Fusion index calculation method based on new energy consumption data

By using Gaussian hybrid model and BiLSTM network and other technical means in the processing of new energy consumption data, the problem of inaccurate calculation results in the existing technology is solved, and more efficient and accurate new energy is achieved Consumption prediction.

CN118916725APending Publication Date: 2024-11-08ELECTRIC POWER RES INST OF EAST INNER MONGOLIA ELECTRIC POWER +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410600704.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-05-15
Publication Date
2024-11-08

Smart Images

  • Figure CN118916725A_ABST
    Figure CN118916725A_ABST
Patent Text Reader

Abstract

The invention discloses a fusion index calculation method based on new energy consumption data, and belongs to the technical field of new energy consumption calculation, in the method, a space-time synchronization graph convolutional neural network performs high-dimensional feature extraction on a local space-time diagram, and fully mines a space-time association relationship of historical data. Mechanism knowledge stored in the new energy consumption knowledge graph is used for guiding the model, complex mathematical modeling and model solving are avoided, the solving speed is high, and the prediction precision is high; a feature processing module, a missing data identification module, an abnormal data identification module and an algorithm module are arranged in the preprocessing module, and mutation data in the data are identified by using a k-means algorithm; abrupt abnormal data may exist in the data set, and the original new energy consumption data needs to be standardized; according to the new energy consumption data clustering method, the optimal clustering center is selected, and the new energy consumption data of each month is subjected to clustering identification by selecting the proper number of the clustering centers, so that the accuracy and the calculation efficiency in the subsequent calculation process are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy consumption calculation, and specifically relates to a fusion index calculation method based on new energy consumption data. Background Art

[0002] With the continuous improvement of the national economic level and the continuous advancement of the "dual carbon" goals, the proportion of new energy in the power grid is increasing. The output of new energy in the regional power grid is characterized by randomness and strong volatility, and the load of the regional power grid is also random and volatile. When the output and load of new energy in the regional power grid fluctuate beyond the forecast, due to the minimum technical output constraints, ramp constraints, minimum operating time constraints, maintenance constraints of thermal power generating units, regional power grid power balance constraints, transmission section flow constraints, and energy storage unit output characteristics constraints, the total power of the overall regional power grid cannot be balanced, resulting in new energy consumption problems in the regional power grid.

[0003] However, common calculation methods cannot process and identify abnormal data and missing data in the reference data during the calculation process, which will cause the subsequent calculation results to be inaccurate and affect the overall calculation accuracy. Summary of the invention

[0004] The purpose of the present invention is to provide a method for calculating a fusion index based on new energy consumption data in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a method for calculating a fusion index based on new energy consumption data, comprising the following steps:

[0006] S1: Collection of advanced new energy consumption data: negotiate with the tax and power parties on industry types and internal content and determine the required content;

[0007] S2: Process the data on new energy consumption. Among the massive amount of new energy historical operation data, there is a large amount of data that affects the regional new energy consumption. Generally, with 15 minutes as a time section, the new energy historical operation data is sorted into time series data according to the corresponding time section, and the data is obtained from the new energy operation ledger. Clean up the new energy historical data, remove abnormal data and fill in missing values, and provide data support for the regional new energy consumption prediction algorithm;

[0008] S3: Then use the algorithm module to perform preprocessing data calculations, and use the algorithm module to cluster the samples generated by Gaussian mixture model clustering using the Gaussian mixture model. When clustering, the weights of each feature are manually assigned based on some historical data statistical results;

[0009] S4: Determine whether the captured data of the stored data in step S3 is equal to the captured data of the adjacent previous stored data. If they are equal, set the captured data of the current stored data to a null value. If they are not equal, retain the captured data of the current stored data and send the read stored data to the master station.

