A method for distributed extraction of typical power consumption patterns and related devices

Abstract

The application provides a method for extracting distributed typical power consumption modes and related equipment, and relates to the field of industrial power big data processing and analysis, and comprises the following steps: collecting power load data of a plurality of industrial users in a target area; processing the power load data of each industrial user to obtain an electricity consumption state matrix of the industrial user; generating a load curve cluster using the electricity consumption state matrices of the plurality of industrial users, and classifying load curves of the same state and the same duration in the load curve cluster into the same category; extracting a typical power consumption mode for each category of load curve to obtain a typical power consumption mode cluster corresponding to each industrial user; and fusing the typical power consumption mode clusters of the plurality of industrial users to obtain a typical power consumption mode. The method can effectively improve the accuracy and representativeness of the extraction of the typical power consumption mode, and improve the efficiency of mode extraction and data analysis.

Description

Technical Field

[0001] This invention relates to the field of industrial power big data processing and analysis, and in particular to a method and related equipment for distributed extraction of typical power consumption patterns. Background Technology

[0002] Benefiting from the rapid development of industrial manufacturing and new-generation high-tech information technology industries, the electrification level of various industries has continued to improve, leading to a surge in demand for electricity. According to statistics from the National Energy Administration, my country's total electricity consumption in 2021 reached 8,312.8 billion kilowatt-hours, of which industrial electricity consumption was 5,509 billion kilowatt-hours, accounting for 66.27% of total electricity consumption and a staggering 98.15% of secondary industry electricity consumption. This demonstrates that industrial electricity load constitutes a significant proportion of the overall electricity demand, especially in regions with heavy industrial development.

[0003] Industrial load has become a significant component of my country's electricity load. With the rapid development of smart grids and the continuous advancement of smart meter technology and electricity consumption data collection technology, an increasing number of smart meters are being deployed in households and industries. The frequent collection of user electricity load data has led to a significant increase in the amount of data. Simultaneously, influenced by market conditions and external factors, industrial users' electricity consumption patterns exhibit high uncertainty, complexity, and diversity, making it difficult to uncover underlying patterns and posing a significant challenge to further data analysis and decision support. To a certain extent, this may even become a major factor hindering the future development of smart grids and intelligent power systems.

[0004] Data-driven extraction of user electricity consumption patterns is a crucial foundation for unlocking the value of power big data applications. Scientifically sound extraction can uncover the electricity consumption habits of industrial users, revealing potential patterns and further improving the efficiency of data analysis. On one hand, power suppliers can provide flexible, personalized, and precise user services based on industrial users' electricity consumption patterns, while also positively promoting intelligent resource management and real-time scheduling of smart power systems to improve their operational efficiency. On the other hand, industrial users can understand their factory's production patterns and electricity consumption based on their own electricity consumption preferences, providing valuable information for developing scientifically sound production plans and energy-saving and emission-reduction strategies.

[0005] However, existing methods for extracting typical user electricity consumption patterns are facing the following challenges:

[0006] First, there is little consideration given to extracting typical electricity consumption patterns at the industrial user level;

[0007] Second, the drift effect of user electricity consumption behavior is ignored in the pattern extraction process, which makes most typical electricity consumption patterns have certain temporal and spatial similarities. In fact, similar electricity consumption patterns can be characterized by the cross-sectional curves of typical electricity consumption patterns with drift effects.

[0008] Third, it is mainly designed for single users and rarely triggers from multiple user levels to extract typical electricity consumption patterns that are generally representative. At the same time, the process of extracting electricity consumption behavior patterns from multiple users faces challenges such as large data volume and low computational efficiency.

[0009] In summary, there is an urgent need for a new method for extracting typical electricity consumption patterns to address the various challenges faced by existing technologies. Summary of the Invention

[0010] This invention provides a method and related equipment for distributed extraction of typical power consumption patterns, with the aim of improving the accuracy of power consumption pattern extraction and the efficiency of data calculation and analysis for industrial users.

[0011] To achieve the above objectives, the present invention provides a distributed method for extracting typical electricity consumption patterns, comprising:

[0012] Step 1: Collect power load data from multiple industrial users within the target area;

[0013] Step 2: Process the power load data of each industrial user to obtain the power consumption status matrix of the industrial users.

