Industrial power consumption data classification method considering energy consumption characteristics

Abstract

The invention relates to the technical field of data classification, and discloses an industrial power consumption data classification method considering energy consumption characteristics, which comprises the steps of data acquisition, preprocessing, feature extraction, data classification and data integration, data acquisition covers multiple parameters, the data source is wide, and the applicability is high. In the preprocessing link, through data marking, missing value filling and abnormal value replacement, the data accuracy and integrity are guaranteed, and a foundation is built for subsequent analysis; in feature extraction, multiple feature values are determined from the dimensions of power, electricity consumption, electricity consumption time and the like, and electricity consumption features are described in multiple directions; according to data classification, the characteristic value is compared with a preset threshold value, and multi-dimensional judgment is combined, so that classification is more accurate, and the actual power utilization condition of the industry is met; the data is integrated to form a comprehensive classification result, which is convenient for direct application in the aspects of power planning, industrial layout and the like. On the whole, the method is helpful to accurately grasp the industrial electricity utilization rule, and provides a scientific and effective basis for reasonable allocation of electric power resources, electricity utilization management and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of data classification, and in particular to a method for classifying industrial electricity data taking energy consumption characteristics into consideration. Background Art

[0002] With the acceleration of industrialization and the continued growth of energy demand, the rational allocation and effective management of power resources across various industries have become critical issues that require urgent resolution. Accurately understanding and classifying the power consumption characteristics of different industries is crucial for optimizing power resource allocation, formulating differentiated energy policies, ensuring the stable operation of power systems, and promoting sustainable industrial development.

[0003] In traditional industrial electricity management, there is often a lack of refined analysis and classification methods for industrial electricity data. On the one hand, the data collection process may be irregular, resulting in uneven data quality, such as inconsistent collection frequency, incomplete data records, and a large number of missing values ​​and outliers. These problems seriously affect the subsequent accurate judgment of industrial electricity characteristics. On the other hand, previous analysis methods are mostly relatively simple, focusing only on the overall scale of electricity consumption or simple time distribution. They fail to comprehensively consider industrial electricity characteristics from multiple dimensions such as power fluctuations, peak and valley changes in electricity consumption, and fluctuations in electricity consumption time. They are unable to deeply explore the subtle differences in electricity consumption behavior among different industries, and it is difficult to meet the requirements of modern power management for refined classification and precise regulation of industrial electricity consumption.

[0004] At the same time, with the development of power monitoring technology, power monitoring systems and smart meters and other equipment have been widely used in the field of industrial power consumption. These devices can collect massive amounts of power consumption data, providing a rich data basis for in-depth analysis of industrial power consumption characteristics. However, how to make full use of these data resources and develop a scientific, efficient and comprehensive industrial power consumption data classification method has become a hot topic and difficulty in current research in the power field. It is based on this background that the present invention proposes a method for classifying industrial power consumption data that takes energy consumption characteristics into consideration. It aims to effectively solve the problems existing in traditional methods through systematic data collection, preprocessing, feature extraction, classification and integration, and realize accurate classification of industrial power consumption data, providing strong support for the scientific management and rational utilization of power resources. Summary of the Invention

[0005] The purpose of the present invention is to provide an industrial electricity data classification method that takes energy consumption characteristics into consideration, thereby solving the technical problems raised in the background technology.

[0006] The purpose of the present invention can be achieved through the following technical solutions: A method for classifying industrial electricity consumption data taking into account energy consumption characteristics includes: Step 1: Data Collection: Collect industrial electricity consumption data, including data parameters corresponding to power, electricity consumption, and electricity consumption time; Step 2: Preprocessing: Perform data labeling, missing value filling, and outlier replacement processing on industrial electricity data, and obtain data sequences corresponding to different data parameters after missing value filling and outlier replacement based on the processing results; Step 3: Feature extraction: Based on the pre-processed multiple data sequences, feature extraction processing is performed on each data sequence, and multiple feature values ​​of the industrial electricity data are determined; Step 4: Data classification: Perform classification analysis based on multiple feature values ​​extracted from features, and then determine the classification results of industrial electricity data through classification analysis; Step 5: Data integration: The classification results based on data classification are integrated to form a comprehensive industrial electricity data classification result, which is then stored in the database.

[0007] As a further solution of the present invention: wherein, the data source for collecting industrial electricity consumption data is the power monitoring system and the equipment corresponding to the smart meter.

[0008] As a further solution of the present invention: the data marking processing method is as follows: In a specified period, it is divided into several standard time periods with uniform time intervals. Then the power, power consumption and power consumption time collected in each standard time period are respectively formed into corresponding data sequences, and they are recorded as power data sequences P={P t}、Electricity consumption data series Y={Y t}, electricity consumption time data series S={S t}; Where t is the standard period number in the specified period, t=1, 2, ... n, and n represents the number of collection time nodes in the specified period; The power parameter in the power data sequence is: in a standard period, a number of power values ​​are obtained according to a predetermined acquisition frequency, and then the average value of the several power values ​​is used as the power parameter of the standard period; The power consumption parameters in the power consumption data series are: the power consumption between the beginning and end of the standard period; The power consumption time parameter in the power consumption time data series is: the power consumption time between the start and end stages of the standard period.