[0010] S5: Finally, the data processed in step S4 is completed;

[0011] S6: Then, construct the input feature matrix of the regional power grid new energy consumption prediction algorithm. The processed historical operation data of the power grid is spliced ​​and integrated to form a time series data input feature matrix and a non-time series data input feature matrix. At the same time, the time series feature matrix and the non-time series feature matrix are substituted into the BiLSTM network model for training. According to whether there is a new energy consumption problem in the historical time series regional power grid, a label matrix is ​​constructed, and the model weight is continuously updated to improve the model accuracy;

[0012] S7: Finally, the new energy forecast value, load forecast value, and corresponding power grid ledger data within the forecast period are used to forecast new energy for the regional power grid, and the new energy consumption situation of the regional power grid within the evaluation period can be obtained;

[0013] S8: Display the data results: The final analysis and calculation results are tabulated and displayed to complete the entire fusion index calculation process based on new energy consumption data.

[0014] In a preferred embodiment, in step S2, missing data identification is also required. Missing data generally exists in the form of null values ​​in the building operation database. The null value format cannot be directly identified by R statements, and the null value needs to be converted into a recognizable "NA" format; the step of identifying missing data in the data set is mainly divided into two steps. First, the building new energy consumption data set in the database is imported into an Excel file, and the file format is modified to a .csv or other format file; secondly, the R language identifies the path to save the file, imports the data of the file into the R language environment, and identifies the missing data in the data set through the is.na function.

[0015] In a preferred embodiment, in step S2, in order to more clearly express the number and proportion of missing data in the entire data set, the table function counts the number of data contained in the missing and non-missing data, and the aggr function draws a missing data proportion chart; then abnormal data identification is performed. When abnormal data identification is performed, continuous repeated data is first identified, and the same data in each month's new energy consumption data is first identified, and continuous and non-continuous identical data exist in these data; then it is determined whether long-term continuous repeated data appears in these identical data; then the mutation data is identified; since there may be missing data in each month's new energy consumption data, the missing data must be deleted first.

[0016] In a preferred embodiment, in step S3, the collected data is conditionally judged to identify null values ​​and negative values ​​and the null values ​​are removed, and then the processed data is subjected to the quantile statistics method to identify abnormal values ​​and remove the null values, wherein the specific steps of the quantile statistics method are as follows: S1, Q1, Q2, Q3, Q4 represent four quartile values ​​arranged in ascending order, respectively; S2, let IQR = Q3-Q1; S3, let the threshold = average value + 20*IQR, if the electricity consumption is higher than the threshold, it is judged as an abnormal value and removed.

[0017] In a preferred embodiment, in step S3, the Gaussian mixture model uses the Gaussian probability density function (normal distribution curve) to accurately quantify things. It is a model that decomposes things into several models based on the Gaussian probability density function (normal distribution curve); the process of clustering using the Gaussian mixture model is the "inverse process" of generating data samples using the Gaussian mixture model: given the number of clusters k, through a given data set, using a certain parameter estimation method, deriving the parameters of each mixed component; mean vector μ, covariance matrix Σ and mixing coefficient α; each multivariate Gaussian distribution component corresponds to a cluster after clustering; the advantage of using the Gaussian mixture model for clustering is that after projection, the sample points do not obtain a definite classification label, but obtain the probability of each class, which is important information; the Gaussian mixture model can be used not only for clustering, but also for probability density estimation.

[0018] In a preferred embodiment, in step S5, the completion of real-time new energy consumption data is as follows: if the 10 adjacent values ​​before and after the quasi-real-time data show regular fluctuations, that is, the absolute values ​​of the fluctuations of each two adjacent points are the same, the data is completed according to the fluctuation law; if the adjacent values ​​do not show regular fluctuations, the generated average electricity consumption value of the same type of day and the same electricity consumption time is used for completion; the completion of daily electricity consumption data is as follows: if the absolute values ​​of the adjacent gaps of the 4 adjacent values ​​before and after the daily data are all less than 0.1, it means that the electricity consumption is stable, and the adjacent value averaging method is used for completion; if the absolute values ​​of the gaps between the adjacent values ​​before and after are unstable, the generated average electricity consumption value of the same type of day is used for completion.

