A gas load prediction method and device, electronic equipment and storage medium

Abstract

The application provides a gas load prediction method and device, electronic equipment and storage medium. The method comprises the following steps: obtaining historical original data in a target area; performing feature extraction on the historical original data to obtain data features of the historical original data; determining specified influence factor data related to the data features of the historical original data according to the data features of the historical original data; comparing the specified influence factor data with historical actual influence factor data to obtain change amount data of the influence factor data; training prediction sample data and the change amount data of the influence factor data by using a gas load prediction model to determine a first prediction value of the gas load; collecting pipeline pressure data in the target area, and calling a pipeline storage data prediction model to train the pipeline pressure data to determine a pipeline storage gas prediction value; and determining a gas load prediction value according to the first prediction value and the pipeline storage gas prediction value. The application solves the problem of accurately predicting the usage of natural gas.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, and particularly relates to a gas load prediction method and device, electronic equipment and storage medium. BACKGROUND

[0002] With the rapid development of natural gas industry, the scheduling management, planning operation and operation optimization of natural gas pipelines must rely on accurate load prediction technology. More accurate prediction of natural gas consumption is of great significance for the optimization scheduling of future natural gas pipeline networks connecting cities, reasonable planning of gas consumption by gas companies and optimization operation of urban gas pipeline networks.

[0003] The existing gas load prediction method predicts the gas load according to weather influencing factors and natural gas consumption. However, in real natural gas application scenarios, factors such as holidays and heating periods will greatly affect the consumption of natural gas. In addition, since most natural gas is transported by long-distance pipelines, there will be residual natural gas in the pipelines, and the content of residual natural gas cannot be ignored in prediction. Therefore, the existing prediction method cannot accurately predict the consumption of natural gas in combination with real natural gas application scenarios. SUMMARY

[0004] Therefore, the embodiments of the present disclosure provide a gas load prediction method and device, electronic equipment and storage medium to solve the problem that the existing prediction method cannot accurately predict the consumption of natural gas in combination with real natural gas application scenarios.

[0005] In a first aspect, the embodiments of the present disclosure provide a gas load prediction method, comprising:

[0006] obtaining historical original data in a target area, the historical original data including historical actual consumption data of natural gas and historical pipeline storage data of natural gas;

[0007] performing feature extraction on the historical original data to obtain data features of the historical original data;

[0008] determining specified influencing factor data related to the data features of the historical original data according to the data features of the historical original data;

[0009] comparing the specified influencing factor data with historical actual influencing factor data to obtain change amount data of the influencing factor data;

[0010] training the prediction sample data and the change amount data of the influencing factor data by using a gas load prediction model to determine a first prediction value of the gas load, wherein the gas load prediction model is a neural network model based on a knowledge distillation framework;

[0011] Collect pipeline pressure data in the target area, call the pipeline data prediction model to train the pipeline pressure data, and determine the pipeline gas consumption prediction value;

[0012] According to the first prediction value and the pipeline gas consumption prediction value, the gas consumption load prediction value is determined.

[0013] In a second aspect, the embodiment of the present disclosure provides a gas consumption load prediction device, which comprises:

[0014] The data acquisition module is configured to acquire historical original data in the target area, and the historical original data comprises historical actual use data of natural gas and historical pipeline data of natural gas;

[0015] The feature extraction module is configured to perform feature extraction on the historical original data to obtain data features of the historical original data;

[0016] The feature determination module is configured to determine specified influence factor data related to the data features of the historical original data according to the data features of the historical original data;

[0017] The data comparison module is configured to compare the specified influence factor data with the historical actual influence factor data to obtain change amount data of the influence factor data;

[0018] The determination module is configured to train the prediction sample data and the change amount data of the influence factor data by using a gas consumption load prediction model to determine a first prediction value of the gas consumption load, wherein the gas consumption load prediction model is a neural network model based on a knowledge distillation framework; collect pipeline pressure data in the target area, call the pipeline data prediction model to train the pipeline pressure data, and determine a pipeline gas consumption prediction value; and according to the first prediction value and the pipeline gas consumption prediction value, the gas consumption load prediction value is determined.

[0019] In a third aspect, the embodiment of the present disclosure provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.

[0020] In a fourth aspect, the embodiment of the present disclosure provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the above method when executed by a processor.

[0021] Compared with the prior art, the beneficial effects of the embodiments of the present disclosure are that: in combination with the use scene of natural gas, the residual data in the natural gas pipeline is taken as part of the historical original data of the target area, the data characteristics of actual gas use and the data characteristics of pipeline residues are obtained, the actual natural gas use situation can be more accurately and comprehensively reflected; and the specified influence factor data is obtained by analyzing the historical original data, the influence factor data is compared with the actual influence factor data, the change amount of the influence factor is obtained, the influence factor is further accurately processed, the first prediction value is obtained according to the sample data and the change amount of the influence factor, the pipeline gas prediction value is obtained according to the pipeline pressure value, the final gas load prediction value is obtained in combination with the first prediction value and the pipeline gas prediction value, and accurate prediction is realized. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without creative labor.