[0009] As a further solution of the present invention: the missing value filling processing method is as follows: Within a specified period, statistics are performed to determine whether each data series contains missing values ​​in each standard time period, and then the missing values ​​are filled.

[0010] As a further solution of the present invention: the missing value statistics method is: in the corresponding data sequence, if the t-th standard time period does not contain numerical data, it means that the t-th standard time period in the corresponding data sequence contains missing values.

[0011] As a further solution of the present invention: the filling process is as follows: Taking power data as an example, in the power data series, when the tth standard period contains missing values; Then obtain the power parameters of the standard period adjacent to the standard period, that is, P t-1 and P t+1 ; Among them, P t-1 represents the power parameter in the previous standard period adjacent to the tth standard period, P t+1 represents the power parameter in the next standard period adjacent to the t-th standard period; Afterwards via: Calculate the filling value PT in the tth standard period in the power data sequence t , fill value PT t Used to supplement the power data sequence and serve as the power parameter P in the tth standard period t use.

[0012] As a further solution of the present invention: wherein, in the power data sequence, when the value of t in the missing value Pt is 1 or n; Then by:

[0013] Calculate the fill value PT in the first or nth standard period in the power data sequence t ; Its formula is: When t=1, the average value of the power parameters in the three standard periods adjacent to the t-th standard period is calculated as the filling value in the first standard period; When t=n, the average value of the power parameters in the first three standard time periods adjacent to the t-th standard time period is calculated as the filling value in the n-th standard time period.

[0014] As a further solution of the present invention: the abnormal value replacement processing method is as follows: Select the power data, calculate the mean and standard deviation of all power parameters in the power data series, and mark them as P p and P b ; Then, each power parameter in the power data sequence is combined with its corresponding mean value and standard deviation to determine the outliers: If P t >P p +3×P b and Pt>P p +3×P b If at least one of the following is true, the corresponding power parameter P t Determined as an outlier; If P p +3×P b ≤P t ≤P p +3×P b , then the corresponding power parameter P t Determined as an outlier; Then the power parameter P will be determined as an outlier. t Make data corrections; According to the outlier replacement processing method of the power data series, outlier replacement processing is performed on other data series.

[0015] As a further solution of the present invention: the data correction method is as follows: In the power data series, obtain all power parameters that are not outliers, then calculate their average value and mark it as P1 p ; Then its average value is used as the outlier corresponding to all power parameters P t Corrected value.

[0016] As a further solution of the present invention: the specific method of feature extraction is as follows: Step 2.1: Extract the power data sequence P t ; Then through:

[0017] Calculate the variance BD corresponding to the power parameter within the specified period P , and use it as the load fluctuation characteristic value; Among them, P p is the mean of all power parameters in the power data series; Step 2.2: Extract the electricity consumption data sequence Y t ; Then, the power consumption parameter with the largest value is extracted from the power consumption data sequence and recorded as Ymax; At the same time, the power consumption parameter with the minimum value is extracted from the power consumption data sequence and recorded as Ymin; Then through: C Y =Ymax-Ymin Calculate the peak-valley difference C of electricity consumption within a specified period Y , and use it as the peak-valley characteristic value of electricity consumption; Also through:

[0018] Calculate the average value Y corresponding to the power consumption parameter within the specified period P , Then the power consumption parameters of each standard period are compared with Y p Compare and assign the digital "0" or "1" to the power consumption parameter of each standard period according to the comparison result, and record the assignment result as F t ; The method is as follows: When Y t ≥Y p , then let F t The value of is 1, that is, the power consumption parameter of the corresponding standard period is assigned by the number "1"; When Y t <Y p , then let F t The value of is 0, that is, the power consumption parameter of the corresponding standard period is assigned with the number "0"; This process is repeated until the results of all standard time periods are determined. Afterwards through:

[0019] Calculate the ratio R of the number of standard periods where the power consumption parameter exceeds the average value of all power consumption parameters to the total number of standard periods Y , and record it as the second peak-valley characteristic value of electricity consumption; Step 2.2: Extract the electricity consumption time data sequence S t ; Then through:

[0020] Calculate the variance BS corresponding to the power consumption time parameter within the specified period P , and use it as the characteristic value of electricity consumption time fluctuation; Among them, S p is the mean of all power consumption time parameters in the power data series.

[0021] As a further solution of the present invention: the specific method of data classification is as follows: Step 3.1: Load fluctuation characteristic value BD P BDY with pre-set load fluctuation characteristic classification threshold P For comparison: If BD P > (1 + γ) × BDYP , then the industry is classified as a high-fluctuation-load industry; If (1+γ)×BDY P ≤BD P ≤(1+γ)×BDY P , then the industry is classified as a medium-fluctuating load industry; If BD P <(1-γ)×BDY P , then the industry is classified as a low-fluctuation-load industry; Wherein, γ is a preset threshold correction coefficient, wherein γ<1; Step 3.2 The peak-valley characteristic value of electricity consumption is C Y and the pre-set peak-to-valley difference feature classification threshold CY Y Compare and compare the peak and valley characteristic values ​​of electricity consumption Y Compared with the pre-set peak-to-valley ratio feature classification threshold RY Y For comparison: If C Y >CY Y and R Y >RY Y If all the contrast formulas in the formula are established, the industry is classified as a high peak-valley difference industry; If C Y >CY Y and R Y >RY Y If only one of the contrasting formulas is established, the industry is classified as a medium peak-valley difference industry; If C Y >CY Y and R Y >RY Y If all the contrast formulas in the formula do not hold true, the industry will be classified as a low peak-valley difference industry; Step 3.3: Set the power consumption time fluctuation characteristic value BS P The pre-set power consumption time fluctuation characteristic classification threshold BSY P For comparison: If BS P >(1+γ)×BSY P , then the industry is classified as a high-fluctuation electricity consumption industry; If (1+γ)×BSY P ≤BS P ≤(1+γ)×BSY P , then the industry is classified as a medium-fluctuating electricity consumption industry; If BS P <(1-γ)×BSY P, then the industry is classified as a low-fluctuation electricity consumption time industry.

