Power consumption abnormity identification method based on density and clustering algorithm

Abstract

The invention discloses a power consumption abnormity identification method based on density and a clustering algorithm, and belongs to the field of power supply and consumption management. Comprising the steps of data collection; preprocessing the data; analyzing user energy consumption characteristics; constructing an energy consumption characteristic index library; screening power utilization abnormity suspected users based on a density calculation model; performing normal user clustering analysis; and identifying a real abnormal user. A density and clustering algorithm is applied to an enterprise user electricity consumption abnormity identification process in a power supply management process, abnormal suspicion users are preliminarily screened by using a density model, and then FCM clustering analysis is performed on residual normal electricity consumption users, so that electricity consumption characteristics of the normal users are obtained; and finally, through similarity analysis of power utilization characteristics of suspected users and normal users, real power utilization abnormal users are screened out. The system aims to monitor illegal power consumption behaviors of key energy-consuming enterprises, supervise and urge the key energy-consuming enterprises to produce and operate according to rules, and guarantee economic benefits of power grid enterprises. The method can be widely applied to the fields of operation management of power supply companies and power utilization monitoring of enterprise users.

Description

Technical Field

[0001] The present invention belongs to the field of power supply and consumption management, and in particular relates to a method for identifying power consumption anomalies for key energy-consuming enterprises. Background Art

[0002] With the vigorous development of the national economy, the electricity consumption level of various industries has been continuously improved, and abnormal monitoring of enterprise electricity consumption has become one of the focuses of supervision and management of my country's power sector.

[0003] In the actual power supply and consumption management process, it was found that some key energy-consuming enterprises, in order to save costs, resorted to stealing electricity or violating electricity contracts, which seriously affected the order of the entire society and damaged the economic benefits of power grid companies.

[0004] Therefore, it is imperative to strengthen the monitoring of abnormal electricity consumption by enterprises and assist regulatory authorities in achieving precise law enforcement against enterprises with abnormal electricity consumption in order to protect the economic benefits of power grid companies.

[0005] Abnormal detection of enterprise electricity consumption can serve as an alarm for enterprises' illegal electricity consumption, ensure the normal operation of enterprise facilities, and urge enterprises to produce and operate in compliance with the law.

[0006] Most of the existing anomaly detection methods are suitable for detecting whether users' electricity consumption is abnormal. However, due to the difficulties such as large data volume, no abnormal enterprise user labels, and complex electricity consumption patterns in various industries, the anomaly detection of enterprise electricity consumption has brought challenges.

[0007] The invention patent with the authorization announcement date of 2020.11.13 and the authorization announcement number of CN 108776276 B discloses a "power consumption anomaly detection method and system". First, the power consumption characteristic data and the operating time period of the power consumption equipment are obtained, and then the power consumption characteristic data is used to determine whether to generate an over-limit alarm according to the predetermined standard range; the power consumption characteristic data is used to calculate the dynamic time regularization distance, and determine whether to generate a mode alarm according to the dynamic time regularization distance; the operating time period of the power consumption equipment is used to determine whether to generate a vulnerability alarm according to the predetermined operating time period. This technical solution can automatically and comprehensively detect power consumption anomalies, overcoming the defects of low accuracy and effectiveness brought about by manual detection of power consumption anomalies. However, since it uses electricity consumption characteristic data to determine whether to generate an over-limit alarm based on a predetermined standard range, and the electricity consumption data of key energy-consuming enterprises have difficulties such as large data volume, no abnormal enterprise user tags, and complex electricity consumption patterns in various industries, the actual application process has high requirements for system computing power and server hardware. In addition, the electricity consumption database has a huge amount of data and a wide variety of types, resulting in slow retrieval or aggregation speed, which is not conducive to quickly obtaining monitoring data results, and also hinders the real-time and timeliness of enterprise electricity consumption anomaly detection and management.