[0023] Figure 1 is a joint learning architecture schematic diagram of an embodiment of the present disclosure;

[0024] Figure 2 is a flow schematic diagram of a gas load prediction method provided by an embodiment of the present disclosure;

[0025] Figure 3 is a prediction flow schematic diagram of a gas load prediction model provided by an embodiment of the present disclosure;

[0026] Figure 4 is a structure schematic diagram of a gas load prediction device provided by an embodiment of the present disclosure;

[0027] Figure 5 is a structure schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0028] In the following description, specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. However, persons skilled in the art should understand that the present disclosure can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present disclosure with unnecessary details.

[0029] Joint learning refers to comprehensively utilizing various AI (Artificial Intelligence) technologies under the premise of ensuring data security and user privacy, jointly mining data value by multiple parties, and giving birth to new intelligent formats and modes based on joint modeling. Joint learning has at least the following characteristics:

[0030] (1) Participating nodes control the weak centralized joint training mode of self-owned data to ensure data privacy and security in the process of co-creating intelligence.

[0031] (2) In different application scenarios, AI algorithms, privacy protection calculations are used to screen and / or combine, and various model aggregation optimization strategies are established to obtain high-level and high-quality models.

[0032] (3) Under the premise of ensuring data security and user privacy, based on various model aggregation optimization strategies, methods for improving the performance of joint learning engines are obtained, wherein the performance methods can be to improve the overall performance of the joint learning engine by solving problems including parallel computing architecture, information interaction under large-scale cross-domain network, intelligent perception, and abnormal processing mechanism.

[0033] (4) Obtain the needs of multiple users in each scenario, determine the real contribution of each joint participant through a mutual trust mechanism, and allocate incentives.

[0034] Based on the above method, an AI technology ecosystem based on joint learning can be established to fully realize the value of industry data and promote the landing of vertical field scenarios.

[0035] A gas load prediction method and device according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0036] Figure 1 is a schematic diagram of a joint learning architecture according to an embodiment of the present disclosure. As shown in Figure 1 , the joint learning architecture can include a server (central node) 101 and a participant 102, a participant 103, and a participant 104.

[0037] In the federated learning process, the basic model can be established by the server 101, and the server 101 sends the model to the participants 102, 103 and 104 which establish a communication connection with the server 101. The basic model can also be established by any participant and uploaded to the server 101, and the server 101 sends the model to other participants which establish a communication connection with the server 101. The participants 102, 103 and 104 construct the model according to the downloaded basic structure and model parameters, train the model using local data, obtain updated model parameters, and upload the updated model parameters to the server 101 in an encrypted manner. The server 101 aggregates the model parameters sent by the participants 102, 103 and 104, obtains global model parameters, and transmits the global model parameters back to the participants 102, 103 and 104. The participants 102, 103 and 104 iterate the respective models according to the received global model parameters until the models finally converge, thereby realizing the training of the models. In the federated learning process, the data uploaded by the participants 102, 103 and 104 is the model parameters, and the local data is not uploaded to the server 101, and all participants can share the final model parameters, so that common modeling can be realized on the basis of ensuring data privacy. It should be noted that the number of participants is not limited to three as described above, but can be set as needed, and the embodiments of the present disclosure do not limit this.

[0038] Figure 2 is a flowchart of a gas load prediction method provided by an embodiment of the present disclosure. Figure 2 The gas load prediction method of Figure 1 may be executed by a server or a participant. As shown in Figure 2 , the gas load prediction method comprises:

[0039] S201, obtaining historical raw data in a target area, the historical raw data comprising historical actual use data of natural gas and historical pipe storage data of natural gas;

[0040] The historical actual use data is the amount of natural gas used in the target area, and the historical pipe storage data is the residual data of the natural gas pipeline in the target area.

[0041] S202, performing feature extraction on the historical raw data to obtain data features of the historical raw data;

[0042] The data features of the historical raw data are used to represent the data features of the amount of natural gas used in the target area and the data features of the residual data of the natural gas pipeline in the target area.

[0043] S203, determining specified influence factor data related to the data features of the historical raw data according to the data features of the historical raw data;

[0044] Specifically, the specified influencing factor data related to the data characteristics of the historical raw data can be determined by a preset data analysis method, wherein the specified influencing factor data includes any one or more of weather data, heating period data, and holiday data.

[0045] S204, comparing the specified influencing factor data with the historical actual influencing factor data to obtain variation data of the influencing factor data;

[0046] Specifically, the specified influencing factor data is compared with the historical actual influencing factor data to determine the variation of the influencing factor data.

[0047] S205, training the prediction sample data and the variation data of the influencing factor data by using the gas consumption load prediction model to determine the first prediction value of the gas consumption load;

[0048] The gas consumption load prediction model is a neural network model based on a knowledge distillation framework.

[0049] Specifically, the prediction sample data and the variation are input into the pre-trained gas consumption load prediction model to determine the first prediction value of the gas consumption load, wherein the gas consumption load prediction model is a neural network model based on a knowledge distillation framework.