[0022] As a further solution of the present invention: the specific method of data classification is as follows: Step 4.1: Preliminary clustering based on K-Means++ algorithm First, for multiple industrial users, the multiple eigenvalues ​​obtained in the feature extraction step are combined into a feature vector, Xi=[BD P，i 、C Y，i 、R Y，i ,BS P，i ]; Where i = 1, 2, ..., m, m represents the number of industrial users; Subsequently, the K-Means++ algorithm is used to perform preliminary clustering of industry users in the following manner: First, randomly select a data point as the first cluster center; Then, for each data point Xi, calculate its shortest distance to the selected cluster center, and select the data point with the largest shortest distance as the next cluster center; Repeat this process until k cluster centers are selected, where the value of k is estimated based on the approximate number of industry types. Then, the data points are assigned to the cluster with the smallest distance based on the shortest distance between the data points and the cluster center; The distance calculation formula is:

[0023] Where, is the shortest distance between the data point and the cluster center, and g = 1, 2, ... v, v is the dimension of the feature vector, j = 1, 2, ... k, C j is the jth cluster center, C j，g is the gth eigenvector in the jth cluster center, X i，g is the gth feature vector within the i-th industry user; Step 4.2: Secondary classification and adjustment based on hierarchical clustering Based on the clustering results of K-Means++ algorithm, hierarchical clustering algorithm is used for secondary classification and adjustment; First, consider each K-Means++ cluster as an initial cluster and calculate the similarity between clusters; the specific method is: The average linkage method is used to calculate the similarity between clusters. Two clusters are selected and recorded as A and B respectively. Then:

[0024] Calculate the similarity S(A, B) between A and B; Where d(x, y) represents the distance between two data points x and y, x represents the nth data point in cluster A, and y represents the nth data point in cluster A, which is obtained by the Euclidean distance formula. |A| and |B| represent the number of data points in cluster A and cluster B, respectively. Then, the cluster pairs with the highest similarity are merged and the cluster hierarchy is gradually constructed until it reaches the corresponding preset number of categories or the similarity between clusters is lower than the corresponding preset threshold.

[0025] Beneficial effects of the present invention: Accurate Classification: Through multi-dimensional feature extraction and a meticulous classification process, industrial electricity consumption data can be accurately classified. Taking into account data parameters such as power, electricity consumption, and electricity consumption time, multiple features are extracted, including load fluctuation characteristic values, peak-valley characteristic values, and electricity consumption time fluctuation characteristic values. Classification is performed based on the comparison of these characteristics with preset thresholds and advanced clustering algorithms. This effectively distinguishes industries with different electricity consumption characteristics, such as those with high load fluctuations, high peak-valley differences, and high fluctuations in electricity consumption time, providing an accurate basis for the refined management of power resources.

[0026] Improved Data Integrity and Accuracy: During the data preprocessing phase, data labeling, missing value filling, and outlier replacement were performed. Data labeling clarified the data structure and facilitated subsequent analysis. Missing value filling effectively addressed missing data by calculating the appropriate mean of adjacent data or the mean of multiple adjacent data in specific locations, ensuring the integrity of the data sequence. Outlier replacement, through rigorous outlier identification and correction, improved data accuracy, prevented abnormal data from interfering with classification results, and thus enhanced the reliability of the entire classification method.

[0027] Adaptability and Flexibility: The preset threshold correction coefficient γ can be adjusted according to the characteristics of different regions and industrial structures, allowing the classification method to adapt to the needs of various scenarios. Furthermore, the intelligent selection of cluster centers in the K-Means++ algorithm and the calculation of inter-cluster similarity based on the average linkage method in the hierarchical clustering algorithm, as well as the flexible construction of cluster hierarchies based on preset conditions, make this method highly adaptable to users of different sizes and types, enabling reasonable classification based on actual conditions, and applicable to both large-scale industrial and small-scale commercial power users.

[0028] Efficient data processing process: The entire process from data collection to data integration is well-designed and efficient. Data collection comes from common power monitoring systems, smart meters, and other equipment, which are widely available and easy to obtain. Each step is closely connected. For example, feature extraction is based on preprocessing results, data classification is based on feature extraction results, and data integration effectively organizes and stores classification results. This enables the entire method to quickly and stably process large amounts of industrial electricity data, meeting the requirements of modern power management for data processing timeliness. This helps power departments to keep abreast of industrial electricity consumption and make reasonable resource allocation and management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below with reference to the accompanying drawings.