[0019] In a preferred embodiment, in step S6, a BiLSTM network model is used to predict the consumption of new energy in the regional power grid. BiLSTM is a bidirectional LSTM, which is mainly a combination of forward and backward LSTM network models.

[0020] In a preferred embodiment, in step S6, the BiLSTM model can better extract power grid time series data information compared with the LSTM network model. LSTM is an improvement on the traditional recurrent neural network model, adding a memory module and a forgetting module to solve the problems of gradient explosion and gradient disappearance in deep learning.

[0021] In a preferred embodiment, in step S7, the data is compared and matched according to the corresponding identification bits based on the contents provided by both parties; the data comparison module is used to collect standard signals and compare them with the voltage and current signals collected by the data acquisition module.

[0022] In a preferred embodiment, in step S7, the new energy data comparison module also includes one or more groups of standard voltage transformers PT and current transformers CT to collect standard voltage and current data for comparison with the data collected by the data collection module at the same time.

[0023] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0024] In the present invention, the spatiotemporal synchronous graph convolutional neural network extracts high-dimensional features from the local spatiotemporal graph, and fully explores the spatiotemporal correlation of historical data. The model is guided by the mechanism knowledge stored in the new energy consumption knowledge graph, which avoids complex mathematical modeling and model solving, and has fast solving speed and high prediction accuracy; the preprocessing module is internally provided with a feature processing module, a missing data identification module, an abnormal data identification module and an algorithm module, so that the data used for index calculation of the new energy consumption data fusion has been processed, so that the abnormal data and missing data in the reference data can be processed and identified, and the mutation data in the data can be identified by using the k-means algorithm; there may be mutation abnormal data in the data set, and the original new energy consumption data must be standardized; then the best clustering center is selected, and the new energy consumption data of each month is clustered and identified by selecting the appropriate number of clustering centers, so that the mutation data can be accurately identified; thereby improving the accuracy and calculation efficiency in the subsequent calculation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the process principle of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0027] Reference Figure 1 ,

[0028] A method for calculating a fusion index based on new energy consumption data includes the following steps:

[0029] S1: Collection of advanced new energy consumption data: negotiate with the tax and power parties on industry types and internal content and determine the required content;

[0030] S2: Process the data on new energy consumption. Among the massive amount of new energy historical operation data, there is a large amount of data that affects the regional new energy consumption. Generally, with 15 minutes as a time section, the new energy historical operation data is sorted into time series data according to the corresponding time section, and the data is obtained from the new energy operation ledger. Clean up the new energy historical data, remove abnormal data and fill in missing values, and provide data support for the regional new energy consumption prediction algorithm;

[0031] S3: Then use the algorithm module to perform preprocessing data calculations, and use the algorithm module to cluster the samples generated by Gaussian mixture model clustering using the Gaussian mixture model. When clustering, the weights of each feature are manually assigned based on some historical data statistical results;

[0032] S4: Determine whether the captured data of the stored data in step S3 is equal to the captured data of the adjacent previous stored data. If they are equal, set the captured data of the current stored data to a null value. If they are not equal, retain the captured data of the current stored data and send the read stored data to the master station.

[0033] S5: Finally, the data processed in step S4 is completed;

[0034] S6: Then, construct the input feature matrix of the regional power grid new energy consumption prediction algorithm. The processed historical operation data of the power grid is spliced ​​and integrated to form a time series data input feature matrix and a non-time series data input feature matrix. At the same time, the time series feature matrix and the non-time series feature matrix are substituted into the BiLSTM network model for training. According to whether there is a new energy consumption problem in the historical time series regional power grid, a label matrix is ​​constructed, and the model weight is continuously updated to improve the model accuracy;

[0035] S7: Finally, the regional power grid new energy forecast is performed using the new energy forecast value, load forecast value, and the corresponding power grid ledger data within the forecast period, and the new energy consumption situation of the regional power grid within the evaluation period can be obtained;

[0036] S8: Display the data results: The final analysis and calculation results are tabulated and displayed to complete the entire fusion index calculation process based on new energy consumption data.