[0050] S206, collecting pipeline pressure data in the target area, calling a pipeline data prediction model to train the pipeline pressure data, and determining a pipeline gas prediction value.

[0051] S207, determining a gas consumption load prediction value according to the first prediction value and the pipeline gas prediction value.

[0052] Specifically, in order to combine the real use scene of natural gas and obtain accurate and comprehensive gas consumption load prediction data, in the historical data acquisition stage, the electronic device collects the actual natural gas consumption in the target area as historical actual use data. Due to the particularity of the natural gas transportation scene, residual natural gas data in the natural gas pipeline data in the target area is also collected, and the residual data in the natural gas pipeline is taken as historical pipeline storage data.

[0053] Specifically, the historical raw data is extracted by feature extraction, which can be filled with missing values by filling the actual gas consumption data and the historical pipeline storage data after filling with specified methods. According to the detection result, the filled historical actual gas consumption data and the filled historical pipeline storage data are replaced by abnormal values to obtain the historical raw data after preprocessing; according to a preset feature extraction algorithm, a first data feature group is extracted from the historical raw data after preprocessing;

[0054] The historical raw data after preprocessing is clustered and analyzed, and the historical raw data after preprocessing is divided into multiple data groups; the data in each data group is analyzed for characteristics, the common characteristics in the group are determined, and the characteristics between multiple data groups are analyzed to determine the difference characteristics between the groups; the second data characteristic group is determined according to the common characteristics in the group and the difference characteristics between the groups; and the data characteristics of the historical raw data are determined according to the first data characteristic group and the second data characteristic group.

[0055] It should be noted that feature extraction refers to a series of processing of historical raw data, and the historical raw data is refined into data characteristics combined with actual application scenarios, so as to facilitate subsequent related calculations through algorithms and models. It should be noted that the data characteristics of the historical raw data are used to represent the data characteristics of the natural gas usage and the data characteristics of the natural gas pipeline residue in the target area.

[0056] Specifically, according to the data characteristics of the historical raw data, the specified influence factor data related to the data characteristics of the historical raw data can be obtained by obtaining the influence factor data corresponding to the historical raw data in the target area; the historical raw data and the influence factor data are respectively processed by dimensionless processing, the processed historical raw data is taken as a control series, and the processed influence factor data is taken as a comparison series; the correlation coefficient of the control series and the comparison series is calculated, and the correlation degree of the influence factor data and the historical raw data is determined according to the correlation coefficient; the correlation degree is sorted according to a preset rule, and the influence factor data with a correlation degree higher than a preset threshold is taken as the specified influence factor data.

[0057] Further, the data characteristics of the historical raw data are analyzed by a pre-set data analysis method to determine the specified influence factor characteristics related to the historical raw data, and the specified influence factor data is determined according to the specified influence factor characteristics. That is, the influence factor data affecting the natural gas usage is extracted from the data characteristics of the historical raw data. Since the natural gas usage is closely related to the weather, the heating period, and the holidays, for example, the natural gas usage in sunny weather is less than that in rainy weather, the natural gas usage during the heating period is significantly different from that during the non-heating period, and the natural gas usage sharply decreases during the National Day holiday, the Spring Festival, and other specific holidays. Therefore, by analyzing the historical raw data of natural gas, the influence factor data affecting the historical raw data is obtained, and the specified influence factor data obtained in this way can more accurately represent the influence size on the natural gas usage.

[0058] Specifically, the real influence factor data corresponding to the historical raw data in the time is obtained, the real influence factor data corresponding to the historical time is compared with the specified influence factor data, and the change amount of the influence factor is obtained, for example, the obtained specified influence factor data is the temperature of 20 degrees in the weather data, and the temperature in the real weather data in the corresponding time is 25 degrees, and then the change amount of the weather data is 5 degrees. That is, the disclosure embodiment considers the change of the actual factor, so that more accurate data range of the influence factor data can be obtained.

[0059] Specifically, the change amount data of the prediction sample data and the influence factor data is trained by using the gas load prediction model to determine the first prediction value of the gas load. The data set of the historical gas data and the influence factor data can be collected; the data in the data set of the influence factor data is retrieved from the pre-constructed knowledge distillation teacher model to determine the historical gas data as the output soft target and / or the output prediction gas data; the loss function of the teacher model is modified according to the output soft target, and the teacher model is trained; the data in the data set of the influence factor data is input into the pre-constructed knowledge distillation student model, and the prediction gas data is taken as the target value of the student model, and the student model is trained to obtain the required gas load prediction model.

[0060] Further, the prediction sample data and the change amount are jointly input into the pre-trained gas load prediction model to determine the first prediction value of the gas load. The gas load prediction model is a neural network model based on a knowledge distillation framework. It should be noted that the knowledge distillation framework mainly uses a larger already trained network to teach a smaller network exactly what to do step by step, and then the small network is trained to learn the accurate behavior of the large network by trying to replicate the output of the large network at each layer. The framework mainly consists of two parts, Teacher model and Student model. The Teacher model of the present application compensates for the problem that the collection frequency of gas data and weather influence factor data is different in gas prediction through a deep learning network, and can simulate an hourly data model of the influence factor data. The Student model realizes the hourly real-time prediction of the gas data through a random forest. The prediction sample data and the sample data change amount are jointly input into the model, the influence factor change is considered, and a more accurate first prediction value can be obtained.