[0014] Step 3: Generate a load curve set cluster using the power consumption status matrix of multiple industrial users, and group the load curves of the same state and duration in the load curve set cluster into one category;

[0015] Step 4: Extract typical power consumption patterns for each type of load curve to obtain a set of typical power consumption patterns for each industrial user.

[0016] Step 5: Merge the typical power consumption patterns of multiple industrial users to obtain typical power consumption patterns.

[0017] Furthermore, step 1 includes:

[0018] For each industrial user within the target area, collect N days' worth of electricity load data from the industrial user's smart meter; the sampling frequency for electricity load data sampling is T per day.

[0019] Preprocess the power load data for N days to obtain power load data X;

[0020] The power load data X is normalized to obtain the normalized power load data X′;

[0021] Where X = [x1, x2, ..., x] d , ..., x N ], and x d =[x d，1 x d，2 , ..., x d，t , ..., x d，T ], x d，t This represents the daily load data collected by industrial users during the t-th time period on day d, where t ranges from [1, T].

[0022] Furthermore, step 2 includes:

[0023] For each industrial user's power load data, the normalized power load data X′ is processed to obtain the average power load μ of the industrial users;

[0024] The daily load data is divided into load states based on the average power load μ, resulting in a load state division into base state load and excited state load.

[0025] The load state matrix S of industrial users is obtained based on the average power load μ and the partitioning results.

[0026] Furthermore, step 2 also includes:

[0027] The average power load μ is

[0028]

[0029] Where N represents the number of sampling days, and T represents the frequency of sampling power load data per day. x′ d，t x represents the normalized power load value for time period t on day d. d，t This represents the daily load data collected by the industrial user at time t on day d. max(X) represents the maximum load value in the historical power load data of the industrial user, and min(X) represents the minimum load value in the historical power load data of the industrial user.

[0030] pass The load state of power load data is divided into basic state load and excited state load. When S d，t =0 indicates that the load state is the base state load, when S d，t =1 indicates that the load state is an excited load.

[0031] Based on the average power load μ and the load status of industrial users, the load status matrix S of each industrial user is obtained as follows:

[0032]

[0033] Among them, S d，t Let S represent the power load status of an industrial user at time t on day d, and S d，t ∈{0,1}, N represents the number of sampling days, and T represents the frequency of sampling power load data per day.

[0034] Furthermore, step 3 includes:

[0035] Given the load state matrix S for each industrial user, the load state set at time t on day d is {s}. d，t} t＝1，2，…T ;

[0036] When t=1 and S d，t When = 0, time t is taken as the trigger time of the ground state load, and when s d，t When = 1, time t is taken as the end time of the ground state load;

[0037] When t=1 and S d，t When = 1, time t is taken as the trigger time of the excited state load, and when s d，t When t = 0, time t is taken as the end time of the excited state load.

[0038] For each industrial user's load status, based on the trigger and end times of the load status, the i-th load curve Lbas for that industrial user on day d with respect to the base state load is generated. d，i and the j-th load curve Lexc with respect to the excited-state load d，j .

[0039] Load curve Lbas d，i and load curve Lexc d，j Each load is encapsulated separately to obtain the set of load curves (Load_bas) for the industrial user on day d, representing the base state load. d and the set of load curves for excited-state loads, Load_exc d .

[0040] Load curve set Load_bas d and the load curve set Load_exc d Encapsulation is performed to obtain the load curve set cluster Load_model for all load states of industrial users on day d. d The Load_model set will collect load curves from all load conditions over N days. dThe load curve set cluster Load_model for industrial users is obtained by encapsulation.

[0041] The load curves with the same duration under the same load state in the Load_model set are grouped into one category, resulting in two types of load curve sets: LenLoad_bas, which represents the load curves of industrial users under the base state load, and LenLoad_exc, which represents the load curves of industrial users under the excited state load.

[0042] Furthermore, step 4 includes:

[0043] For each type of load curve, the distance between any two load curves X and Y of the same duration is obtained as DTW(X,Y) using a distance metric method based on DTW.