[0037] In step S2, missing data identification is also required. Missing data generally exists in the form of null values ​​in the building operation database. The null value format cannot be directly identified through R statements, and the null value needs to be converted into a recognizable "NA" format; the steps to identify missing data in the dataset are mainly divided into two steps. First, the building new energy consumption dataset in the database is imported into an Excel file, and the file format is modified to a .csv or other format file; secondly, the R language identifies the path to save the file, imports the data of the file into the R language environment, and identifies the missing data in the dataset through the is.na function.

[0038] In step S2, in order to more clearly express the number and proportion of missing data in the entire data set, the table function counts the number of data contained in the missing and non-missing data, and the aggr function draws a graph of the proportion of missing data; then abnormal data identification is performed. When performing abnormal data identification, continuous repeated data is first identified, and the same data in each month's new energy consumption data is first identified. These data contain continuous and non-continuous identical data; then it is determined whether long-term continuous repeated data appear in these identical data; then the mutation data is identified; since there may be missing data in each month's new energy consumption data, the missing data must be deleted first.

[0039] In step S3, the collected data is conditionally judged to identify null values ​​and negative values ​​and the null values ​​are removed, and then the processed data is subjected to the quantile statistical method to identify abnormal values ​​and remove the null values. The specific steps of the quantile statistical method are as follows: S1. Use Q1, Q2, Q3, and Q4 to represent the four quartile values ​​in ascending order; S2. Let IQR = Q3-Q1; S3. Let the threshold = average value + 20*IQR. If the electricity consumption is higher than the threshold, it is judged as an abnormal value and removed.

[0040] In step S3, the Gaussian mixture model uses the Gaussian probability density function (normal distribution curve) to accurately quantify things. It is a model that decomposes things into several models based on the Gaussian probability density function (normal distribution curve); the process of clustering using the Gaussian mixture model is the "inverse process" of generating data samples using the Gaussian mixture model: given the number of clusters k, through a given data set, using a certain parameter estimation method, derive the parameters of each mixed component; mean vector μ, covariance matrix Σ and mixing coefficient α; each multivariate Gaussian distribution component corresponds to a cluster after clustering; the advantage of using the Gaussian mixture model for clustering is that after projection, the sample points do not get a definite classification label, but the probability of each class, which is important information; the Gaussian mixture model can be used not only for clustering, but also for probability density estimation.

[0041] In step S5, the completion of real-time new energy consumption data is as follows: if the 10 adjacent values ​​before and after the quasi-real-time data show regular fluctuations, that is, the absolute values ​​of the fluctuations of each two adjacent points are the same, the data is completed according to the fluctuation law; if the adjacent values ​​do not show regular fluctuations, the generated average electricity consumption value of the same type of days and the same electricity consumption time is used for completion; the completion of daily electricity consumption data is as follows: if the absolute values ​​of the adjacent 4 values ​​before and after the daily data are less than 0.1, it means that the electricity consumption is stable, and the adjacent value averaging method is used for completion; if the absolute values ​​of the adjacent value differences are unstable, the generated average electricity consumption value of the same type of days is used for completion.

[0042] In step S6, the BiLSTM network model is used to predict the consumption of new energy in the regional power grid. BiLSTM is a bidirectional LSTM, which is mainly a combination of forward and backward LSTM network models. Compared with the LSTM network model, the BiLSTM model can better extract power grid time series data information. LSTM is an improvement on the traditional recurrent neural network model, adding a memory module and a forgetting module, which solves the problems of gradient explosion and gradient disappearance in deep learning.

[0043] In step S7, the data is compared and matched according to the corresponding identification bits based on the contents provided by both parties; the data comparison module is used to collect standard signals and compare them with the voltage and current signals collected by the data acquisition module; the new energy data comparison module also includes one or more groups of standard voltage transformers PT and current transformers CT to collect standard voltage and current data for comparison with the data collected by the data acquisition module at the same time.

