, perform normalization processing to obtain the normalized data to be analyzed, and determine the candidate cluster to which it belongs, that is, the detailed classification of the cluster to which it belongs.

[0106] In this embodiment, if there is no suitable classification for the natural gas customer in the stored detailed classification of the cluster, a new classification is created for the natural gas customer, and the corresponding model input parameters are obtained through training based on the historical gas consumption data of the natural gas customer and supplemented into the model input parameter library. In this way, during the natural gas customer prediction process, by initializing the model input initial parameters of the prediction model, the model input parameter library can be continuously supplemented.

[0107] A12. Obtain the optimal input parameter candidate sequence set corresponding to the candidate cluster from the model input parameter library.

[0108] In this embodiment, in the model input parameter library, search for the input parameter sequence set that conforms to the classification to which the natural gas customer belongs. As an optional embodiment, the input parameter sequence set may include multiple groups of model input parameters. If so, according to the preset screening strategy, screen out the optimal group of model input parameters as the final model input parameters, or by displaying the input parameter sequence set to the natural gas user, receiving the selection instruction input by the natural gas user, and obtaining the input parameter sequence corresponding to the selection instruction as the model input parameters of the prediction model for initialization.

[0109] A13. Based on the optimal input parameter candidate sequence set, assign values to the prediction model to obtain a gas load prediction model.

[0110] A14. Input the normalized data to be analyzed into the gas load prediction model to obtain the gas load prediction data of the natural gas customers to be predicted.

[0111] In this embodiment, as an alternative embodiment, refer to Figure 1 As shown, the method may further include the following steps:

[0112] Step S105. For each cluster, generate a model input parameter sequence set according to the model input parameters of the prediction model for this cluster; based on the model input parameter sequence sets corresponding to each cluster, construct a model input parameter library.

[0113] In this embodiment, based on the prediction model name, the input parameter sequence set, and the cluster corresponding to this input parameter sequence set, a model input parameter library is formed. By summarizing the model input parameter sequence sets of the prediction models corresponding to each type of natural gas customer obtained, a model input parameter library containing all the prediction models corresponding to all classified natural gas customers is obtained.

[0114] The method for initializing the model input parameters of the natural gas customer gas load prediction model proposed in this embodiment, through the historical gas consumption data scale, gas consumption patterns, etc. of natural gas customers, first classifies natural gas customers at multiple levels according to the industry, then according to the machine learning algorithm clustering, and finally according to the natural gas customer gas consumption data scale. Based on the historical gas consumption data of natural gas customers classified at each level, train multiple groups of initial model input parameters of the prediction model, evaluate each group of initial model input parameters according to multiple index evaluation functions such as R2, MAPE, RMSE, and MAE, select the best model input parameters from them, obtain the best model input parameter sequences of the prediction models corresponding to natural gas customers classified at each level, and use these model input parameters to initialize the prediction model to obtain a gas load prediction model for predicting the gas load of natural gas customers classified at this level, which can effectively improve the prediction accuracy of the gas load prediction model and achieve the purpose of quickly and accurately predicting the gas demand of natural gas customers.

[0115] In this embodiment, as an alternative embodiment, refer to Figure 1 As shown, the method further includes: Step S106. Update and optimize the input parameter candidate sequence set. Specifically, it may include the following steps:

[0116] Obtain other input parameter candidate sequence sets corresponding to the candidate clusters from the model input parameter library;

[0117] For each set of other input parameter candidate sequence sets, the prediction model is respectively assigned to normalize the prediction model with the input of the data to be analyzed, so as to obtain the sub-optimal prediction data of the gas consumption load of the natural gas customers to be predicted;

[0118] Based on the actual gas consumption load data corresponding to the gas consumption load prediction data of the natural gas customers to be predicted, the gas consumption load prediction data and the sub-optimal prediction data of the gas consumption load are evaluated;

[0119] If the evaluation accuracy of the sub-optimal prediction data of the gas consumption load is greater than the evaluation accuracy of the gas consumption load prediction data, the input parameter candidate sequence set corresponding to the sub-optimal prediction data of the gas consumption load is used to replace the optimal input parameter candidate sequence set.