[0008] The invention patent application with the application publication date of 2024.02.13 and the application publication number of CN 117554714A discloses a "power consumption anomaly detection method, power consumption anomaly detection device and electronic device", including: calculating the power consumption reference value of the target user according to the new energy impact coefficient, the first historical data and the second historical data; obtaining the actual power consumption data of the target user at the current moment, and calculating the absolute value of the difference between the power consumption reference value and the actual power consumption data of the target user at the current moment, and obtaining the target difference; when the target difference is greater than the second threshold, determining that the target user's power consumption is abnormal. This method solves the problem of poor power consumption anomaly detection effect when new energy is connected. Similarly, since it is necessary to calculate the target user's electricity consumption reference value based on the new energy impact coefficient, the first historical data and the second historical data during the data processing process, if the various impact coefficients and historical data mentioned above are deviated or erroneous, the basis of the entire data processing will be deviated. At the same time, it focuses on increasing the impact of the new energy impact coefficient on abnormal electricity consumption, but it is still based on the existing database and data processing model. It is necessary to build a user electricity consumption prediction model and use historical electricity consumption data to build a training sample set, which results in a long implementation cycle and a large workload for data training. It is not conducive to obtaining monitoring data results quickly and in real time, and has a certain lag effect on the real-time detection of enterprise electricity consumption anomalies and the timeliness of power supply operation management. Summary of the invention

[0009] The technical problem to be solved by the present invention is to provide a method for identifying abnormal electricity consumption based on density and clustering algorithms. It applies density and clustering algorithms to the process of identifying abnormal electricity consumption of enterprise users in the power supply management process, uses the density model to preliminarily screen abnormal suspected users, and then performs FCM clustering analysis on the remaining normal electricity users to obtain the electricity consumption characteristics of normal users. Finally, through the similarity analysis of the electricity consumption characteristics of suspected users and normal users, the real abnormal electricity users are screened out. It aims to monitor the illegal electricity consumption behavior of key energy-consuming enterprises, urge key energy-consuming enterprises to produce and operate in accordance with the law, and ensure the economic benefits of power grid enterprises.

[0010] The technical solution of the present invention is to provide a method for identifying abnormal power consumption based on density and clustering algorithms, including the collection of various power data in the power supply area, which is characterized by:

[0011] S1) Data collection;

[0012] S2) data preprocessing;

[0013] S3) Analysis of user energy consumption characteristics;

[0014] S4) constructing an energy characteristic index library;

[0015] S5) Screening suspected users of abnormal electricity consumption based on the density calculation model;

[0016] S6) Cluster analysis of normal users;

[0017] S7) Identification of truly abnormal users.

[0018] Specifically, in the step S1, the monthly electricity consumption, time-sharing electricity consumption, contract capacity, 96-point load and other electricity data of the electricity users in the region are matched through fields such as the enterprise electricity user number and the unified social credit code; these data can be collected in real time through devices such as smart meters, sensors and monitoring systems, and can also be collected offline through historical data.

[0019] Specifically, in step S2, the collected data is cleaned and normalized, invalid or erroneous data records are deleted during the cleaning process, and all data are converted into numerical values ​​between 0 and 1 during the normalization process to facilitate subsequent analysis and modeling.

[0020] Specifically, in step S3, data statistics methods are used to analyze the peak, off-peak and peak electricity consumption and seasonal electricity consumption of key energy-consuming enterprises, and the energy consumption characteristics of key energy-consuming enterprises are sorted out.

[0021] Specifically, in step S4, based on the above-mentioned analysis of user energy consumption characteristics, indicators such as peak / flat / valley electricity consumption proportion, summer / winter electricity consumption proportion, load rate, etc. are selected to construct a user characteristic indicator library, and abnormal electricity users are identified through cluster analysis.

[0022] Specifically, in step S5, density calculation is performed on the sample characteristic index data based on the density model, and cluster analysis technology is used to identify abnormal electricity users in combination with the spatial distribution of the samples.

[0023] Furthermore, in the step S6, users with a relatively high density are regarded as normal electricity users, and the characteristic data of these users are clustered and analyzed using the FCM clustering algorithm to obtain the patterns and characteristics of normal electricity users.

[0024] Furthermore, in the step S7, based on the power consumption characteristic index, similarity analysis is performed between users suspected of abnormal power consumption and users with normal power consumption, and users with less similarity are determined to be users with real abnormal power consumption.

