A method and system for identifying bus loads
By using LSTM neural network and fuzzy clustering algorithm, combined with power timing data, time data and meteorological data, the problem of difficult to predict the power changes in busbars of large power supply nodes and accurately identify large-scale and diversified loads in the prior art is solved, and high-precision load identification and prediction are achieved.
Patent Information
- Application Number
- CN201910591109.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-07-02
- Publication Date
- 2025-06-20
- Estimated Expiration
- 2039-07-02
AI Technical Summary
The prior art is difficult to effectively predict the load power changes of large-scale power supply node buses, and it is difficult to accurately identify the load types of large-scale and diversified loads.
The fuzzy clustering algorithm that combines maximum value normalization and particle swarm optimization is adopted to collect power timing data, time data and meteorological data to perform load identification and feature extraction to achieve accurate identification of bus load.
Accurate identification of large-scale and diversified loads is achieved, and the accuracy and reliability of bus load prediction are improved.
Smart Images
Figure CN110490220B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power system automation analysis, and in particular to a method and system for identifying bus loads. Background Art
[0002] In the power grid, some large power supply nodes (Bulk Supply Points, BSPs), such as the equivalent load nodes of transformers in 220 kV substations of provincial power grids and 10 kV feeder nodes of regional power grids, are usually connected to a large number of load resources of different types and with very different characteristics. How to effectively predict the changes in the load power of these bus nodes and how to effectively utilize the aggregation characteristics of the flexible load responses connected to these nodes are the key concerns of power grid dispatchers. Determining the types and quantities of loads connected to large power supply nodes is the basis and key for carrying out the above research. However, in the actual power grid, due to problems such as measurement costs, inaccurate network models, and system ownership, it is often difficult to directly obtain the composition of BSP loads through measuring devices. Existing load identification and extraction methods have the following problems: First, existing load resource identification technologies mostly target small-scale household users, and the research methods are difficult to be used for the identification of large-scale and diverse load-side resources; second, existing load identification algorithms are sensitive to factors such as load classification, individual scale, load models, and external environments, and the identification accuracy is limited.
[0003] Therefore, a method capable of accurately identifying large-scale and diverse loads on a bus is needed. Summary of the Invention
[0004] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide a method and system for identifying bus loads, which can accurately identify the load types of large-scale and diverse loads on a bus by using the ability of the LSTM neural network to handle complex data with non-linear and interactive effects.
[0005] The purpose of the present invention is achieved by the following technical solutions:
[0006] A method for identifying bus loads, the improvement being that the method includes:
[0007] Collecting load identification data of the bus to be identified;
[0008] Determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified;
[0009] Wherein, the load identification data includes: power time series data, time data, and meteorological data.
[0010] Preferably, the types of the electrical loads include: industrial loads, agricultural loads, municipal domestic electrical loads, and transportation electrical loads.
[0011] Preferably, determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified includes:
[0012] Substituting the load identification data of the bus to be identified into a pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified.
[0013] Furthermore, the training process of the pre-established LSTM neural network includes:
[0014] Collecting the load power time series data corresponding to various types of electrical loads on the bus, the corresponding time data, and meteorological data at the historical moment of the bus;
[0015] Successively preprocessing the load power time series data corresponding to various types of electrical loads on the bus, the corresponding time data, and meteorological data at the historical moment of the bus using the maximum value normalization method, and performing clustering analysis and data fitting using the fuzzy clustering algorithm based on particle swarm optimization with the Davies-Bouldin index as the evaluation index for clustering effectiveness to obtain the load identification data of the virtual bus.
[0016] Taking a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and taking the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus as the output layer training samples of the initial LSTM neural network, and training the initial LSTM neural network to obtain the pre-established LSTM neural network.
[0017] Furthermore, the verification process of the pre-established LSTM neural network includes:
[0018] Substituting another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network.
[0019] Comparing the output data with the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus. If the relative error between the output data and the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus is less than a preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.
[0020] A bus load identification system, the improvement is that the system includes:
[0021] An acquisition module for acquiring the load identification data of the bus to be identified;
[0022] A determination module, configured to determine the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified;
[0023] Wherein, the load identification data includes: power time series data, time data, and meteorological data.