[0044] Based on the DTW(X,Y) load curve AP clustering algorithm, the load curve clusters LenLoad_bas and LenLoad_exc are clustered to extract the typical power consumption pattern set RepLoad_Model for the p-th industrial user under different load conditions. p .

[0045] The textFile() function in SparkContext is called to create an RDD of electricity load data for each industrial user. Then, the Spark distributed parallel computing framework is used to extract patterns from the RDD to obtain a typical electricity consumption pattern set cluster RepLoad_Model corresponding to each industrial user.

[0046] Furthermore, step 5 includes the following:

[0047] Based on typical electricity consumption patterns, multiple industrial users are classified to obtain industrial user clusters;

[0048] We perform big data analysis and visualization processing on typical electricity consumption patterns and industrial user clusters to obtain visualized data analysis reports.

[0049] The present invention also provides a system for distributed extraction of typical electricity consumption patterns, comprising:

[0050] The data acquisition module is used to collect power load data from multiple industrial users within the target area.

[0051] The data processing module is used to process the power load data of each industrial user separately to obtain the power consumption status matrix of the industrial users.

[0052] The data classification module is used to generate a cluster of load curves using the power consumption status matrix of multiple industrial users, and to group load curves of the same state and duration in the cluster into one category.

[0053] The data extraction module is used to extract typical power consumption patterns for each type of load curve, thereby obtaining a set of typical power consumption patterns for each industrial user.

[0054] The data fusion module is used to merge clusters of typical power consumption patterns from multiple industrial users to obtain typical power consumption patterns.

[0055] The present invention also provides a computer-readable medium, comprising: a computer program for storing a computer program, which, by executing the computer program, enables the above-described method for distributed extraction of typical electricity consumption patterns.

[0056] The present invention also provides a device for distributed extraction of typical power consumption patterns, for implementing the above-mentioned method for distributed extraction of typical power consumption patterns, characterized in that it includes: a memory and a processor;

[0057] Memory is used to store computer programs;

[0058] A processor is used to execute computer programs stored in memory.

[0059] The above-mentioned solution of the present invention has the following beneficial effects: The present invention considers the industrial user level, collects power load data from multiple industrial users, processes, extracts, and merges the power load data from multiple industrial users to obtain typical power consumption patterns, and classifies multiple industrial users according to the typical power consumption patterns to obtain industrial user clusters. This can effectively improve the accuracy and representativeness of typical power consumption pattern extraction, and also improve the efficiency of pattern extraction and data analysis. By performing big data analysis and visualization processing on typical power consumption patterns and industrial user clusters, visualized data analysis reports are obtained and recommended to power suppliers or industrial users, which can realize decision support, demand response, and energy conservation and emission reduction strategies.

[0060] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0061] Figure 1 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0062] To make the technical problems, solutions, and advantages of this invention clearer, a detailed description will be provided below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0063] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0064] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a locking connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0065] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0066] This invention addresses existing problems by providing a distributed method and related equipment for extracting typical electricity consumption patterns.

[0067] like Figure 1 As shown, embodiments of the present invention provide a distributed method for extracting typical electricity consumption patterns, targeting industrial power big data, including:

[0068] Step 1: Collect power load data from multiple industrial users within the target area;

[0069] Step 2: Process the power load data of each industrial user to obtain the power consumption status matrix of the industrial users.

[0070] Step 3: Generate a load curve set cluster using the power consumption status matrix of multiple industrial users, and group the load curves of the same state and duration in the load curve set cluster into one category;

[0071] Step 4: Extract typical power consumption patterns for each type of load curve to obtain a set of typical power consumption patterns for each industrial user.

[0072] Step 5: Merge the typical power consumption patterns of multiple industrial users to obtain typical power consumption patterns.

[0073] Specifically, step 1 includes:

[0074] For each industrial user within the target area, collect N days' worth of electricity load data from the industrial user's smart meter; the sampling frequency for electricity load data sampling is T per day.

[0075] Preprocess the power load data for N days to obtain power load data X;

[0076] Where X = [x1, x2, ..., x] d , ..., x N ], and x d =[x d，1 x d，2 , ..., x d，t , ..., x d，T ], x d，t This represents the daily load data collected by industrial users during time period t on day d, where t ranges from [1, T].