[0120] In this embodiment, taking the current time as June and predicting the gas consumption load of natural gas customers in July as an example, the gas consumption load prediction data for July is obtained. After July has passed, the actual gas consumption load data for July of this natural gas customer can be obtained. Based on this actual gas consumption load data, the gas consumption load prediction data predicted based on each group of model input parameters can be verified. If the prediction result corresponding to the optimal input parameter candidate sequence set is not ideal, other groups of model input parameter values (input parameter candidate sequences) can be selected for input. If the prediction effect of this optimal input parameter candidate sequence set is the best, then this group of data is continued to be retained as the optimal model input parameters.

[0121] In this embodiment, based on the influence of the model input parameters on the prediction result of the gas consumption load prediction model, the natural gas customers are first classified, and the optimal model input parameters are respectively determined according to the data characteristics of each type of natural gas customer. In the actual application process, by matching the classification to which the natural gas customer to be predicted belongs, the optimal model input parameters corresponding to this classification are selected, and the model input parameter library storing the optimal model input parameters can be continuously supplemented, so as to effectively improve the prediction effect of the prediction model.

[0122] Based on the same inventive concept, an embodiment of the present invention further provides a device for initializing the input parameters of a gas consumption load prediction model. Referring to Figure 4 as shown, this device may include:

[0123] A normalization module 401, configured to obtain the historical gas consumption data of natural gas customers within a pre-designed calculation time range, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data;

[0124] In this embodiment, as an optional embodiment, the normalization module 401 includes:

[0125] An extreme value selection unit, configured to select the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value from the obtained historical gas consumption data;

[0126] A difference calculation unit, used for calculating the difference between the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value, to obtain a normalized difference;

[0127] The normalization unit is used to calculate the difference between each historical gas usage data and the historical gas usage data with the smallest value, and calculate the quotient of the difference and the normalized difference to obtain the normalized data of the historical gas usage data.

[0128] In this embodiment, as an optional embodiment, taking the historical gas consumption data as monthly gas consumption data as an example, the monthly gas consumption data is normalized using the following formula:

[0129] X i =(x i -min(∑x n )) / (max(∑x n )-min(∑x n ))

[0130] In this embodiment, as an optional embodiment, the normalization module 401 is further used for:

[0131] The historical gas consumption data of each natural gas customer is collected and processed into monthly gas consumption data, and the preset calculation time range is in years; the monthly gas consumption data is cleaned and processed.

[0132] In this embodiment, as an optional embodiment, cleaning the monthly gas consumption data includes:

[0133] For each natural gas customer, determine whether the monthly gas consumption data of the natural gas customer is continuous. If not, remove the monthly gas consumption data of the natural gas customer.

[0134] In this embodiment, processing the monthly gas consumption data includes:

[0135] For the monthly gas consumption data to be cleaned, the missing values, abnormal values ​​and repeated values ​​in the monthly gas consumption data are processed according to a preset processing strategy to obtain the monthly gas consumption data that meets the quality requirements.

[0136] The clustering and grading module 402 is used to classify the natural gas customers into cluster levels based on the industry to which the natural gas customers belong, the pre-set clustering algorithm, the user's gas consumption data volume and normalized data;

[0137] In this embodiment, as an optional embodiment, the clustering classification module 402 includes:

[0138] A classification unit is used to, for each normalized data, obtain the industry to which the natural gas customer corresponding to the normalized data belongs, query the classification result set. If the industry is in the classification result set, place the normalized data in the industry classification corresponding to the classification result set; if the industry is not in the classification result set, create a new industry classification in the classification result set and place the normalized data in the newly created industry classification.

[0139] A clustering unit is used to, for each industry classification in the classification result set, adopt a pre-set clustering algorithm to cluster the natural gas customers under the industry classification, and obtain a clustering cluster set containing multiple clustering clusters for the industry classification.

[0140] A sub-classification unit is used to, for each clustering cluster in the clustering cluster set, according to a pre-set normalized classification threshold, perform threshold sub-division on the normalized data included in the clustering cluster, and obtain multiple clustering cluster sub-classifications included in the clustering cluster.