[0061] Specifically, in the actual application scenario, there is also some natural gas in the natural gas pipeline, and the residual amount in the natural gas pipeline is usually ignored when predicting the use of the front gas. Therefore, in the embodiment of the present disclosure, the pipeline pressure value in the target area is collected, and the pipeline pressure value is input into the pre-trained pipeline data prediction model to determine the pipeline gas prediction value through the pipeline pressure value.

[0062] Specifically, the gas consumption load prediction value is determined according to the first prediction value and the pipe-stored gas prediction value.

[0063] According to the technical scheme provided by the embodiment of the present disclosure, by combining the use scene of natural gas, taking the residual data in the natural gas pipeline as part of the historical original data of the target region, the data characteristics of the actual gas consumption and the data characteristics of the pipeline residual are obtained, which can more accurately and comprehensively reflect the actual use of natural gas; and the specified influence factor data is obtained by analyzing the historical original data, and the change amount of the influence factor is obtained by comparing the influence factor data with the actual influence factor data, so that the influence factor is further accurately processed, the first prediction value is obtained according to the sample data and the change amount of the influence factor, and the pipe-stored gas prediction value is obtained according to the pipeline pressure value, and the final gas consumption load prediction value is obtained by combining the first prediction value and the pipe-stored gas prediction value, so as to realize accurate prediction.

[0064] In some embodiments, the historical original data is subjected to feature extraction to obtain data characteristics of the historical original data, specifically including: the historical original data is subjected to missing value filling, and the filled actual gas consumption data and the filled historical pipe-stored data are subjected to abnormal value detection by a specified method; according to the detection result, the filled historical actual gas consumption data and the filled historical pipe-stored data are subjected to abnormal value replacement to obtain preprocessed historical original data; a first data feature group in the preprocessed historical original data is extracted according to a preset feature extraction algorithm; the preprocessed historical original data is subjected to cluster analysis, and the preprocessed historical original data is divided into a plurality of data groups; the data in each data group is subjected to feature analysis to determine the common features in the group, and the feature analysis is performed between the plurality of data groups to determine the difference features between the groups; a second data feature group is determined according to the common features in the group and the difference features between the groups; and the data characteristics of the historical original data are determined according to the first data feature group and the second data feature group.

[0065] Specifically, since the obtained historical original data of natural gas is obtained from different sources, there are abnormal data in the historical original data, and therefore, the historical original data needs to be preprocessed. The actual natural gas historical original data is less in quantity due to the particularity of the data, and therefore, each piece of data needs to be processed as useful data. The historical original data can be subjected to missing value filling, and the filling data can be determined according to the average data of the similar time or combined with experience setting. After the historical original data is subjected to missing value filling, the filled data is subjected to abnormal value detection, and the detected abnormal value is replaced with a normal value, wherein the abnormal value detection method can be a simple statistical method or other detection methods, which are not limited here.

[0066] After the historical raw data is preprocessed, feature data extraction is performed on the historical raw data after preprocessing. First, feature extraction is performed on the historical raw data after preprocessing to obtain a first data feature group of the historical raw data, and the first data feature group is used to represent data features common to the historical raw data. The extraction algorithm can be a principal component analysis algorithm or an independent component analysis algorithm. Then, clustering analysis is performed on the historical raw data after preprocessing, and the historical raw data is divided into multiple data groups. Feature analysis is performed on the data in each data group in each subgroup, and the value frequency of each data variable in the corresponding data group is counted. The distribution of the value frequency is more concentrated as the common feature in the data group, and the common feature in the group is taken as the common feature in the group.

[0067] Feature analysis is performed between multiple data groups, and the data variable with a more obvious difference in value frequency between two different data groups is taken as the difference feature between the two data groups. The common feature in the group and the difference feature between the groups are combined to form a second data feature group. The second data feature group is used to represent more detailed data features of the historical raw data. The data features in the first data feature group and the second data feature group are taken as the data features of the historical raw data.

[0068] According to the technical scheme provided by the embodiments of the present disclosure, the historical raw data is preprocessed to avoid the influence of missing values and outliers in the data on the feature extraction effect. By performing overall feature extraction on the historical raw data and feature extraction on the grouped data, and combining the two feature extraction methods to obtain data features, the accuracy and comprehensiveness of the data features of the historical raw data can be effectively ensured.

[0069] In some embodiments, by specifying the data analysis method, the specified influence factor data related to the feature vector of the historical raw data is determined, specifically including: obtaining the influence factor data corresponding to the historical raw data in the target area; performing dimensionless processing on the historical raw data and the influence factor data respectively, taking the processed historical raw data as a control series and the processed influence factor data as a comparison series; calculating the correlation coefficient of the control series and the comparison series, and determining the correlation degree of the influence factor data and the historical raw data according to the correlation coefficient; sorting the correlation degrees according to a preset rule, and taking the influence factor data with a correlation degree higher than a preset threshold as the specified influence factor data.