[0077] Perform maximum-min normalization on the power load data X to obtain the normalized power load data X′.

[0078] In this embodiment of the invention, industrial users within the target area refer to all industrial users within an industrial park or a certain range who have significant electricity consumption and noticeable power load during normal working hours. These users may have the same electricity consumption behavior pattern or multiple different electricity consumption behavior patterns. The power load data X is subjected to max-min normalization to obtain the normalized industrial user power load data X′. The normalized power load data X′ is then transmitted to a distributed data storage system for storage. Here, x′... d，t Let X represent the normalized power load value for the t-th time period on day d, max(X) represent the maximum load value in the historical power load data of industrial users, and min(X) represent the minimum load value in the historical power load data of industrial users.

[0079] Specifically, step 2 includes:

[0080] For each industrial user's power load data, the normalized power load dataset X′ is processed to obtain the average power load μ of the industrial users;

[0081] The daily load data is divided into load states based on the average power load μ, resulting in a load state division into base state load and excited state load.

[0082] The load state matrix S of industrial users is obtained based on the average power load μ and the partitioning results.

[0083] Specifically, based on the power load data of multiple industrial users, the average power load μ for each industrial user is obtained. The average power load μ is then used as the standard to determine the load level of the industrial users. The formula for calculating the average power load μ is as follows:

[0084]

[0085] This invention is based on the energy consumption situation in actual industrial electricity use, and is triggered by... The load state of power load data is divided into base state load and excited state load. Base state load refers to the basic load energy consumption generated by the use of low-energy-consuming electrical equipment in a factory or workshop when no work is being performed. Excited state load refers to the high load energy consumption generated by the use of high-energy-consuming electrical equipment in a factory or workshop when work is being performed. When S d，t =0 indicates that the load state is the base state load, when S d，t =1 indicates that the load condition is an excited state load;

[0086] Based on the average power load μ and the power consumption status of industrial users, the load state matrix S of each industrial user is obtained as follows:

[0087]

[0088] Among them, S d，t Let S represent the power load status of an industrial user at time t on day d, and S d，t ∈{0,1}, N represents the number of sampling days, and T represents the frequency of sampling power load data per day.

[0089] Specifically, step 3 includes:

[0090] Given the load state set on day d in the power consumption state matrix S for each industrial user, let {S d，t} t＝1，2，…T ;

[0091] When t=1 and S d，t = 0 or when t ≠ 1 and S d，t =0, S d，t-1 When S = 1, time t is taken as the trigger time of the ground state load. d，t When = 1, time t is taken as the end time of the ground state load;

[0092] When t=1 and S d，t =1 or when t≠1 and S d，t =1, S d，t-1 When S = 0, time t is taken as the trigger time of the excited state load. d，t When t = 0, time t is taken as the end time of the excited state load.

[0093] For each industrial user's load status, based on the trigger and end times of the load status, the i-th load curve Lbas for that industrial user on day d with respect to the base state load is generated. d，i and the j-th load curve Lexc with respect to the excited-state load d，j ;

[0094] Load curve Lbas d，i and load curve Lexc d，j Each load is encapsulated separately to obtain the set of load curves (Load_bas) for the industrial user on day d, representing the base state load. d and the set of load curves for excited-state loads, Load_exc d ;

[0095] Load curve set Load_bas d and the load curve set Load_exc d Encapsulation is performed to obtain the load curve set cluster Load_model for all load states of industrial users on day d. d The Load_model set will collect load curves from all load conditions over N days. d The load curve set cluster Load_model for industrial users is obtained by encapsulation.

[0096] In this embodiment of the invention, based on the triggering and ending times of the load state, the load curves at times [ts1, te1] and [ts2, te2] can be identified as the base state load and the activated state load, respectively, and are used as the i-th load curve Lbas of the industrial user on day d regarding the base state load. d，i and the j-th electronic model Lexc regarding the excited-state load d，j They are represented as follows:

[0097]

[0098]

[0099] Based on the load state identification method described above, multiple load curves of industrial users under different load states are obtained on day d. These load curves are then encapsulated into a set of load curves for the industrial user on day d under the base load, named Load_bas. d ={Lbas d，1 Lbas d，2 ...} and the set of load curves with respect to excited-state loads, Load_exc d ={Lexc d，1 Lexc d，2 , ...}.