[0044] In the present invention, the spatiotemporal synchronous graph convolutional neural network extracts high-dimensional features from the local spatiotemporal graph, and fully explores the spatiotemporal correlation of historical data. The model is guided by the mechanism knowledge stored in the new energy consumption knowledge graph, which avoids complex mathematical modeling and model solving, and has fast solving speed and high prediction accuracy; the preprocessing module is internally provided with a feature processing module, a missing data identification module, an abnormal data identification module and an algorithm module, so that the data used for index calculation of the new energy consumption data fusion has been processed, so that the abnormal data and missing data in the reference data can be processed and identified, and the mutation data in the data can be identified by using the k-means algorithm; there may be mutation abnormal data in the data set, and the original new energy consumption data must be standardized; then the best clustering center is selected, and the new energy consumption data of each month is clustered and identified by selecting the appropriate number of clustering centers, so that the mutation data can be accurately identified; thereby improving the accuracy and calculation efficiency in the subsequent calculation process.

[0045] It should be noted that, in this article, relational terms such as first and second, etc. 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.

[0046] The above description enables those skilled in the art to implement or use 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 will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for calculating a fusion index based on new energy consumption data, characterized in that: The steps include: S1: Collection of advanced new energy consumption data: negotiate the industry type and content with the tax and power parties and determine the required content; S2: Process the data on new energy consumption. Among the massive amount of new energy historical operation data, there is a large amount of data that affects the regional new energy consumption. Generally, with 15 minutes as a time section, the new energy historical operation data is sorted into time series data according to the corresponding time section, and the data is obtained from the new energy operation ledger. The new energy historical data is cleaned, abnormal data is removed and missing values ​​are filled, providing data support for the regional new energy consumption prediction algorithm. S3: Then use the algorithm module to perform preprocessing data calculations, and use the algorithm module to cluster the samples generated by Gaussian mixture model clustering using the Gaussian mixture model. When clustering, the weights of each feature are manually assigned based on some historical data statistical results; S4: Determine whether the captured data of the stored data in step S3 is equal to the captured data of the adjacent previous stored data. If they are equal, set the captured data of the current stored data to a null value. If they are not equal, retain the captured data of the current stored data and send the read stored data to the master station. S5: Finally, the data processed in step S4 is completed; S6: Then, construct an input feature matrix of the regional power grid new energy consumption prediction algorithm; splice and integrate the processed power grid historical operation data to form a time series data input feature matrix and a non-time series data input feature matrix; At the same time, the time series feature matrix and the non-time series feature matrix are substituted into the BiLSTM network model for training; a label matrix is ​​constructed based on whether there is a problem of new energy consumption in the regional power grid in the historical time series, and the model weights are continuously updated to improve the model accuracy; S7: Finally, the new energy forecast value, load forecast value, and corresponding power grid ledger data within the forecast period are used to forecast new energy for the regional power grid, and the new energy consumption situation of the regional power grid within the evaluation period can be obtained; S8: Display the data results: The final analysis and calculation results are tabulated and displayed to complete the entire fusion index calculation process based on new energy consumption data.

2. A method for calculating a fusion index based on new energy consumption data according to claim 1, characterized in that: In step S2, missing data identification is also required. Missing data generally exists in the form of null values ​​in the building operation database. The null value format cannot be directly identified by R statements, and the null value needs to be converted into a recognizable "NA" format. The step of identifying missing data in the data set is mainly divided into two steps. First, the building new energy consumption data set in the database is imported into an Excel file, and the file format is modified to a .csv or other format file; Secondly, R language identifies the path where the file is saved, imports the data of the file into the R language environment, and uses the is.na function to identify missing data in the data set.