[0141] A model training module 403, for each clustering for which clustering level division is performed, the user trains a prediction model based on the normalized data of the natural gas customers in the clustering, and obtains the model input parameters of the prediction model for the clustering.

[0142] In this embodiment, as an optional embodiment, the model training module 403 includes:

[0143] A model selection unit is used to determine the prediction model of the natural gas customers in the target clustering from the clusterings obtained by performing clustering level division.

[0144] An assignment unit is used to randomly assign values to the model input parameters of the determined prediction model to obtain multiple groups of different initial model input parameters.

[0145] A training unit is used to, for each group of initial model input parameters, respectively train the prediction model based on the normalized data included in the target clustering to adjust the initial model input parameters until the prediction model converges, and obtain each group of model input parameters corresponding to each group of initial model input parameters.

[0146] An evaluation unit is used to, for each group of model input parameters, respectively assign values to the prediction model, predict using the assigned prediction model based on the normalized data included in the target clustering, obtain the prediction results corresponding to each group of model input parameters respectively, and evaluate each prediction result respectively using a pre-set error algorithm.

[0147] A parameter determination unit is used to obtain the model input parameters of the prediction model for the target clustering based on the evaluation results.

[0148] The data prediction module 404 is configured to obtain the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, initialize the prediction model by using the obtained model input parameters, and predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the gas consumption load prediction data of the natural gas customer to be predicted.

[0149] In this embodiment, as an alternative embodiment, refer to Figure 4 As shown, the device may further include:

[0150] The parameter library construction module 405 is configured to, for each cluster, generate a model input parameter sequence set according to the model input parameters of the prediction model for the cluster; and construct a model input parameter library based on the model input parameter sequence sets corresponding to each cluster.

[0151] In this embodiment, as an alternative embodiment, the data prediction module 404 includes:

[0152] The cluster determination unit is configured to obtain the normalized data to be analyzed of the natural gas customer to be predicted, and determine the candidate clusters for hierarchical clustering division to which the natural gas customer to be predicted belongs;

[0153] The parameter extraction unit is configured to obtain the optimal input parameter candidate sequence set corresponding to the candidate cluster from the model input parameter library;

[0154] The initialization unit is configured to assign values to the prediction model based on the optimal input parameter candidate sequence set to obtain a gas consumption load prediction model;

[0155] The prediction unit is configured to input the normalized data to be analyzed into the gas consumption load prediction model to obtain the gas consumption load prediction data of the natural gas customer to be predicted.

[0156] In this embodiment, as another alternative embodiment, refer to Figure 4 As shown, the device further includes:

[0157] The parameter optimization module 406 is configured to obtain other input parameter candidate sequence sets corresponding to the candidate cluster from the model input parameter library; for each group of other input parameter candidate sequence sets, assign values to the prediction model respectively, input the normalized data to be analyzed into the assigned prediction model to obtain the sub-optimal gas consumption load prediction data of the natural gas customer to be predicted; evaluate the gas consumption load prediction data and the sub-optimal gas consumption load prediction data according to the actual gas consumption load data corresponding to the gas consumption load prediction data of the natural gas customer to be predicted; if the evaluation accuracy of the sub-optimal gas consumption load prediction data is greater than the evaluation accuracy of the gas consumption load prediction data, replace the optimal input parameter candidate sequence set with the input parameter candidate sequence set corresponding to the sub-optimal gas consumption load prediction data.

[0158] Based on the same inventive concept, an embodiment of the present invention further provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the model input parameter initialization method in any of the above possible implementation manners are implemented.

[0159] Optionally, the storage medium may be a non-transitory computer-readable storage medium. For example, the non-transitory computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0160] Based on the same inventive concept, referring to Figure 5 , an embodiment of the present invention further provides an electronic device, including a memory 101 (such as a non-volatile memory), a processor 102, and a computer program stored on the memory 101 and executable on the processor 102. When the processor 102 executes the program, the steps of the model input parameter initialization method in any of the above possible implementation manners are implemented, which is equivalent to the model input parameter initialization device as described above. Of course, the processor can also be used to process other data or perform operations. The electronic device may be a device such as a PC, a server, a terminal, etc.