[0070] Specifically, according to the corresponding time of the historical raw data, the influence factor data at the same time in the target area is obtained, for example, the weather data in the corresponding time period, whether the corresponding time period is a heating period, and whether the corresponding time period is a holiday, the influence factor data and the historical raw data are respectively processed in a dimensionless manner, so that the influence factor data and the historical raw data can be compared and calculated. The processed historical raw data is used as a control series, and the processed influence factor data is used as a comparison series. The correlation coefficient of the control series and the comparison series is calculated, and the correlation degree of the influence factor data and the historical raw data is determined according to the correlation coefficient. It should be noted that the correlation degree is used to represent the correlation degree of the influence factor data and the historical raw data. The greater the correlation degree of the influence factor data and the historical raw data, the greater the influence of the corresponding influence factor on the gas data. The correlation degrees are sorted according to the numerical values, and the influence factor data with a correlation degree higher than a preset threshold is used as specified influence factor data. It should be noted that the preset threshold can be determined according to the actual situation of the user. When the requirement for the influence factor is strict, data ranked in the top 70% and having a correlation degree greater than a specified value can be selected.

[0071] According to the technical scheme provided by the embodiments of the present disclosure, the historical raw data and the influence factor data are processed in a dimensionless manner, so that the two types of data can be compared and calculated. Through correlation degree calculation, the influence of the influence factor on the natural gas usage is quantitatively displayed, and a more accurate influence factor type can be obtained.

[0072] In some embodiments, a data set of historical gas usage data and influence factor data is collected; a pre-constructed knowledge distillation teacher model is called to train data in the data set of influence factor data, and the historical gas usage data is determined as an output soft target and / or an output predicted gas usage data; the loss function of the teacher model is modified according to the output soft target, and the teacher model is trained; data in the data set of influence factor data is input into a pre-constructed knowledge distillation student model, and the predicted gas usage data is used as a target value of the student model, so that the student model is trained to obtain a required gas load prediction model.

[0073] Specifically, a teacher model of a multi-layer feedforward neural network is constructed, including an input layer, two hidden layers, and an output layer. The input variable dimension of the input layer is 32 dimensions, the number of neurons of the two hidden layers is 512 and 256 respectively, and the output layer is used to output the predicted value of the gas data. The ReLU function is selected as the activation function, which is beneficial to improve the training speed and efficiency of the model. The student model is constructed by using the random forest algorithm Xgboost. Xgboost trains a tree in each round, so that the loss function can be minimized. The loss function not only measures the fitting error of the model, but also increases the regularization term, that is, the penalty term of the complexity of each tree, to prevent overfitting.

[0074] The data set of historical gas consumption data and influencing factor data is constructed, the influencing factor data is input into the pre-constructed teacher model, the historical gas consumption data is taken as an output soft target, and predicted gas consumption data is output. According to the output soft target, the loss function of the teacher model is modified, and the teacher model is trained. The influencing factor data is input into the pre-constructed student model, the predicted gas consumption data is taken as a target value of the student model, the student model is trained, and a required gas consumption load prediction model is obtained, as shown in Figure 3 Figure 3 The prediction process of the gas consumption load prediction model provided by the embodiments of the present disclosure is provided, the influencing factor data and the natural gas consumption data in the sample data are respectively input into the teacher model and the student model, the teacher model outputs the predicted gas consumption data at the hour level, and the data is taken as a soft target and a target value of the student model, and the final prediction result is obtained through the student model.

[0075] According to the technical scheme provided by the embodiments of the present disclosure, the Teacher can solve the problem that the gas consumption data and the influencing factor data have different collection frequencies in gas consumption prediction through a deep learning network, a data model of the influencing factor data at the hour level can be simulated, and the Student model can realize real-time prediction of the gas consumption data at the hour level through a random forest algorithm.

[0076] In some embodiments, according to the first prediction value and the pipe storage gas consumption prediction value, a gas consumption load prediction value is determined, specifically including: calculating a ratio of the pipe storage gas consumption prediction value and the first prediction value to obtain a first ratio; calculating a ratio of the historical pipe storage data and the historical actual gas consumption data to obtain a second ratio; if a difference between the first ratio and the second ratio is greater than a preset threshold, the first prediction value and the pipe storage gas consumption prediction value are weighted to obtain the gas consumption load prediction value.