[0100] Based on the aforementioned load state identification method and load curve set encapsulation strategy, the industrial user's load curve set cluster Load_bas is used to define the base state load. d ={Lbas d，1 Lbas d，2 ...} and the load curve Load_exc with respect to the excited state load. d ={Lexc d，1 Lexc d，2 Encapsulate the load curves (load_model) of the industrial user under all load conditions on day d, and obtain the load curve set cluster Load_model. d = {Load_bas d Load_exc d} will collect the load curve set cluster Load_model for all load conditions over N days. d Further encapsulation is performed to obtain the set of load curves for industrial users: Load_model = {Load_bas, Load_exc}, Load_bas = {Load_bas1, ..., Load_bas}. N}, Load_exc={Load_exc1,..., Load_exc N The load curves with the same duration under the same load condition in the Load_model set are grouped into one category, resulting in the industrial user load curve set set LenLoad_bas under the base state load and the load curve set set LenLoad_exc under the excited state load.

[0101] Specifically, step 4 includes:

[0102] For each type of load curve, the distance between any two load curves X and Y of the same duration is obtained as DTW(X,Y) using a distance metric method based on DTW.

[0103] Based on the DTW(X,Y) load curve AP clustering algorithm, the load curve clusters LenLoad_bas and LenLoad_exc are clustered to extract the typical power consumption pattern set RepLoad_Model for the p-th industrial user under different load conditions. p ;

[0104] The textFile() function in SparkContext is called to create an RDD of electricity load data for each industrial user. Then, the Spark distributed parallel computing framework is used to extract patterns from the RDD to obtain a typical electricity consumption pattern set cluster RepLoad_Model corresponding to each industrial user.

[0105] In this embodiment of the invention, load curves with the same duration under the same load state in the load curve set cluster Load_model are grouped into one category, thereby obtaining the load curve set LenLoad_bas for industrial users under the base state load with respect to the load duration. i and the set of load curves LenLoad_exc with respect to load duration under excited-state load. i The calculation formula is as follows:

[0106]

[0107]

[0108] Among them, LenLoad_bas i and LenLoad_exc i Let |·| represent the set of load curves for ground state load and excited state load, respectively, and let |·| represent the total number of elements in the set, i.e., the base number.

[0109] Based on the above set of load curves classified by duration under different load conditions, the set of load curves for industrial users under base load is obtained as LenLoad_bas = {LenLoad_bas1, LenLoad_bas2, ...} and the set of load curves under excited load as LenLoad_exc = {LenLoad_exc1, LenLoad_exc2, ...}.

[0110] Specifically, in this embodiment of the invention, the load curve clustering algorithm based on Dynamic Time Warping (DTW) is an AP clustering algorithm based on Dynamic Time Warping (DTW), which includes the following steps:

[0111] Step 41: Based on the DTW distance metric method, obtain the set of power consumption patterns for industrial users, LenLoad_bas. iThe similarity matrix S;

[0112] Step 42: Set the initial reference degree P and the number of iterations M. Generally, the reference degree P = median(S) and the number of iterations M = 50.

[0113] Step 43, obtain the industrial user's power consumption mode set LenLoad_bas i The attraction matrix R and the belonging matrix A;

[0114] Step 44: Iteratively update the attraction matrix R and the membership matrix A until the cluster centers no longer change during several iterations or the predetermined number of iterations is reached;

[0115] Step 45: Output the cluster centers and the power consumption patterns included in each cluster center, and add the cluster centers of each cluster to the typical power consumption pattern RepLoad_Model of the p-th industrial user. p In the process, clustering is completed until all power consumption patterns of the industrial user, including the clusters LenLoad_bas and LenLoad_exc, are clustered.

[0116] The distance metric method based on DTW is used to describe the similarity between load curves with the same duration under the same condition. For any two load curves X and Y with the same duration, the distance metric formula DTW(X,Y) between load curves X and Y can be expressed as follows:

[0117] DTW(X,Y)=minC(X,Y)

[0118] in, This represents the Euclidean distance between the load points x and y.