[0161] As Figure 5 shown, the electronic device generally may further include: a memory 103, a network interface 104, and an internal bus 105. In addition to these components, other hardware may also be included, which will not be elaborated here.

[0162] It should be noted that the above model input parameter initialization device can be implemented by software. As a logically meaningful device, it is formed by the processor 102 of the electronic device where it is located reading the computer program instructions stored in the non-volatile memory into the memory 103 for running.

[0163] The embodiments of the subject matter and functional operations described in this specification can be implemented in the following: digital electronic circuits, tangible embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. The embodiments of the subject matter described in this specification can be implemented as one or more computer programs, that is, one or more modules in computer program instructions encoded on a tangible non-transitory program carrier to be executed by a data processing device or to control the operation of the data processing device. Alternatively or additionally, the program instructions can be encoded on an artificially generated propagated signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode information and transmit it to a suitable receiver device for execution by the data processing device. The computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.

[0164] The processes and logical flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform the corresponding functions by operating on input data and generating output. The processes and logical flows can also be performed by, for example, FPGA (Field Programmable Gate Array) or ASIC (Application Specific Integrated Circuit) of special logic circuits, and the apparatus can also be implemented as special logic circuits.

[0165] Computers suitable for executing computer programs include, for example, general and / or special microprocessors, or any other type of central processing unit. Generally, the central processing unit will receive instructions and data from a read-only memory and / or a random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, etc., or the computer will be operably coupled to such mass storage devices to receive data from them or transfer data to them, or both. However, a computer is not necessarily required to have such devices. In addition, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name just a few.

[0166] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special logic circuits.

[0167] Although this specification contains many specific implementation details, these should not be construed as limiting the scope of any invention or the scope of what is claimed, but rather as mainly describing the features of specific embodiments of a particular invention. Certain features described in multiple embodiments in this specification can also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment can also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may function in certain combinations as described above and even be claimed as such initially, one or more features from a claimed combination can in some cases be removed from that combination, and the claimed combination can be directed to a sub-combination or a variation of a sub-combination.

[0168] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or requiring that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0169] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0170] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0171] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features applied herein.

Claims

1. A method for initializing input parameters of a gas load prediction model, characterized in that, Including: Obtain the historical gas consumption data of natural gas customers within a pre-designed time range for budgeting, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data; Based on the industry to which the natural gas customers belong, a pre-set clustering algorithm, the scale of the user gas consumption data volume, and the normalized data, perform hierarchical clustering on the natural gas customers; For each cluster subjected to hierarchical clustering, train the prediction model based on the normalized data of the natural gas customers in this cluster to obtain the model input parameters of the prediction model for this cluster; Obtain the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, and initialize the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the gas load prediction data of the natural gas customer to be predicted.

2. The method for initializing model input parameters according to claim 1, characterized in that The method further includes: For each cluster, generate a model input parameter sequence set based on the model input parameters of the prediction model for this cluster; Based on the model input parameter sequence sets corresponding to each cluster, construct a model input parameter library.

3. The method for initializing model input parameters according to claim 2, wherein The step of obtaining the model input parameters corresponding to the target cluster to which the natural gas customer to be predicted belongs, initializing the prediction model with the obtained model input parameters to predict the normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain the gas load prediction data of the natural gas customer to be predicted includes: Obtain the normalized data to be analyzed of the natural gas customer to be predicted, and determine the candidate clusters for hierarchical clustering to which the natural gas customer to be predicted belongs; From the model input parameter library, obtain the optimal input parameter candidate sequence set corresponding to the candidate cluster; Based on the optimal input parameter candidate sequence set, assign values to the prediction model to obtain a gas load prediction model; Input the normalized data to be analyzed into the gas load prediction model to obtain the gas load prediction data of the natural gas customer to be predicted.

4. The method for initializing model input parameters according to claim 3, characterized in that, The method further includes: From the model input parameter library, obtain other input parameter candidate sequence sets corresponding to the candidate cluster; For each group of other input parameter candidate sequence sets, assign values to the prediction model respectively, input the normalized data to be analyzed into the prediction model with assigned values, and obtain the sub-optimal gas load prediction data of the natural gas customer to be predicted; Evaluate the gas load prediction data and the sub-optimal gas load prediction data according to the actual gas load data corresponding to the gas load prediction data of the natural gas customer to be predicted; If the evaluation accuracy of the sub-optimal gas load prediction data is greater than the evaluation accuracy of the gas load prediction data, replace the optimal input parameter candidate sequence set with the input parameter candidate sequence set corresponding to the sub-optimal gas load prediction data.