[0077] ​Specifically, in combination with the actual natural gas application scene, in large cities, natural gas is mostly transmitted by long-distance pipelines, and the residual natural gas content in the long-distance pipeline cannot be ignored and also belongs to part of the natural gas scheduling link. However, in smaller areas, the residual in the natural gas pipeline is less and can be ignored. Therefore, whether the pipe inventory data needs to be added to the final predicted value can be determined by the ratio of the pipe inventory to the actual consumption. The ratio of the pipe inventory consumption prediction value to the first prediction value is calculated as the first ratio, and the first ratio is used to represent the data proportion obtained by prediction. The ratio of the historical pipe inventory data to the historical actual gas consumption data is calculated as the second ratio, and the second ratio is used to represent the actual proportion of the historical pipe inventory data in the historical actual gas consumption data. Whether the first ratio needs to add pipe inventory data is determined according to the second ratio. If the difference between the first ratio and the second ratio is greater than a preset threshold, the pipe inventory consumption prediction value and the first prediction value are weighted to obtain the final gas load prediction value.

[0078] According to the technical scheme provided by the embodiments of the present disclosure, the gas load prediction data is obtained by combining the actual natural gas application scene, the prediction data is closer to the actual scene, and the accuracy of the prediction data is increased.

[0079] In some embodiments, the pipe pressure value is input into the pre-trained pipe inventory data prediction model to determine the pipe inventory consumption prediction value. Previously, the historical pipe pressure value and the historical pipe inventory data are used to construct a pipe inventory data set. The historical pipe pressure value in the pipe inventory data set is input into the pre-constructed neural network model, the historical pipe inventory data is used as the output target of the neural network model, the model parameters are adjusted, and the model is trained to determine the required pipe inventory data prediction model.

[0080] Specifically, the neural network model is pre-constructed, the historical pipe pressure value and the historical pipe inventory data are used to train the neural network model, the model parameters are adjusted according to the historical pipe inventory data, and the required pipe inventory data prediction model is determined.

[0081] According to the technical scheme provided by the embodiments of the present disclosure, the pipe inventory prediction data is obtained by the pipe inventory data prediction model, the calculation time is saved, and the accuracy of the pipe inventory prediction data is further enhanced.

[0082] In some embodiments, the gas load is used to represent the natural gas consumption in the target area.

[0083] All the optional technical solutions described above can be combined to form optional embodiments of the present application, which will not be described here.

[0084] The following is an embodiment of the device of the present disclosure, which can be used to execute the method embodiment of the present disclosure. For details not disclosed in the device embodiment of the present disclosure, please refer to the method embodiment of the present disclosure.

[0085] Figure 4 is a schematic diagram of a gas load prediction device provided by an embodiment of the present disclosure. As shown in Figure 4 , the gas load prediction device comprises:

[0086] The data acquisition module 401 is configured to acquire historical raw data in the target area, and the historical raw data comprises historical actual use data of natural gas and historical pipe storage data of natural gas, wherein the historical actual use data is the natural gas consumption in the target area, and the historical pipe storage data is the residual data of the natural gas pipeline in the target area.

[0087] The feature extraction module 402 is configured to extract features from the historical raw data to obtain data features of the historical raw data, wherein the data features of the historical raw data are used to represent the data features of the natural gas consumption in the target area and the data features of the natural gas pipeline residual in the target area.

[0088] The feature determination module 403 is configured to determine specified influence factor data related to the data features of the historical raw data according to the data features of the historical raw data, wherein the specified influence factor data comprises any one or more of weather data, heating period data and holiday data.

[0089] The data comparison module 404 is configured to compare the specified influence factor data with the historical actual influence factor data to obtain change amount data of the influence factor data.

[0090] The determination module 405 is configured to train the prediction sample data and the change amount data of the influence factor data by using a gas load prediction model, to determine a first prediction value of the gas load, wherein the gas load prediction model is a neural network model based on a knowledge distillation framework; to collect pipeline pressure data in the target area, to retrieve a pipe storage data prediction model to train the pipeline pressure data, to determine a pipe storage gas prediction value; and to determine a gas load prediction value according to the first prediction value and the pipe storage gas prediction value.

[0091] According to the technical scheme provided by the embodiment of the present disclosure, in combination with the use scene of natural gas, the residual data in the natural gas pipeline is taken as part of the historical original data of the target region, the data features of actual gas use and the data features of pipeline residual are obtained, the actual natural gas use situation can be more accurately and comprehensively reflected; and the specified influence factor data is obtained by analyzing the historical original data, the influence factor data is compared with the actual influence factor data, the change amount of the influence factor is obtained, the influence factor is further accurately processed, the first prediction value is obtained according to the sample data and the change amount of the influence factor, the pipeline gas use prediction value is obtained according to the pipeline pressure value, the final gas use load prediction value is obtained in combination with the first prediction value and the pipeline gas use prediction value, and accurate prediction is realized.

[0092] In some embodiments, the feature extraction module 402 is also configured to perform missing value filling on the historical original data, perform abnormal value detection on the filled actual gas use data and the filled historical pipeline storage data through a specified manner, perform abnormal value replacement on the filled historical actual gas use data and the filled historical pipeline storage data according to the detection result, and obtain the preprocessed historical original data; extract a first data feature group in the preprocessed historical original data according to a preset feature extraction algorithm; perform clustering analysis on the preprocessed historical original data, and divide the preprocessed historical original data into a plurality of data groups; perform feature analysis on the data in each data group to determine the common features in the group, and perform feature analysis on a plurality of data groups to determine the difference features between the groups; determine a second data feature group according to the common features in the group and the difference features between the groups; and determine the data features of the historical original data according to the first data feature group and the second data feature group.