3. A method for calculating a fusion index based on new energy consumption data according to claim 1, characterized in that: In the step S2, in order to more clearly express the number and proportion of missing data in the entire data set, the table function counts the number of data contained in the missing and non-missing data, and the aggr function draws a missing data proportion chart; then abnormal data identification is performed. When abnormal data identification is performed, continuous repeated data is first identified, and the same data in each month's new energy consumption data is first identified. These data contain continuous and non-continuous identical data; then it is determined whether long-term continuous repeated data appears in these identical data; then the mutation data is identified; since there may be missing data in each month's new energy consumption data, the missing data must be deleted first.

4. A method for calculating a fusion index based on new energy consumption data according to claim 1, characterized in that: In step S3, the collected data is conditionally judged to identify null values ​​and negative values ​​and the null values ​​are removed, and then the processed data is subjected to the quantile statistics method to identify abnormal values ​​and remove the null values, wherein the specific steps of the quantile statistics method are as follows: S1, Q1, Q2, Q3, Q4 represent four quartile values ​​arranged in ascending order respectively; S2, IQR=Q3-Q1; S3, threshold=average value+20*IQR, if the electricity consumption is higher than the threshold, it is judged as an abnormal value and removed.

5. The method for calculating the fusion index based on new energy consumption data according to claim 1, characterized in that: In step S3, the Gaussian mixture model uses the Gaussian probability density function (normal distribution curve) to accurately quantify things. It is a model that decomposes things into several models based on the Gaussian probability density function (normal distribution curve); the process of clustering using the Gaussian mixture model is the "inverse process" of generating data samples using the Gaussian mixture model: given the number of clusters k, through a given data set, a certain parameter estimation method is used to derive the parameters of each mixed component; mean vector μ, covariance matrix Σ and mixing coefficient α; each multivariate Gaussian distribution component corresponds to a cluster after clustering; the advantage of using the Gaussian mixture model for clustering is that after projection, the sample points do not obtain a definite classification label, but obtain the probability of each class, which is important information; the Gaussian mixture model can be used not only for clustering, but also for probability density estimation.

6. A method for calculating a fusion index based on new energy consumption data according to claim 1, characterized in that: In step S5, the completion of the real-time new energy consumption data is as follows: if the 10 adjacent values ​​before and after the quasi-real-time data show regular fluctuations, that is, the absolute values ​​of the fluctuations of each two adjacent points are the same, the data is completed according to the fluctuation law; if the adjacent values ​​do not show regular fluctuations, the generated average power consumption value of the same power consumption time on the same day is used for completion; the completion of the daily power consumption data is as follows: if the absolute values ​​of the adjacent gaps of the 4 adjacent values ​​before and after the daily data are all less than 0.1, it means that the power consumption is stable, and the adjacent value averaging method is used for completion; if the absolute values ​​of the gaps between the adjacent values ​​before and after are unstable, the generated average power consumption value of the same day is used for completion.

7. A method for calculating a fusion index based on new energy consumption data according to claim 1, characterized in that: In step S6, a BiLSTM network model is used to predict the consumption of new energy in the regional power grid; BiLSTM is a bidirectional LSTM, which is mainly a combination of forward and backward LSTM network models.

8. The method for calculating the fusion index based on new energy consumption data according to claim 1, characterized in that: In step S6, compared with the LSTM network model, the BiLSTM model can better extract power grid time series data information; and LSTM is an improvement on the traditional recurrent neural network model, adding a memory module and a forgetting module, which solves the problems of gradient explosion and gradient disappearance in deep learning.

9. The method for calculating the fusion index based on new energy consumption data according to claim 1, characterized in that: In step S7, the data is compared and matched according to the corresponding identification bits based on the contents provided by both parties; the data comparison module is used to collect standard signals and compare them with the voltage and current signals collected by the data acquisition module.

10. The method for calculating the fusion index based on new energy consumption data according to claim 1, characterized in that: In the step S7, the new energy data comparison module also includes one or more groups of standard voltage transformers PT and current transformers CT to collect standard voltage and current data for comparison with the data collected by the data collection module at the same time.