[0119] Based on the load curve clusters LenLoad_bas under base state load, LenLoad_exc under excited state load, and the load curve clustering algorithm based on DTW(X,Y), the load curve RepLoad_Model of the p-th industrial user under different load states is extracted. p This is to characterize similar electricity consumption patterns.

[0120] In this embodiment of the invention, a Spark Application runtime environment is constructed, SparkContext is started, and the textFile() function in SparkContext is called to create a load data RDD for the electricity load data of each industrial user stored in the distributed data storage system. The load status identification method, load curve extraction algorithm, and DTW-based AP clustering algorithm for load curves, as proposed above, are encapsulated in the machine learning algorithm library MLlib in Spark. SparkContext sends a request to the resource manager to extract typical load curves in the Executor. Based on the Spark distributed parallel computing framework, the Executor calls the relevant algorithms for typical load curve extraction encapsulated in the MLlib algorithm library. A distributed computing mode is adopted for the load data RDD files of multiple users to extract and cluster the typical load curves of multiple industrial users, thereby obtaining a typical electricity consumption pattern set cluster RepLoad_Model = {RepLoad_Model1, ..., RepLoad_Model2} for multiple industrial users. p ...}, where RepLoad_Model p This represents the set of typical electricity consumption patterns for the p-th industrial user.

[0121] Resilient Distributed Datasets (RDDs) are a format of resilient distributed datasets primarily used for distributed parallel computing on the Spark platform. RDDs are read-only collections of objects distributed across a cluster, composed of multiple partitions. They are constructed through transformation operations and automatically reconstructed after failures to achieve resilience. An RDD can be viewed as a Spark object, stored in memory or on disk. The in-memory distributed datasets enabled by the Spark distributed parallel computing framework, besides providing interactive queries, can optimize iterative workloads and can be used to build large-scale, low-latency data analysis applications.

[0122] Specifically, step 5 includes:

[0123] Representative typical power consumption patterns RepLoad_Agg are obtained from the typical power consumption pattern set cluster RepLoad_Model to characterize the overall power consumption behavior pattern of multiple industrial users, and the representative typical power consumption patterns RepLoad_Agg are fed back to the distributed data storage system.

[0124] Specifically, after step 5, the following also includes:

[0125] Based on the representative typical power consumption pattern RepLoad_Agg, industrial users are classified to obtain industrial user clusters with similar power consumption behavior patterns, and the industrial user clusters are fed back to the distributed data storage system.

[0126] Big data analysis and visualization processing are performed on typical power consumption patterns and industrial user clusters in distributed data storage systems to obtain visualized data analysis reports.

[0127] In this embodiment of the invention, big data analysis and visualization technologies are used to process typical electricity consumption patterns and industrial user clusters fed back to the distributed data storage system to obtain visualized data analysis reports. These reports are then transmitted to power suppliers and industrial users to assist them in formulating power strategies and energy-saving and emission-reduction plans, thereby achieving decision support and demand response.

[0128] Embodiments of the present invention also provide a system for distributed extraction of typical electricity consumption patterns, comprising:

[0129] The data acquisition module is used to collect power load data from multiple industrial users within the target area.

[0130] The data processing module is used to process the power load data of each industrial user separately to obtain the power consumption status matrix of the industrial users.

[0131] The data classification module is used to generate a cluster of load curves using the power consumption status matrix of multiple industrial users, and to group load curves of the same state and duration in the cluster into one category.

[0132] The data extraction module is used to extract typical power consumption patterns for each type of load curve, thereby obtaining a set of typical power consumption patterns for each industrial user.

[0133] The data fusion module is used to merge clusters of typical power consumption patterns from multiple industrial users to obtain typical power consumption patterns.

[0134] Embodiments of the present invention also provide a computer-readable medium, comprising: a computer program for storing a computer program, which, by executing the computer program, enables the above-described method for distributed extraction of typical electricity consumption patterns.

[0135] An embodiment of the present invention also provides a device for distributed extraction of typical power consumption patterns, used to implement the above-described method for distributed extraction of typical power consumption patterns, characterized in that it includes: a memory and a processor;

[0136] Memory is used to store computer programs;

[0137] A processor is used to execute computer programs stored in memory.