[0093] In some embodiments, the feature determination module 403 is also configured to obtain influence factor data corresponding to the historical original data in the target region; perform dimensionless processing on the historical original data and the influence factor data respectively, take the processed historical original data as a control series, and take the processed influence factor data as a comparison series; calculate the correlation degree coefficient of the control series and the comparison series, determine the correlation degree of the influence factor data and the historical original data according to the correlation degree coefficient; sort the correlation degrees according to a preset rule, and take the influence factor data with a correlation degree higher than a preset threshold as the specified influence factor data.

[0094] In some embodiments, the system further includes a model determination module 406, configured to collect a dataset of historical gas consumption data and influencing factor data; retrieve data from the dataset of influencing factor data used for training a pre-built knowledge distillation teacher model, determine historical gas consumption data as the output soft target and / or output predicted gas consumption data; modify the loss function of the teacher model according to the output soft target, and train the teacher model; input the data from the dataset of influencing factor data into the pre-built knowledge distillation student model, use the predicted gas consumption data as the target value of the student model, and train the student model to obtain a gas load prediction model that meets the requirements.

[0095] In some embodiments, the determining module 405 is further configured to calculate the ratio of the predicted gas consumption value in the pipeline to the first predicted value to obtain a first ratio; calculate the ratio of historical pipeline data to historical actual gas consumption data to obtain a second ratio; and if the difference between the first ratio and the second ratio is greater than a preset threshold, perform a weighted calculation on the first predicted value and the predicted gas consumption value in the pipeline to obtain a predicted gas load value.

[0096] In some embodiments, the model determination module 406 is further configured to construct a pipe storage dataset based on historical pipe pressure values ​​and historical pipe storage data; input the historical pipe pressure values ​​in the pipe storage dataset into a pre-constructed neural network model; use the historical pipe storage data as the output target of the neural network model; adjust the model parameters; and train the model to determine a pipe storage data prediction model that meets the requirements.

[0097] In some embodiments, gas load is used to represent the amount of natural gas used within a target area.

[0098] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this disclosure.

[0099] Figure 5 This is a schematic diagram of the electronic device 5 provided in an embodiment of this disclosure. Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, it implements the steps in the various method embodiments described above. Alternatively, when the processor 501 executes the computer program 503, it implements the functions of each module / unit in the various device embodiments described above.

[0100] Exemplarily, the computer program 503 can be divided into one or more modules / units, which are stored in the memory 502 and executed by the processor 501 to complete the present disclosure. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 503 in the electronic device 5.

[0101] The electronic device 5 can be a desktop computer, a notebook computer, a palm computer, a cloud server and the like. The electronic device 5 can include but is not limited to the processor 501 and the memory 502. Those skilled in the art can understand that the electronic device 5 can include more or less components, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus and the like. Figure 5 The electronic device 5 is only an example and does not constitute a limitation on the electronic device 5, and can include more or less components than the diagram, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus and the like.

[0102] The processor 501 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0103] The memory 502 can be an internal storage unit of the electronic device 5, for example, a hard disk or a memory of the electronic device 5. The memory 502 can also be an external storage device of the electronic device 5, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card and the like equipped on the electronic device 5. Further, the memory 502 can include both the internal storage unit and the external storage device of the electronic device 5. The memory 502 is used to store computer programs and other programs and data required by the electronic device. The memory 502 can also be used to temporarily store data that has been output or will be output.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is taken as an example, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or software. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above system can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0105] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0106] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present disclosure.

[0107] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices / equipment and methods can be implemented in other ways. For example, the device / equipment embodiments described above are only schematic. For example, the division of modules or units is only a logical function division, and actual implementation can have another division manner. Multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed mutual units can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0108] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the present embodiment.

[0109] In addition, each functional unit in each of the embodiments of the present disclosure can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0110] If the integrated module / unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can be completed by the computer program instructing the related hardware, and the computer program can be stored in the computer readable storage medium. When the processor executes the computer program, the steps of each method embodiment described above can be implemented. The computer program can include computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electric carrier signals and telecommunication signals.

[0111] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure.