[0024] Preferably, the types of the electrical loads include: industrial loads, agricultural loads, municipal domestic electricity loads, and transportation electricity loads.
[0025] Preferably, the determination module is configured to:
[0026] Substitute the load identification data of the bus to be identified into a pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified.
[0027] Further, the training process of the pre-established LSTM neural network includes:
[0028] Collect the load power time series data corresponding to various types of electrical loads on the bus, and their corresponding time data and meteorological data at the historical moment of the bus;
[0029] Successively preprocess the load power time series data corresponding to various types of electrical loads on the bus, and their corresponding time data and meteorological data at the historical moment of the bus by using the maximum value normalization method, and perform clustering analysis and data fitting by using the fuzzy clustering algorithm based on particle swarm optimization with the Davies-Bouldin index as the evaluation index of clustering effectiveness to obtain the load identification data of the virtual bus;
[0030] Use a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and use the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus as the output layer training samples of the initial LSTM neural network to train the initial LSTM neural network to obtain the pre-established LSTM neural network.
[0031] Further, the verification process of the pre-established LSTM neural network includes:
[0032] Substitute another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network;
[0033] Compare the output data with the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus. If the relative error between the output data and the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus is less than the preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.
[0034] Compared with the closest prior art, the beneficial effects of the present invention are:
[0035] The technical solution provided by the present invention collects the load identification data of the bus to be identified; determines the load sequences corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified, and realizes the accurate identification of the load types of large-scale and diversified loads through the learning ability of the LSTM neural network. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 is a flowchart of a bus load identification method provided by the present invention;
[0037] Figure 2 is a structure diagram of a multi-layer LSTM neural network for bus load identification provided by an embodiment of the present invention;
[0038] Figure 3 is a structure diagram of a bus load identification system provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0039] The following further details the specific embodiments of the present invention with reference to the accompanying drawings.
[0040] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0041] The present invention provides a bus load identification method. Utilizing the advantages of big data analysis and artificial intelligence in dealing with complex data with non-linearity and interactive influence, first, clustering analysis is performed on various types of load data of the power grid to be analyzed, so as to obtain the electrical characteristics of the main typical loads in this area and construct a classification typical load electrical characteristic library; on this basis, multi-source information such as external meteorology, season, date, and time is analyzed to form a multi-information source for effectively identifying the bus load composition; finally, based on the LSTM neural network, the load types included in a specific large power supply node are identified, such as Figure 1As shown, the method includes:
[0042] Step 101, collect the load identification data of the bus to be identified;
[0043] Step 102, determine the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified;
[0044] Among them, the load identification data includes: power time series data, time data, and meteorological data.
[0045] The types of the electrical loads include: industrial load, agricultural load, municipal domestic electricity load, and transportation electricity load.
[0046] The step 102, as Figure 2 shown, includes:
[0047] Substitute the load identification data of the bus to be identified into the pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified.
[0048] The training process of the pre-established LSTM neural network includes:
[0049] Collect the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus, and their corresponding time data and meteorological data;
[0050] Successively preprocess the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus, and their corresponding time data and meteorological data by using the maximum value normalization method, and use the fuzzy clustering algorithm based on particle swarm optimization to perform clustering analysis and data fitting with the Davies-Bouldin index as the evaluation index of clustering effectiveness to obtain the load identification data of the virtual bus;
[0051] Use a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and use the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus as the output layer training samples of the initial LSTM neural network to train the initial LSTM neural network to obtain the pre-established LSTM neural network.
[0052] The verification process of the pre-established LSTM neural network includes:
[0053] Substitute another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network;
[0054] Compare the output data with the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus. If the relative error between the output data and the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus is less than a preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.
[0055] Based on the above solution, the optimal embodiment provided by the present invention may include the following steps:
[0056] Step 1: Collect the power time series data of various types of electrical loads on the bus in a standard year from the power acquisition system or the user energy management system; perform clustering analysis based on the power time series data of various types of electrical loads, and use the clustering centers of various loads as the characteristic curves of various typical loads.