5. The method for initializing model input parameters according to any one of claims 1 to 4, characterized in that The step of performing normalization processing on the obtained historical gas consumption data includes: From the obtained historical gas consumption data, select the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value; Calculate the difference between the historical gas consumption data with the largest value and the historical gas consumption data with the smallest value to obtain a normalization difference; For each piece of historical gas consumption data, calculate the difference between this historical gas consumption data and the historical gas consumption data with the smallest value, and calculate the quotient of this difference and the normalized difference to obtain the normalized data of this historical gas consumption data.

6. The method for initializing model input parameters according to any one of claims 1 to 4, characterized in that Based on the industry to which the natural gas customers belong, a preset clustering algorithm, the scale of the user gas consumption data volume, and the normalized data, the hierarchical clustering of natural gas customers is carried out, including: For each piece of normalized data, obtain the industry to which the natural gas customer corresponding to this normalized data belongs, query the classification result set. If the industry is in the classification result set, place this normalized data in the industry classification corresponding to the classification result set. If the industry is not in the classification result set, create a new industry classification in the classification result set and place this normalized data in the newly created industry classification; For each industry classification in the classification result set, use the preset clustering algorithm to cluster the natural gas customers under this industry classification to obtain a clustering cluster set containing multiple clustering clusters for this industry classification; For each clustering cluster in the clustering cluster set, according to the preset normalized classification threshold, perform threshold subdivision on the normalized data contained in this clustering cluster to obtain multiple clustering cluster sub-classifications contained in this clustering cluster.

7. The method for initializing model input parameters according to any one of claims 1 to 4, characterized in that Based on the normalized data of the natural gas customers in this clustering, train the prediction model to obtain the model input parameters of the prediction model for this clustering, including: From the clusters obtained by the hierarchical clustering, determine the prediction model of the natural gas customers in the target cluster; Randomly assign values to the model input parameters of the determined prediction model to obtain multiple groups of different initial model input parameters; For each group of initial model input parameters, respectively train the prediction model according to the normalized data contained in the target cluster to adjust the initial model input parameters until the prediction model converges, and obtain each group of model input parameters corresponding to each group of initial model input parameters; For each group of model input parameters, respectively assign values to the prediction model, and based on the normalized data contained in the target cluster, use the assigned prediction model to make predictions to obtain the prediction results corresponding to each group of model input parameters respectively, and use the preset error algorithm to evaluate each prediction result respectively; Based on the evaluation results, obtain the model input parameters of the prediction model for the target cluster.

8. An input parameter initialization device for a gas load prediction model, characterized in that, Including: A normalization module, which is used to obtain the historical gas consumption data of natural gas customers within a preset calculation time range, and perform normalization processing on the obtained historical gas consumption data to obtain normalized data; A clustering and grading module, which is used to perform hierarchical clustering on natural gas customers based on the industry to which the natural gas customers belong, a preset clustering algorithm, the scale of the user gas consumption data volume, and the normalized data; A model training module, which is used to train the prediction model for each cluster obtained by the hierarchical clustering according to the normalized data of the natural gas customers in this cluster to obtain the model input parameters of the prediction model for this cluster; A data prediction module, configured to obtain model input parameters corresponding to a target cluster to which a natural gas customer to be predicted belongs, initialize a prediction model by using the obtained model input parameters, and predict normalized data to be analyzed of the natural gas customer to be predicted, so as to obtain gas load prediction data of the natural gas customer to be predicted.

9. A storage medium, characterized in that, A program or instruction is stored on a storage medium, and when the program or instruction is run by a processor, the method for initializing input parameters of a gas load prediction model according to any one of claims 1 to 7 is implemented.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, the method for initializing input parameters of a gas load prediction model according to any one of claims 1 to 7 is implemented.

Images