[0025] Furthermore, the density calculation model is as follows:

[0026]

[0027] Among them, d ij is the Euclidean distance between two sample points, calculated from the characteristic data of each sample indicator; dc is the cutoff distance. The cutoff distance is about 2% of the sample set;

[0028] Based on the above density calculation model, density calculation is performed on all users, and the density threshold is set to 1.02. Users with a density less than 1.02 are considered to be suspected of abnormality.

[0029] Furthermore, based on the energy consumption characteristic indicators of each user, similarity analysis is performed between users suspected of abnormal electricity consumption and users with normal electricity consumption; users whose similarity difference is greater than the set upper limit are users with real abnormal electricity consumption;

[0030] Alternatively, by comparing the recent electricity usage data and historical electricity usage data of a user suspected of abnormal electricity usage, when it is found that the similarity difference is greater than a set upper limit, the user is confirmed to be a user suspected of abnormal electricity usage;

[0031] The upper limit is set to 1.5.

[0032] Compared with the prior art, the advantages of the present invention are:

[0033] 1. The technical solution of the present invention is mainly used for identifying abnormal electricity consumption of key energy-consuming enterprises through cluster analysis through enterprise electricity consumption data, such as monthly electricity consumption, time-sharing electricity consumption, contract capacity, 96-point load and other data; constructing a characteristic index library of indicators such as peak / peak / flat / valley electricity consumption proportion, summer / winter electricity consumption proportion, load rate, etc., which can screen out users with real abnormal electricity consumption and ensure that the economic benefits of power grid enterprises are not lost;

[0034] 2. The technical solution of the present invention can better monitor the abnormal electricity consumption of enterprises, urge enterprises to produce and operate in accordance with the law, and ensure that the economic benefits of power grid enterprises are fully realized; the present invention targets the problems of electricity consumption of key energy-consuming enterprises, and continuously optimizes them according to business application feedback, further improves and standardizes the indicator data, and supports the practical application of abnormal electricity consumption identification;

[0035] 3. The technical solution of the present invention can be applied to improve the safety of electricity use and energy consumption monitoring of key enterprises; by analyzing the electricity consumption data of key energy-consuming enterprises, it is possible to monitor the energy consumption of key energy-consuming enterprises, discover abnormal electricity consumption and potential problems, and take timely measures to adjust and optimize, thereby ensuring the safe and stable production and operation of enterprises and reducing the frequency of accidents;

[0036] 4. The technical solution of the present invention, the anomaly identification method based on density and clustering algorithm can provide technical support for power grid enterprises to monitor enterprise electricity consumption and promote the realization of benefits for power grid enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1It is a schematic diagram of the present invention for identifying abnormal power consumption of equipment in key energy-consuming enterprises. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] The technical solution of the present invention provides a method for identifying abnormal electricity consumption based on density and clustering algorithms, which aims to monitor the illegal electricity consumption behavior of key energy-consuming enterprises and urge them to produce and operate in compliance with the law.

[0040] Figure 1 The specific implementation steps of the technical solution of the present invention are as follows:

[0041] S1 Data Collection: Collect various data sources in the region, including monthly electricity, time-of-use electricity, contract capacity, 96-point load and other power data. These data can be obtained through sensors, monitoring equipment and public databases.

[0042] S2 Data preprocessing: Clean and preprocess the collected data, including data cleaning, data normalization, and missing value processing, etc. This ensures the accuracy and consistency of the data.

[0043] S3 User energy consumption characteristics analysis: Based on the collected power data, energy consumption analysis is carried out by time and season.

[0044] Construction of S4 energy consumption characteristic index library: Based on user energy consumption analysis, select indicators such as peak / flat / valley electricity consumption proportion, summer / winter electricity consumption proportion, load rate, etc. to construct a user energy consumption characteristic index library.

[0045] S5 screens suspected users of abnormal electricity consumption based on the density calculation model: density calculation is performed on sample characteristic indicator data based on the density model. Users with smaller density are considered to be suspected of abnormal users, and users with larger density are considered to be normal electricity users.

[0046] S6 Cluster analysis of normal users: For normal electricity users, the FCM clustering algorithm is used to perform cluster analysis on the characteristic data of these users to obtain the patterns and characteristics of normal electricity users.

[0047] S7 Identification of truly abnormal users: Based on various energy consumption characteristic indicators, users suspected of abnormal electricity usage are analyzed for similarity with users with normal electricity usage. Users with greater similarity are truly abnormal electricity users.