[0138] In summary, the embodiments of the present invention first collect power load data from multiple industrial users, process, extract, and fuse the power load data of multiple industrial users to obtain typical power consumption patterns, and classify multiple industrial users according to these typical power consumption patterns to obtain industrial user clusters. This can effectively improve the accuracy and representativeness of typical power consumption pattern extraction, and also improve the efficiency of pattern extraction and data analysis. By performing big data analysis and visualization processing on typical power consumption patterns and industrial user clusters, visualized data analysis reports are generated and recommended to power suppliers or industrial users, enabling decision support, demand response, and energy conservation and emission reduction strategies.

[0139] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for distributed extraction of typical electricity consumption patterns, characterized in that, include: Step 1: Collect power load data from multiple industrial users within the target area; Step 2: Process the power load data of each industrial user to obtain the power consumption status matrix of the industrial user, including: For each of the aforementioned industrial users' power load data, the standardized power load data is processed separately. The process is performed to obtain the average power load of the industrial users. ; Based on the average power load The daily load data is divided into load states, resulting in a load state division into ground state load and excited state load. Based on the average power load The load state matrix of the industrial user is obtained from the partitioning results. ; Step 3: Generate a load curve cluster using the power consumption status matrix of multiple industrial users, and group load curves of the same state and duration in the load curve cluster into one category, including: Define the load state matrix S for each of the industrial users. Heavenly The set of load states at time t is ,and , This indicates the frequency of daily power load data sampling; when and Sometimes or when and , At that time, The time is used as the trigger time for the ground state load. and At that time, The time is taken as the end time of the ground state load; when and Sometimes or when and , At that time, The time is taken as the triggering time of the excited state load. and At that time, The time is taken as the end time of the excited-state load; For each industrial user's load status, based on the trigger and end times of the load status, the industrial user's load status is generated in the [number]th [phase]. Tianzhong's first regarding ground state load A load curve and regarding the excited state load A load curve ; The load curve and the load curve They are packaged separately to obtain the industrial user in the first... Tianzhong's set of load curves for ground-state load and a set of load curves related to excited-state loads ; The set of load curves and the set of load curves Encapsulation is performed to obtain the industrial user in the first... d A collection of load curves under all load conditions for a given day. , will collect A collection of load curves under all load conditions for a given day. Encapsulation is performed to obtain the set of load curves for the industrial users. ; The load curve set cluster Load curves with the same duration under the same load condition are grouped into one category, resulting in two types of load curve clusters. These two types of load curve clusters are typical load curve clusters of the industrial user under the base state load. and the set of typical load curves of the industrial users under excited-state load. ; Step 4: Extract typical electricity consumption patterns for each type of load curve to obtain a typical electricity consumption pattern set cluster for each industrial user, including: For each type of load curve, a distance metric method based on DTW is used to obtain any two load curves of the same duration. and The distance between them is ; based on The load curve AP clustering algorithm is used to cluster the load curves. and the set of load curves Perform clustering and extract the first... A collection of typical electricity consumption patterns of industrial users under different load conditions ; The `textFile()` function in the Spark Context is called to create an RDD containing electricity load data for each industrial user. Then, using the Spark distributed parallel computing framework, patterns are extracted from the RDD to obtain a cluster of typical electricity consumption patterns corresponding to each industrial user. ; Step 5: Merge the typical power consumption patterns of multiple industrial users to obtain typical power consumption patterns.

2. The method for distributed extraction of typical electricity consumption patterns according to claim 1, characterized in that, Step 1 includes: For each industrial user within the target area, data is collected from the smart meters of those industrial users. N Daily power load data; the sampling frequency for daily power load data is [number]. ; Regarding the N The daily power load data is preprocessed to obtain the power load data. ; Power load data Perform standardization processing to obtain standardized power load data. ; in, ,and , Indicates that industrial users are in the first Heavenly Daily load data collected during the specified time period, The range of values ​​is .