[0030] Figure 1 This is a system block diagram of an industrial electricity data classification method that takes energy consumption characteristics into consideration according to the present invention. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention. Example

[0032] See also Figure 1 As shown, the present invention is a method for classifying industrial electricity data taking into account energy consumption characteristics, including: Step 1: Data Collection: Collect industrial electricity consumption data, including data parameters corresponding to power, electricity consumption, and electricity consumption time; Among them, the data sources for collecting industrial electricity consumption data are power monitoring systems, smart meters and other equipment; Step 2: Preprocessing: Data marking is performed on the data parameters corresponding to power, power consumption, and power consumption time in industrial power consumption data: The specific method is as follows: In a specified period, it is divided into several standard time periods with uniform time intervals. Then the power, power consumption and power consumption time collected in each standard time period are respectively formed into corresponding data sequences, and they are recorded as power data sequences P={P t}、Electricity consumption data series Y={Y t}, electricity consumption time data series S={S t}; Where t is the standard period number in the specified period, t=1, 2, ... n, and n represents the number of collection time nodes in the specified period; The power parameter in the power data sequence is: in a standard period, a number of power values ​​are obtained according to a predetermined acquisition frequency, and then the average value of the several power values ​​is used as the power parameter of the standard period; The power consumption parameters in the power consumption data series are: the power consumption between the beginning and end of the standard period; The electricity usage time parameters in the electricity usage time data series are: the electricity usage time between the start and end of the standard period; Step 3: Feature extraction: Step 2.1: Extract the power data sequence P t ; Then through:

[0033] Calculate the variance BD corresponding to the power parameter within the specified period P , and use it as the load fluctuation characteristic value; Among them, P p is the mean of all power parameters in the power data series; Step 2.2: Extract the electricity consumption data sequence Y t ; Then, the power consumption parameter with the largest value is extracted from the power consumption data sequence and recorded as Ymax; At the same time, the power consumption parameter with the minimum value is extracted from the power consumption data sequence and recorded as Ymin; Then through: C Y =Ymax-Ymin Calculate the peak-valley difference C of electricity consumption within a specified period Y , and use it as the peak-valley characteristic value of electricity consumption; Also through:

[0034] Calculate the average value Y corresponding to the power consumption parameter within the specified period P , Then the power consumption parameters of each standard period are compared with Y p Compare and assign the digital "0" or "1" to the power consumption parameter of each standard period according to the comparison result, and record the assignment result as F t ; The method is as follows: When Y t ≥Y p , then let F t The value of is 1, that is, the power consumption parameter of the corresponding standard period is assigned by the number "1"; When Y t <Y p , then let F t The value of is 0, that is, the power consumption parameter of the corresponding standard period is assigned with the number "0"; This process is repeated until the results of all standard time periods are determined. Afterwards through:

[0035] Calculate the ratio R of the number of standard periods where the power consumption parameter exceeds the average value of all power consumption parameters to the total number of standard periods Y , and record it as the second peak-valley characteristic value of electricity consumption; Step 2.2: Extract the electricity consumption time data sequence S t ; Then through:

[0036] Calculate the variance BS corresponding to the power consumption time parameter within the specified period P , and use it as the characteristic value of electricity consumption time fluctuation; Among them, S p is the mean of all power consumption time parameters in the power data series; Step 4: Data classification: Perform classification analysis based on multiple feature values ​​extracted from features, and then determine the classification results of industrial electricity data through classification analysis; The classification analysis method is as follows: Step 3.1: Load fluctuation characteristic value BD P BDY with pre-set load fluctuation characteristic classification threshold P For comparison: If BD P > (1 + γ) × BDY P , then the industry is classified as a high-fluctuation-load industry; If (1+γ)×BDY P ≤BD P ≤(1+γ)×BDY P , then the industry is classified as a medium-fluctuating load industry; If BD P <(1-γ)×BDY P , then the industry is classified as a low-fluctuation-load industry; Wherein, γ is a preset threshold correction coefficient, wherein γ<1; Step 3.2 The peak-valley characteristic value of electricity consumption is C Y and the pre-set peak-to-valley difference feature classification threshold CY Y Compare and compare the peak and valley characteristic values ​​of electricity consumptionY Compared with the pre-set peak-to-valley ratio feature classification threshold RY Y For comparison: If C Y >CY Y and R Y >RY Y If all the contrast formulas in the formula are established, the industry is classified as a high peak-valley difference industry; If C Y >CY Y and R Y >RY Y If only one of the contrasting formulas is established, the industry is classified as a medium peak-valley difference industry; If C Y >CY Y and R Y >RY Y If all the contrast formulas in the formula do not hold true, the industry will be classified as a low peak-valley difference industry; Step 3.3: Set the power consumption time fluctuation characteristic value BS P The pre-set power consumption time fluctuation characteristic classification threshold BSY P For comparison: If BS P >(1+γ)×BSY P , then the industry is classified as a high-fluctuation electricity consumption industry; If (1+γ)×BSY P ≤BS P ≤(1+γ)×BSY P , then the industry is classified as a medium-fluctuating electricity consumption industry; If BS P <(1-γ)×BSY P , then the industry is classified as a low-fluctuation electricity consumption time industry; Step 5: Data integration: The results obtained based on the load fluctuation characteristic value, the peak and valley characteristic values ​​1 and 2, and the time fluctuation characteristic value of electricity consumption are integrated to form a comprehensive industrial electricity consumption data classification result, which is then stored in the database; For example, the comparison results of the characteristic value corresponding to a certain industrial electricity consumption data and its corresponding threshold are: (1+γ)×BDY P ≤BD P ≤(1+γ)×BDY P 、C Y ≤CY Y 、R Y ≤RY Y ,BS P >(1+γ)×BSY P, then the electricity consumption data of this industry corresponds to the comprehensive industrial electricity consumption data classification result of "medium fluctuation load type, low electricity consumption peak-valley difference type, high fluctuation electricity consumption time type" category industry.