[0048] Example:

[0049] The method described in the technical solution of the present invention is specifically implemented as follows:

[0050] According to step S1, the monthly power consumption, time-sharing power consumption, contract capacity, 96-point load and other power data of power users in the region are matched through fields such as enterprise power user number and unified social credit code. These data can be collected in real time through devices such as smart meters, sensors and monitoring systems, or collected offline through historical data.

[0051] According to step S2, the collected data is cleaned and normalized, invalid or erroneous data records are deleted during the cleaning process, and all data are converted into numerical values ​​between 0 and 1 during the normalization process to facilitate subsequent analysis and modeling.

[0052]

[0053] Among them, x scal is the normalized value, x is the original value, and x min is the minimum value, x max is the maximum value.

[0054] According to step S3, a data statistical method is used to analyze the peak, off-peak and peak electricity consumption ratios and seasonal electricity consumption of key energy-consuming enterprises, and the energy consumption characteristics of key energy-consuming enterprises are sorted out.

[0055] According to step S4, based on the above-mentioned user energy consumption characteristics analysis, indicators such as peak / peak / flat / valley electricity consumption proportion, summer / winter electricity consumption proportion, load rate, etc. are selected to build a user characteristic indicator library, and abnormal electricity users are identified through cluster analysis.

[0056] 1) Contract capacity ratio:

[0057] Contract capacity refers to the capacity allowed for installed electrical equipment. There is a prescribed range standard for the ratio between a user's contract capacity and their actual electricity consumption. When this standard is exceeded, the user faces the risk of abnormal electricity consumption.

[0058] 2) Peak electricity consumption ratio:

[0059] The ratio of peak power consumption to total power consumption reflects the change in peak power consumption:

[0060]

[0061] Among them, peak_eate is the proportion of peak electricity consumption, E(peak) is the peak electricity consumption, and E(population) is the total electricity consumption.

[0062] 3) Percentage of electricity consumption during normal period:

[0063] The ratio of average power to total power, reflecting the change of power in average period:

[0064]

[0065] Among them, flat_rate is the proportion of electricity consumption in flat period, E(flat) is the electricity consumption in flat period, and E(population) is the total electricity consumption.

[0066] 4) Off-peak electricity consumption ratio:

[0067] The ratio of valley power to total power reflects the change of valley power:

[0068]

[0069] Among them, valley_rate is the proportion of electricity consumption in the valley period, E(peak) is the electricity consumption in the valley period, and E(population) is the total electricity consumption.

[0070] 5) Peak electricity consumption ratio:

[0071] The ratio of peak power to total power, reflecting the change of peak power:

[0072]

[0073] Among them, teap_rate is the proportion of electricity consumption during peak period, E(teap) is the electricity consumption during peak period, and E(population) is the total electricity consumption.

[0074] 6) Summer electricity consumption ratio:

[0075] Summer electricity consumption ratio = summer total electricity consumption / annual total electricity consumption;

[0076] 7) Proportion of electricity consumption in winter;

[0077] Winter electricity consumption ratio = total winter electricity consumption / total annual electricity consumption;

[0078] According to step S5, cluster analysis technology is used to identify abnormal electricity users in combination with the spatial distribution of samples.

[0079] Therefore, the concept of density is introduced, and a density calculation model is constructed to measure the spatial distribution of sample data and separate low-density sample points from high-density sample points.

[0080] The density calculation model is shown below.

[0081]

[0082] Among them, d ij is the Euclidean distance between two sample points, calculated from the characteristic data of each sample indicator; d c is the cutoff distance. The cutoff distance is about 2% of the sample set.

[0083] Based on the density calculation model, density calculation is performed on all users, and users with a relatively small density (in this technical solution, the threshold is set to a density less than 1.02) are regarded as users suspected of abnormality.

[0084] According to step S6, users with a relatively large density are regarded as normal electricity users, and the characteristic data of these users are clustered and analyzed using the FCM clustering algorithm to obtain the patterns and characteristics of normal electricity users.

[0085] According to step S7, based on the power consumption characteristic index, by performing similarity analysis between users suspected of abnormal power consumption and users with normal power consumption, users with less similarity are determined to be users with real abnormal power consumption.