Claims

1. A gas load prediction method characterized by, The method comprises the following steps: acquiring historical raw data in a target area, the historical raw data comprising historical actual use data of natural gas and historical pipe storage data of natural gas; extracting features from the historical raw data to obtain data features of the historical raw data; determining specified influence factor data related to the data features of the historical raw data according to the data features of the historical raw data; comparing the specified influence factor data with historical actual influence factor data to obtain change amount data of the influence factor data; training prediction sample data and the change amount data of the influence factor data by using a gas load prediction model to determine a first prediction value of the gas load, wherein the gas load prediction model is a neural network model based on a knowledge distillation framework; collecting pipeline pressure data in the target area, and calling a pipe storage data prediction model to train the pipeline pressure data to determine a pipe storage gas prediction value; determining a gas load prediction value according to the first prediction value and the pipe storage gas prediction value; determining specified influence factor data related to the data features of the historical raw data according to the data features of the historical raw data, specifically comprising: acquiring influence factor data corresponding to the historical raw data in the target area; respectively performing dimensionless processing on the historical raw data and the influence factor data, taking the processed historical raw data as a control series and taking the processed influence factor data as a comparison series; calculating a correlation coefficient of the control series and the comparison series, and determining a correlation degree of the influence factor data and the historical raw data according to the correlation coefficient; sorting the correlation degrees according to a preset rule, and taking the influence factor data with a correlation degree higher than a preset threshold as the specified influence factor data; training prediction sample data and the change amount data of the influence factor data by using a gas load prediction model to determine a first prediction value of the gas load, comprising: collecting a data set of historical gas use data and influence factor data; calling a pre-constructed knowledge distillation teacher model to train data in the data set of the influence factor data, determining the historical gas use data as an output soft target and / or output prediction gas data; modifying a loss function of the teacher model according to the output soft target, and training the teacher model; inputting data in the data set of the influence factor data into a pre-constructed knowledge distillation student model, taking the prediction gas data as a target value of the student model, and training the student model to obtain a required gas load prediction model.

2. The method of claim 1, wherein, The method further comprises the following steps: performing missing value filling on the historical raw data, and detecting outliers of the filled actual gas use data and the filled historical pipe storage data by a specified method; according to the detection result, replacing outliers of the filled historical actual gas use data and the filled historical pipe storage data to obtain preprocessed historical raw data; extracting a first data feature group from the preprocessed historical raw data according to a preset feature extraction algorithm; The historical raw data after the preprocessing is subjected to cluster analysis, and the historical raw data after the preprocessing is divided into multiple data groups; the data in each data group is subjected to feature analysis, and the common features in the group are determined, and the features between the multiple data groups are subjected to feature analysis, and the difference features between the groups are determined; The second data feature group is determined according to the common features in the group and the difference features between the groups; The data features of the historical raw data are determined according to the first data feature group and the second data feature group.

3. The method of claim 1, wherein, The use load prediction value is determined according to the first prediction value and the pipe storage use prediction value, and specifically includes: The ratio of the pipe storage use prediction value and the first prediction value is calculated to obtain a first ratio; The ratio of the historical pipe storage data and the historical actual use data is calculated to obtain a second ratio; If the difference between the first ratio and the second ratio is greater than a preset threshold, the first prediction value and the pipe storage use prediction value are weighted to obtain the use load prediction value.

4. The method of claim 1, wherein, The pipe storage data prediction model is trained by collecting the pipe pressure data in the target area, and the pipe storage use prediction value is determined, which includes: The pipe storage data set is constructed according to the historical pipe pressure value and the historical pipe storage data; The historical pipe pressure value in the pipe storage data set is input into the pre-constructed neural network model, the historical pipe storage data is taken as the output target of the neural network model, the model parameters are adjusted, and the model is trained to determine the required pipe storage data prediction model.

5. The method according to any one of claims 1 to 4, characterized in that, The use load is used to represent the natural gas usage in the target area.

6. An air charge estimation device characterized by comprising: The device includes: The data acquisition module is configured to acquire historical raw data in a target area, including historical actual use data of natural gas and historical pipe storage data of natural gas; The feature extraction module is configured to extract features from the historical raw data to obtain data features of the historical raw data; The feature determination module is configured to determine specified influence factor data related to the data features of the historical raw data according to the data features of the historical raw data; The data comparison module is configured to compare the specified influence factor data with historical actual influence factor data to obtain change amount data of the influence factor data; The determination module is configured to train prediction sample data and change amount data of the influence factor data by using a use load prediction model to determine a first prediction value of the use load, wherein the use load prediction model is a neural network model based on a knowledge distillation framework; pipe pressure data in the target area is collected, a pipe storage data prediction model is called to train the pipe pressure data, and a pipe storage use prediction value is determined; and a use load prediction value is determined according to the first prediction value and the pipe storage use prediction value. The feature determination module is further configured to acquire influence factor data corresponding to the historical raw data in the target region; perform dimensionless processing on the historical raw data and the influence factor data respectively, take the processed historical raw data as a control series, and take the processed influence factor data as a comparison series; calculate a correlation coefficient of the control series and the comparison series, and determine a correlation degree of the influence factor data and the historical raw data according to the correlation coefficient; sort the correlation degrees according to a preset rule, and take the influence factor data with a correlation degree higher than a preset threshold as specified influence factor data. The model determination module is further configured to collect a data set of historical gas consumption data and influence factor data; call data in the data set of influence factor data for training of a pre-constructed knowledge distillation teacher model, determine the historical gas consumption data as an output soft target and / or output predicted gas consumption data; modify a loss function of the teacher model according to the output soft target, and train the teacher model; input the data in the data set of influence factor data into a pre-constructed knowledge distillation student model, take the predicted gas consumption data as a target value of the student model, and train the student model to obtain a required gas consumption load prediction model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 5.

Images