[0037] This embodiment collects industrial electricity consumption data and performs a series of steps such as preprocessing, feature extraction, data classification and data integration in sequence, fully considering various data parameters such as power, electricity consumption, electricity consumption time and their corresponding characteristic values. Based on the comparison with the pre-set threshold, it can more accurately classify industries according to different characteristics such as load fluctuation, peak-to-valley difference in electricity consumption, and fluctuation in electricity consumption time, which helps to comprehensively and carefully grasp the electricity consumption characteristics and laws of different industries, and provide an accurate basis for the rational allocation of power resources and industrial electricity management. The industrial electricity consumption data is analyzed from multiple dimensions such as load fluctuation characteristics, peak-to-valley characteristics and electricity consumption time fluctuation characteristics, avoiding the one-sidedness of single-dimensional evaluation, so that the final comprehensive industrial electricity consumption data classification results are more scientific and have reference value.

[0038] Example 2 As the second embodiment of the present invention, when the present application is specifically implemented, compared with the first embodiment, the technical solution of this embodiment is different from that of the first embodiment only in that, in this embodiment, the preprocessing step is also used to perform missing value filling processing on the data sequence obtained after data labeling; The specific method is as follows: In the specified period, count whether each data series contains missing values ​​in each standard period, and then fill in the missing values; The missing value statistics method is: in the corresponding data sequence, if there is no numerical data in the t-th standard period, it means that there is a missing value in the t-th standard period of the corresponding data sequence; Filling is handled as follows: Taking power data as an example, in the power data series, when the tth standard period contains missing values; Then obtain the power parameters of the standard period adjacent to the standard period, that is, P t-1 and P t+1 ; Among them, P t-1 represents the power parameter in the previous standard period adjacent to the tth standard period, P t+1 represents the power parameter in the next standard period adjacent to the t-th standard period; Afterwards via:

[0039] Calculate the filling value PT in the tth standard period in the power data sequence t , fill value PT tUsed to supplement the power data sequence and serve as the power parameter P in the tth standard period t use; Among them, in the power data series, when the value of t in the missing value Pt is 1 or n; Then by:

[0040] Calculate the fill value PT in the first or nth standard period in the power data sequence t ; Its formula is: When t=1, the average value of the power parameters in the three standard periods adjacent to the t-th standard period is calculated as the filling value in the first standard period; When t=n, the average value of the power parameters in the first three standard time periods adjacent to the t-th standard time period is calculated as the filling value in the n-th standard time period.

[0041] This embodiment adds missing value filling to the data sequence after data labeling in the preprocessing step. This can effectively solve the problems of inaccurate subsequent analysis and biased feature extraction caused by missing data, ensure the integrity of the data, and make the analysis of load fluctuation characteristics, peak and valley characteristics of electricity consumption, and time fluctuation characteristics of electricity consumption based on this data more reliable, thereby improving the accuracy of the entire industrial electricity consumption data classification; after missing value filling, the data sequence that may have been unable to be effectively utilized due to missing data in some time periods is improved, so that the subsequent steps can fully utilize the data of all standard time periods for calculation, comparison, and classification, maximize the role of the data, and improve the usability of the data in industrial electricity consumption classification analysis.

[0042] Example 3 As the third embodiment of the present invention, when the present application is specifically implemented, compared with the first and second embodiments, the technical solution of this embodiment is to combine the solutions of the first and second embodiments. The difference between the technical solution of this embodiment and the first and second embodiments is only in this embodiment; The preprocessing step is also used to perform outlier replacement on the data sequence obtained after data labeling; The specific method is as follows: Taking power data as an example, in the power data series, the mean and standard deviation of all power parameters are calculated and marked as P and p and P b ; Then, each power parameter in the power data sequence is combined with its corresponding mean value and standard deviation to determine the outliers: If P t >P p +3×Pb and Pt>P p +3×P b If at least one of the following is true, the corresponding power parameter P t Determined as an outlier; If P p +3×P b ≤P t ≤P p +3×P b , then the corresponding power parameter P t Determined as an outlier; Then the power parameter P will be determined as an outlier. t Make data corrections; The method is as follows: In the power data series, obtain all power parameters that are not outliers, then calculate their average value and mark it as P1 p ; Then its average value is used as the outlier corresponding to all power parameters P t Correction value.