[0057] Step 2: Collect the standard-year external meteorological information corresponding to the characteristic curve data of the loads in Step 1 from the meteorological data network (or each power grid meteorological database), and combine the time and the characteristic curves of various types of load power consumption associated with the time to form multi-source data for identifying the bus load in this area.
[0058] Step 3: Based on the multi-source data, form a training sample set and a verification sample set for analyzing the composition of the bus load, and train and verify the network parameters of the LSTM neural network respectively, specifically including a data preparation link, a data processing link, a network training link, and a network verification link.
[0059] Step 4: Use the load power time series data of the bus to be identified, its associated time, and external meteorological information as input data, and identify the load types included therein based on the LSTM neural network; the output of the LSTM neural network is the load change sequence corresponding to various types of electrical loads on the bus.
[0060] Among them, the specific steps of Step 1 are as follows:
[0061] Step S1-1: Collect the standard-year power time series data of the bus load in the corresponding climate zone from the power acquisition system or the user energy management system, and use the standard-year power time series data of the bus load as the input quantity for clustering analysis.
[0062] Step S1-2: Preprocess the data using the maximum value normalization method.
[0063] Step S1-3: Use the fuzzy clustering algorithm based on particle swarm optimization, and use the Davies-Bouldin index (DBI) as the evaluation index for clustering effectiveness to perform clustering analysis on various types of loads to obtain load clustering centers with obvious differences.
[0064] Step S1-4: Use the load clustering center as the typical load to form the characteristic curves of various typical loads.
[0065] Among them, the specific steps of Step 3 are as follows:
[0066] Step S3-1: Data preparation stage: Based on the multi-source data identified by the bus load in the standard year, multiply the characteristic curve data of various typical loads with the same time by different coefficients to simulate and synthesize the load identification data of a large number (not less than 10,000) of virtual buses with different load compositions. Take any 80% of the load identification data of the virtual bus as the training sample data of the initial LSTM neural network, and the remaining 20% of the data as the verification sample data of the pre-established LSTM neural network.
[0067] Step S3-2: Data processing stage: Filter, differentiate, and normalize the sample data to obtain the training sample set (X k , Y k ) and the verification sample set (X j , Y j ).
[0068] Among them, X k is the training sample input data of the initial LSTM neural network, representing the time series data of the load power on the bus at the k-th moment and its corresponding external meteorological data, k = 1, 2,..., n; is the training sample output data of the initial LSTM neural network, representing the time series data of the load power of various loads on the bus at the k-th moment; X j is the test sample input data of the pre-established LSTM neural network, representing the time series data of the load power on the bus at the j-th moment and its corresponding external meteorological data, j = 1, 2,..., n; is the test sample data of the pre-established LSTM neural network, representing the time series data of the load power of various loads on the bus at the j-th moment.
[0069] Step S3-3: Network training stage: Use the training sample set (X k , Y k ) to train the initial LSTM neural network to obtain the pre-established LSTM neural network.
[0070] Step S3-4: Network verification stage: Input X j into the pre-established LSTM neural network to obtain the output data, and use the output data to compare with Y j in the verification sample set (X j , Y j ) to analyze the error.
[0071] The present invention also provides a bus load identification system, asFigure 3 As shown, the system includes:
[0072] A collection module for collecting load identification data of the bus to be identified;
[0073] A determination module for determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified;
[0074] Among them, the load identification data includes: power time series data, time data, and meteorological data.
[0075] The types of the electrical loads include: industrial load, agricultural load, municipal domestic electricity load, and transportation electricity load.
[0076] The determination module is used for:
[0077] Substituting the load identification data of the bus to be identified into a pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified.
[0078] The training process of the pre-established LSTM neural network includes:
[0079] Collecting the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus, as well as the corresponding time data and meteorological data;
[0080] Successively preprocessing the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus, as well as the corresponding time data and meteorological data by using the maximum value normalization method, and using the fuzzy clustering algorithm based on particle swarm optimization, and performing clustering analysis and data fitting with the Davies-Bouldin index as the evaluation index of clustering effectiveness to obtain the load identification data of the virtual bus;
[0081] Taking a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and taking the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus as the output layer training samples of the initial LSTM neural network, and training the initial LSTM neural network to obtain the pre-established LSTM neural network.