[0086] The technical solution of the present invention proposes an abnormal identification model that integrates density algorithm and FCM clustering algorithm. First, the density model is used to preliminarily screen abnormal suspected users, and then FCM clustering analysis is performed on the remaining normal electricity users to obtain the electricity consumption characteristics of normal users. Finally, through the similarity analysis of the electricity consumption characteristics of suspected users and normal users, the real abnormal electricity users are screened out; it can better monitor the abnormal electricity consumption of enterprises, urge enterprises to produce and operate in accordance with the law, and continuously adjust and optimize according to business application feedback, and further improve and standardize the indicator data to support the practical application of abnormal electricity consumption identification; by analyzing the electricity consumption data of key energy-consuming enterprises, it can monitor the energy consumption of key energy-consuming enterprises, improve the safety of electricity use and energy consumption monitoring of key enterprises; discover abnormal electricity consumption and potential problems, and take timely measures to adjust and optimize, so as to ensure the safe and stable production and operation of enterprises, reduce the frequency of accidents, and provide technical support for power grid enterprises to monitor enterprise electricity consumption.

[0087] The present invention can be widely used in the fields of operation management of power supply companies and electricity consumption monitoring of enterprise users.

Claims

1. A method for identifying power consumption anomalies based on density and clustering algorithms, including the collection of various power data in the power supply area, characterized by: S1) Data collection; S2) data preprocessing; S3) Analysis of user energy consumption characteristics; S4) constructing an energy characteristic index library; S5) Screening suspected users of abnormal electricity consumption based on the density calculation model; S6) Cluster analysis of normal users; S7) Identification of truly abnormal users.

2. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In the step S1, the monthly power consumption, time-sharing power consumption, contract capacity, 96-point load and other power data of power users in the region are matched through fields such as the enterprise power user number and the unified social credit code; This data can be collected in real time through devices such as smart meters, sensors and monitoring systems, or offline through historical data.

3. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In step S2, the collected data is cleaned and normalized. In the cleaning process, invalid or erroneous data records are deleted. In the normalization process, all data are converted into numerical values ​​between 0 and 1 to facilitate subsequent analysis and modeling.

4. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In the step S3, data statistics methods are used to analyze the peak, off-peak and peak electricity consumption and seasonal electricity consumption of key energy-consuming enterprises, and the energy consumption characteristics of key energy-consuming enterprises are sorted out.

5. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In step S4, based on the above-mentioned analysis of user energy consumption characteristics, indicators such as peak / flat / valley electricity consumption proportion, summer / winter electricity consumption proportion, load rate, etc. are selected to build a user characteristic indicator library, and abnormal electricity users are identified through cluster analysis.

6. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In the step S5, density calculation is performed on the sample characteristic index data based on the density model, and cluster analysis technology is used to identify abnormal electricity users in combination with the spatial distribution of the samples.

7. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In the step S6, users with a relatively large density are regarded as normal electricity users, and the characteristic data of these users are clustered and analyzed using the FCM clustering algorithm to obtain the patterns and characteristics of normal electricity users.

8. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that In step S7, based on the power consumption characteristic index, similarity analysis is performed between users suspected of abnormal power consumption and users with normal power consumption, and users with less similarity are determined to be users with real abnormal power consumption.

9. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that The density calculation model is as follows: Among them, d ij is the Euclidean distance between two sample points, calculated from the characteristic data of each sample indicator; dc is the cutoff distance. The cutoff distance is about 2% of the sample set; Based on the above density calculation model, density calculation is performed on all users, and the density threshold is set to 1.

02. Users with a density less than 1.02 are considered to be suspected of abnormality.

10. The method for identifying abnormal power consumption based on density and clustering algorithm according to claim 1, characterized in that Based on the energy consumption characteristic indicators of each user, users suspected of abnormal electricity consumption and users with normal electricity consumption are analyzed for similarity; users whose similarity difference is greater than the set upper limit are users with real abnormal electricity consumption; Alternatively, by comparing the recent electricity usage data and historical electricity usage data of a user suspected of abnormal electricity usage, when it is found that the similarity difference is greater than a set upper limit, the user is confirmed to be a real user suspected of abnormal electricity usage.

Images