[0043] This embodiment further adds an outlier replacement process to the data sequence after data labeling in the preprocessing stage. By scientifically determining the outliers in the data sequence such as power and correcting them in a reasonable way, the adverse effects of outliers on the overall data characteristics can be avoided, such as preventing outliers from causing deviations in characteristic values ​​such as variance, further optimizing the accuracy of the data, and allowing the various characteristic values ​​extracted based on these data to more truly reflect the actual situation of industrial power consumption, thereby improving the credibility of the industrial power consumption data classification results; after removing the interference of outliers, whether it is the analysis of load fluctuation characteristics, peak and valley characteristics of power consumption, or time fluctuation characteristics of power consumption, it will be more in line with the normal laws of industrial power consumption. In the data classification link, classification is performed based on these more accurate and reliable characteristic values ​​compared with the corresponding thresholds, ensuring the reliability of the entire industrial power consumption data classification process and providing more powerful data support for subsequent related decisions.

[0044] Example 4 As the fourth embodiment of the present invention, when the present application is specifically implemented, compared with the first, second and third embodiments, the only difference between this embodiment and the first, second and third embodiments is that in this embodiment, the classification analysis method in the data classification step is also as follows: Step 4.1: Preliminary clustering based on K-Means++ algorithm First, for multiple industrial users, the multiple eigenvalues ​​obtained in the feature extraction step are combined into a feature vector, Xi=[BD P，i 、C Y，i 、R Y，i ,BS P，i]; Where i = 1, 2, ..., m, m represents the number of industrial users; Subsequently, the K-Means++ algorithm is used to perform preliminary clustering of industry users in the following manner: First, randomly select a data point as the first cluster center; Then for each data point Xi, calculate its shortest distance to the selected cluster center, and select the data point that maximizes the shortest distance as the next cluster center; Repeat this process until k cluster centers are selected, where the value of k is estimated based on the approximate number of industry types. Then, the data points are assigned to the cluster with the smallest distance based on the shortest distance between the data points and the cluster center; The distance calculation formula is:

[0045] Where, is the shortest distance between the data point and the cluster center, and g = 1, 2, ... v, v is the dimension of the feature vector, j = 1, 2, ... k, C j is the jth cluster center, C j，g is the gth eigenvector in the jth cluster center, X i，g is the gth feature vector within the i-th industry user; Step 4.2: Secondary classification and adjustment based on hierarchical clustering Based on the clustering results of K-Means++ algorithm, hierarchical clustering algorithm is used for secondary classification and adjustment; First, consider each K-Means++ cluster as an initial cluster and calculate the similarity between clusters; the specific method is: The average linkage method is used to calculate the similarity between clusters. Two clusters are selected and recorded as A and B respectively. Then: Calculate the similarity S(A, B) between A and B; Where d(x, y) represents the distance between two data points x and y, x represents the nth data point in cluster A, and y represents the nth data point in cluster A, which is obtained by the Euclidean distance formula. |A| and |B| represent the number of data points in cluster A and cluster B, respectively. Then, the cluster pairs with the highest similarity are merged and the cluster hierarchy is gradually constructed until the preset number of categories is reached or the similarity between clusters is lower than the preset threshold. In this embodiment, the hierarchical clustering algorithm can optimize the preliminary classification results of the K-Means++ algorithm, further refine the classification boundaries, and improve the classification accuracy; The data integration steps are as follows: After the classification algorithm unit completes the initial clustering of the K-Means++ algorithm and the secondary classification and adjustment of the hierarchical clustering, the first step is to organize each final category. Each category is assigned a unique category number or label; at the same time, the corresponding industry users in each category are recorded; In this embodiment, it ensures that each industry user can clearly correspond to a classification result, which facilitates subsequent data query and management; The sorted results are then stored in the database.

[0046] This embodiment uses the K-Means++ algorithm for preliminary clustering. By intelligently selecting cluster centers, the clustering results are made more representative and stable. Furthermore, a hierarchical clustering algorithm is used for secondary classification and adjustment. The average linkage method is used to calculate inter-cluster similarity and flexibly merge clusters. This effectively optimizes the preliminary classification results, further refines the classification boundaries, and significantly improves the classification accuracy of industrial electricity consumption data. This multi-stage clustering approach can more accurately capture subtle differences in electricity consumption characteristics between industries, providing more reliable data support for refined management of power resources and precise decision-making. During the data integration step, the resulting categories are organized, each assigned a unique number or label, and the corresponding industrial users within the category are recorded. This approach greatly facilitates subsequent data query and management. Whether the power sector is conducting daily data monitoring and analysis or conducting in-depth research on the electricity consumption of a specific industry, relevant data can be quickly and accurately located, improving work efficiency and reducing data management costs.

[0047] Example 5 As the fifth embodiment of the present invention, when this application is specifically implemented, compared with the first, second, third and fourth embodiments, the technical solution of this embodiment is to combine and implement the solutions of the above-mentioned first, second, third and fourth embodiments.

[0048] This embodiment combines all the advantages of Examples 1 through 4. It performs data labeling, missing value filling, and outlier replacement during the data preprocessing phase, effectively addressing data quality issues and ensuring data integrity and accuracy. Accurate data is the foundation for subsequent analysis and classification, preventing classification errors or biases caused by data issues and making the classification results of industrial electricity consumption data more reliable. From feature extraction to data classification, it comprehensively considers multiple data parameters and their characteristic values, such as power, power consumption, and power consumption time. Classification is performed by comparing with preset thresholds and using advanced clustering algorithms, achieving accurate classification of industrial electricity consumption data from multiple dimensions. Furthermore, during the data integration phase, the classification results are effectively organized and stored, facilitating data query and management. This not only helps the power sector gain a comprehensive and in-depth understanding of the characteristics and patterns of electricity consumption in different industries, providing an accurate basis for the rational allocation of power resources and industrial electricity management, but also improves the overall efficiency and quality of power data management, promoting the intelligent development of the power industry.