[0082] The verification process of the pre-established LSTM neural network includes:
[0083] Substituting another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network;
[0084] Compare the output data with the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus. If the relative error between the output data and the load power time series data corresponding to various types of electrical loads on the bus at the corresponding historical moment of the bus is less than a preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.
[0085] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0086] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or a device for the functions specified in multiple blocks.
[0087] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or a device for the functions specified in multiple blocks.
[0088] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the specified functions in one process Figure 1 one process or multiple processes and / or blocks Figure 1 or a device for the functions specified in multiple blocks.
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent substitutions, and any modification or equivalent substitution that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.
Claims
1. A method for identifying bus loads, characterized in that, The method includes: Collecting the load identification data of the bus to be identified; Determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified; Among them, the load identification data includes: power time series data, time data, and meteorological data; The determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified includes: Substituting the load identification data of the bus to be identified into a pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified; The training process of the pre-established LSTM neural network includes: Collecting the load power time series data corresponding to various types of electrical loads on the bus, the corresponding time data, and meteorological data at the historical moment of the bus; Successively preprocessing the load power time series data corresponding to various types of electrical loads on the bus, the corresponding time data, and meteorological data at the historical moment of the bus by using the maximum value normalization method, and using the fuzzy clustering algorithm based on particle swarm optimization, and performing clustering analysis and data fitting with the Davies-Bouldin index as the evaluation index of clustering effectiveness to obtain the load identification data of the virtual bus; Taking a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and taking the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus as the output layer training samples of the initial LSTM neural network, and training the initial LSTM neural network to obtain the pre-established LSTM neural network.
2. The method according to claim 1, characterized in that, The types of the electrical loads include: industrial loads, agricultural loads, municipal domestic electricity loads, and transportation electricity loads.
3. The method according to claim 1, characterized in that, The verification process of the pre-established LSTM neural network includes: Substituting another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network; Comparing the output data with the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus. If the relative error between the output data and the load power time series data corresponding to various types of electrical loads on the bus at the historical moment of the bus is less than a preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.
4. A system for identifying bus loads, characterized in that, The system includes: A collection module for collecting the load identification data of the bus to be identified; A determination module for determining the load power time series data corresponding to various types of electrical loads on the bus to be identified according to the load identification data of the bus to be identified; Among them, the load identification data includes: power time series data, time data, and meteorological data; The determination module is used for: Substituting the load identification data of the bus to be identified into a pre-established LSTM neural network to obtain the load power time series data corresponding to various types of electrical loads on the bus to be identified; The training process of the pre-established LSTM neural network includes: Collect the time-series data of the load powers corresponding to various types of electrical loads on the bus, the corresponding time data, and the meteorological data at the historical moments of the bus. Successively preprocess the time-series data of the load powers corresponding to various types of electrical loads on the bus, the corresponding time data, and the meteorological data at the historical moments of the bus using the maximum normalization method, and perform clustering analysis and data fitting using the fuzzy clustering algorithm based on particle swarm optimization with the Davies-Bouldin index as the evaluation index for clustering effectiveness to obtain the load identification data of the virtual bus. Use a part of the load identification data in the load identification data of the virtual bus as the input layer training samples of the initial LSTM neural network, and use the time-series data of the load powers corresponding to various types of electrical loads on the bus at the historical moments of the bus as the output layer training samples of the initial LSTM neural network to train the initial LSTM neural network and obtain the pre-established LSTM neural network.
5. The system according to claim 4, characterized in that, The types of the electrical loads include: industrial loads, agricultural loads, municipal domestic electrical loads, and transportation electrical loads.
6. The system according to claim 4, characterized in that, The verification process of the pre-established LSTM neural network includes: Substitute another part of the load identification data in the load identification data of the virtual bus into the pre-established LSTM neural network to obtain the output data of the pre-established LSTM neural network. Compare the output data with the time-series data of the load powers corresponding to various types of electrical loads on the bus at the historical moments of the bus. If the relative error between the output data and the time-series data of the load powers corresponding to various types of electrical loads on the bus at the historical moments of the bus is less than the preset threshold, the pre-established LSTM neural network is qualified; otherwise, the pre-established LSTM neural network is unqualified.