3. The method for distributed extraction of typical electricity consumption patterns according to claim 2, characterized in that, Step 2 also includes: The average power load for in, Indicates the number of sampling days. This indicates the frequency of daily electricity load data sampling. , Indicates the first Heavenly Electricity load values ​​after time period normalization Indicates that industrial users are in the first Heavenly Daily load data collected during the specified time period, This represents the maximum load value in the historical power load data of industrial users. This represents the minimum load value in the historical power load data of industrial users; pass The load state of the power load data is divided into basic load and excited load. This indicates that the load condition is the base state load. This indicates that the load condition is an excited state load; Based on the average power load Based on the load status of industrial users, obtain the load status matrix for each industrial user. S for in, Indicates that industrial users are in the first Heavenly The electrical load status at any given time. Indicates the number of days for sampling.

4. The method for distributed extraction of typical electricity consumption patterns according to claim 1, characterized in that, Step 5 is followed by: Based on the typical electricity consumption patterns, multiple industrial users are classified to obtain industrial user clusters; Big data analysis and visualization processing are performed on the typical electricity consumption patterns and the industrial user clusters to obtain visualized data analysis reports.

5. A system for distributed extraction of typical electricity consumption patterns, characterized in that, include: The data acquisition module is used to collect power load data from multiple industrial users within the target area. The data processing module is used to process the power load data of each industrial user separately to obtain the power consumption status matrix of the industrial user, including: For each of the aforementioned industrial users' power load data, the standardized power load data is processed separately. The process is performed to obtain the average power load of the industrial users. ; Based on the average power load The daily load data is divided into load states, resulting in a load state division into ground state load and excited state load. Based on the average power load The load state matrix of the industrial user is obtained from the partitioning results. ; The data classification module is used to generate a load curve cluster using the electricity status matrix of multiple industrial users, and to group load curves of the same state and duration in the load curve cluster into one category, including: Define the load state matrix S for each of the industrial users. Heavenly The set of load states at time t is ,and , This indicates the frequency of daily power load data sampling; when and Sometimes or when and , At that time, The time is used as the trigger time for the ground state load. and At that time, The time is taken as the end time of the ground state load; when and Sometimes or when and , At that time, The time is taken as the triggering time of the excited state load. and At that time, The time is taken as the end time of the excited-state load; For each industrial user's load status, based on the trigger and end times of the load status, the industrial user's load status is generated in the [number]th [phase]. Tianzhong's first regarding ground state load A load curve and regarding the excited state load A load curve ; The load curve and the load curve They are packaged separately to obtain the industrial user in the first... Tianzhong's set of load curves for ground-state load and a set of load curves related to excited-state loads ; The set of load curves and the set of load curves Encapsulation is performed to obtain the industrial user in the first... d A collection of load curves under all load conditions for a given day. , will collect A collection of load curves under all load conditions for a given day. Encapsulation is performed to obtain the set of load curves for the industrial users. ; The load curve set cluster Load curves with the same duration under the same load condition are grouped into one category, resulting in two types of load curve clusters. These two types of load curve clusters are typical load curve clusters of the industrial user under the base state load. and the set of typical load curves of the industrial users under excited-state load. ; The data extraction module is used to extract typical electricity consumption patterns for each type of load curve, resulting in a set of typical electricity consumption patterns for each industrial user, including: For each type of load curve, a distance metric method based on DTW is used to obtain any two load curves of the same duration. and The distance between them is ; based on The load curve AP clustering algorithm is used to cluster the load curves. and the set of load curves Perform clustering and extract the first... A collection of typical electricity consumption patterns of industrial users under different load conditions ; The `textFile()` function in the Spark Context is called to create an RDD containing electricity load data for each industrial user. Then, using the Spark distributed parallel computing framework, patterns are extracted from the RDD to obtain a cluster of typical electricity consumption patterns corresponding to each industrial user. ; The data fusion module is used to fuse the typical power consumption patterns of multiple industrial users to obtain typical power consumption patterns.

6. A computer-readable medium comprising: The computer program is used to store a computer program, characterized in that, by executing the computer program, it is used to implement the distributed method for extracting typical power consumption patterns as described in any one of claims 1-4.

7. A device for distributed extraction of typical power consumption patterns, characterized in that, The method for implementing the distributed extraction of typical power consumption patterns according to any one of claims 1-4 is characterized in that it comprises: a memory and a processor; The memory is used to store computer programs; The processor is used to execute computer programs stored in the memory.

Images