[0049] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0050] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for classifying industrial electricity data taking into account energy consumption characteristics, characterized in that: include: Step 1: Data collection: Collect industrial electricity consumption data, including data parameters corresponding to power, electricity consumption, and electricity consumption time; Step 2: Preprocessing: The industrial electricity consumption data is processed by data labeling, missing value filling and outlier replacement, and based on the processing results, the data sequences corresponding to different data parameters after missing value filling and outlier replacement are obtained; Step 3: Feature extraction: Based on the preprocessed multiple data sequences, feature extraction processing is performed on each data sequence respectively, and multiple feature values ​​of the industrial electricity consumption data are determined; Step 4: Data classification: Based on the multiple feature values ​​extracted from the features, classification analysis is performed, and then the classification results of the industrial electricity consumption data are determined through classification analysis; Step 5: Data Integration: The classification results based on data classification are integrated to form a comprehensive industrial electricity data classification result, which is then stored in the database.

2. The industrial electricity data classification method considering energy consumption characteristics according to claim 1 is characterized in that: Data labeling is handled as follows: In the specified period, it is divided into several standard time periods with uniform time intervals, and then the power, power consumption, and power consumption time collected in each standard time period are respectively composed into corresponding data sequences, and they are recorded in turn as power data sequences P = {P t }、Electricity consumption data series Y={Y t }、Electricity consumption time data series S={S t }; Among them, t is the standard period in the specified period, t=1, 2, ... n, n represents the number of collection time nodes in the specified period; The power parameter in the power data sequence is: in a standard period, a number of power values ​​are obtained according to a predetermined acquisition frequency, and then the average value of the several power values ​​is used as the power parameter of the standard period; The power consumption parameters in the power consumption data series are: the power consumption between the beginning and end of the standard period; The power consumption time parameter in the power consumption time data series is: the power consumption time between the beginning and end of the standard period.

3. The industrial electricity data classification method considering energy consumption characteristics according to claim 1 is characterized in that: Missing value filling is handled as follows: In the specified period, statistics are conducted to determine whether each data series contains missing values ​​in each standard time period, and then the missing values ​​are filled.

4. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 3, characterized in that: The missing value statistics are as follows: in the corresponding data sequence, if the tth standard period does not contain numerical data, it means that the tth standard period in the corresponding data sequence contains missing values.

5. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 3, characterized in that: The filling is handled as follows: Taking power data as an example, in the power data series, when the tth standard period contains missing values; Then the power parameters of the standard period adjacent to the standard period are obtained, that is, P t-1 and P t+1 ; Among them, P t-1 represents the power parameter in the previous standard period adjacent to the tth standard period, P t+1 represents the power parameter in the next standard period adjacent to the tth standard period; Afterwards via: Calculate the filling value PT in the tth standard period in the power data sequence t , fill value PT t Used to supplement the power data sequence and serve as the power parameter P in the tth standard period t use; Among them, in the power data series, when the t value in the missing value Pt is 1 or n; Then by: Calculate the fill value PT in the first or nth standard period in the power data sequence t ; Its formula is: When t=1, the average value of the power parameters in the next three standard periods adjacent to the tth standard period is calculated as the filling value in the first standard period; When t=n, the average value of the power parameters in the first three standard time periods adjacent to the t-th standard time period is calculated as the filling value in the n-th standard time period.

6. The industrial electricity data classification method considering energy consumption characteristics according to claim 1 is characterized in that: Outlier replacement is handled as follows: Select the power data, calculate the mean and standard deviation of all power parameters in the power data sequence, and mark them as P p and P b ; Then, each power parameter in the power data sequence is combined with its corresponding mean value and standard deviation to determine the outliers: If P t >P p +3×P b and Pt>P p +3×P b If at least one of the following holds true, then the corresponding power parameter P t Determined as an outlier; If P p +3×P b ≤P t ≤P p +3×P b , then the corresponding power parameter P t Determined as an outlier; Then the power parameter P that is judged as an abnormal value t Make data corrections; According to the outlier replacement processing method of the power data series, outlier replacement processing is performed on other data series.

7. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 6, characterized in that: The data correction method is as follows: In the power data series, obtain all power parameters that are not outliers, then calculate their average value and mark it as P1 p ; Then its average value is taken as the outlier corresponding to all power parameters P t Correction value.

8. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 1, characterized in that: The specific method of feature extraction is as follows: Step 2.1: Extract the power data sequence P t ; Then through: Calculate the variance BD corresponding to the power parameter within the specified period P , and use it as the load fluctuation characteristic value; Among them, P p is the mean of all power parameters in the power data sequence; Step 2.2, extract the power consumption data sequence Y t ; Then extract the power consumption parameter with the largest value in the power consumption data sequence and record it as Ymax; At the same time, the power consumption parameter with the minimum value is extracted from the power consumption data sequence and recorded as Ymin; then: C Y =Ymax-Ymin calculates the peak-to-valley difference C of electricity consumption within a specified period Y , and use it as the peak-valley characteristic value of electricity consumption; Also through: Calculate the average value Y corresponding to the power consumption parameter within the specified period P , and then compare the power consumption parameters of each standard period with Y p The comparison is performed, and the power consumption parameters of each standard period are assigned a digital value of "0" or "1" according to the comparison result, and the assignment result is recorded as F t ; The method is as follows: When Y t ≥Y p , then let F t The value of is 1, that is, the power consumption parameter of the corresponding standard period is assigned by the number "1"; When Y t <Y p , then let F t The value of is 0, that is, the power consumption parameter of the corresponding standard period is assigned with the number "0"; And so on, until all the assignment results of all standard time periods are determined; Afterwards via: Calculate the ratio of the number of standard time periods where the power consumption parameter exceeds the average value of all power consumption parameters to the total number of standard time periods. Y , and record it as the peak-valley characteristic value 2 of electricity consumption; Step 2.2: Extract the electricity consumption time data sequence S t ; Then through: Calculate the variance BS corresponding to the power consumption time parameter within the specified period P , and use it as the characteristic value of power consumption time fluctuation; where S p is the mean of all power consumption time parameters in the power data series.

9. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 8, characterized in that: The classification analysis in the data classification step is as follows: Step 3.1: Set the load fluctuation characteristic value BD P With the pre-set load fluctuation characteristic classification threshold BDY P For comparison: If BD P >(1+γ)×BDY P , then the industry is classified as a high fluctuating load industry; If (1+γ)×BDY P ≤BD P ≤(1+γ)×BDY P , then the industry is classified as a medium-fluctuating load industry; If BD P <(1-γ)×BDY P , then the industry is classified as a low-fluctuation-load industry; Wherein, γ is a preset threshold correction coefficient, wherein γ<1; Step 3.2 The peak-valley characteristic value of electricity consumption is C Y The pre-set peak-to-valley difference feature classification threshold CY Y Compare and compare the peak and valley characteristic values ​​of electricity consumption Y Compared with the preset peak-to-valley ratio feature classification threshold RY Y For comparison: If C Y >CY Y and R Y >RY Y If all the comparison formulas in the above are established, the industry will be classified as a high peak-to-valley difference industry; If C Y >CY Y and R Y >RY Y If only one set of comparison formulas is established, the industry will be classified as a medium-peak-valley electricity consumption industry; If C Y >CY Y and R Y >RY Y If all the comparison formulas in the above are not true, the industry will be classified as a low peak-to-valley difference industry; Step 3.3: Set the power consumption time fluctuation characteristic value BS P The pre-set power consumption time fluctuation characteristic classification threshold BSY P For comparison: If BS P >(1+γ)×BSY P , then the industry is classified as an industry with high fluctuation in electricity consumption time; If (1+γ)×BSY P ≤BS P ≤(1+γ)×BSY P , then the industry is classified as a medium-fluctuating electricity consumption time industry; If BS P <(1-γ)×BSY P , then the industry is classified as an industry with low fluctuation in electricity consumption time.

10. The method for classifying industrial electricity data taking into account energy consumption characteristics according to claim 8, characterized in that: The classification analysis in the data classification step is as follows: Step 4.1: Preliminary clustering based on K-Means++ algorithm: First, for multiple industry users, the multiple eigenvalues ​​obtained in the feature extraction step are combined into a feature vector, Xi = [BD P，i , C Y，i , R Y，i ,BS P，i ]; Where i = 1, 2, ..., m, m represents the number of industrial users; Subsequently, the K-Means++ algorithm is used to perform preliminary clustering of industry users in the following way: First, randomly select a data point as the first cluster center; Then for each data point Xi, calculate its shortest distance to the selected cluster center, and select the data point that maximizes the shortest distance as the next cluster center; Repeat this process until k cluster centers are selected, where the value of k is estimated based on the approximate number of industry types; Then, the data points are assigned to the cluster with the smallest distance based on the shortest distance between the data points and the cluster center; The distance calculation formula is: In the formula, is the shortest distance between the data point and the cluster center, and g = 1, 2, ... v, v is the dimension of the feature vector, j = 1, 2, ... k, C j is the jth cluster center, C j，g is the g-th eigenvector in the j-th cluster center, X i，g is the g-th feature vector within the i-th industry user; Step 4.2: Secondary classification and adjustment based on hierarchical clustering: Based on the clustering results of K-Means++ algorithm, hierarchical clustering algorithm is used for secondary classification and adjustment; First, consider each K-Means++ cluster as an initial cluster and calculate the similarity between clusters; the specific method is: The average linkage method is used to calculate the similarity between clusters. Two clusters are selected and recorded as A and B respectively. Then: Calculate the similarity S(A, B) between A and B; where d(x, y) represents the distance between two data points x and y, x represents the number of data points in cluster A, and y represents the number of data points in cluster A, which is obtained by the Euclidean distance formula, and |A| and |B| represent the number of data points in cluster A and cluster B respectively; then start merging the cluster pairs with the highest similarity, and gradually build a clustering hierarchy until it reaches the corresponding preset number of classifications or the similarity between clusters is lower than the corresponding preset threshold.

Images