Harmonic responsibility intelligent division and compensation optimization system and method

Abstract

The invention relates to the technical field of power grid management, in particular to a harmonic wave responsibility intelligent division and compensation optimization system and method, and the method comprises the following steps: S1, dividing a power grid region, and providing a basic framework for subsequent harmonic wave monitoring and responsibility division; s2, harmonic monitoring and data acquisition: providing reliable data support for subsequent analysis: deploying corresponding monitoring equipment at different levels based on a region division result in the step S1, acquiring harmonic data, preprocessing the harmonic data, and outputting high-quality data for subsequent steps; according to the invention, through three-level regional layering and multi-device monitoring, data is preprocessed in combination with PAA and ShapeDTW algorithms, accurate acquisition and asynchronous sampling alignment of power grid harmonic data are realized, and reliable data support is provided for subsequent analysis; a hierarchical compensation strategy is adopted, multi-agent reinforcement learning and syn-position compensation scheduling are combined, dynamic optimization compensation of harmonic waves is achieved, the distortion rate of the harmonic waves is reduced, and meanwhile the energy consumption cost is reduced by 18% compared with a traditional scheme.

Description

Technical Field

[0001] This invention relates to the field of power grid management technology, and in particular to a system and method for intelligent allocation and compensation optimization of harmonic responsibility. Background Technology

[0002] With the development of science and technology, various nonlinear and time-varying electronic devices, such as inverters, rectifiers, and switching power supplies, are widely used, resulting in a significant increase in harmonic components in power systems. Their negative effects are becoming increasingly apparent. "Harmonic pollution" has become one of the main factors affecting power quality; therefore, harmonic control has become an urgent requirement for the development of modern power production.

[0003] The purpose of harmonic liability delineation is to clarify the responsibility of each entity in the power grid for harmonic pollution, providing a basis for defining responsibility in harmonic mitigation. Compensation is necessary because harmonics can increase losses and shorten the lifespan of electrical equipment, and even affect the safe and stable operation of the power grid.

[0004] Traditional methods for intelligent allocation and compensation optimization of harmonic responsibility have shortcomings such as insufficient data acquisition accuracy, difficulty in aligning asynchronous sampling data, low accuracy in tracing and locating harmonic sources, inaccurate responsibility quantification models, lack of dynamic adjustment mechanisms for compensation strategies, and high energy consumption costs.

[0005] To address this, this invention proposes a harmonic responsibility intelligent division and compensation optimization system and method. Through a three-level regional hierarchical monitoring and data preprocessing algorithm, it achieves accurate acquisition and alignment of harmonic data; with the help of blind source separation and graph neural network algorithm, it achieves efficient source tracing and localization of harmonic sources; and by adopting a hierarchical compensation strategy and multi-agent reinforcement learning, it achieves dynamic optimization compensation and energy consumption cost control of harmonics. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a system and method for intelligent division and compensation optimization of harmonic responsibility, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The intelligent allocation and compensation optimization method for harmonic responsibility includes the following steps:

[0009] S1. The division of power grid areas provides a basic framework for subsequent harmonic monitoring and responsibility allocation;

[0010] S2. Harmonic monitoring and data acquisition provide reliable data support for subsequent analysis: Based on the regional division results of step S1, corresponding monitoring equipment is deployed at different levels to collect harmonic data. After preprocessing, high-quality data is output for subsequent steps.

[0011] S3. Harmonic source tracing and location, providing locational basis for liability determination: Based on the preprocessed data collected in step S2, the mixed harmonic signals are separated and the harmonic source locations are located, clarifying the source area of ​​harmonic pollution;

[0012] S4. Harmonic liability determination provides a quantitative basis for compensation strategy: Based on the location of the harmonic source determined in step S3, combined with the monitoring data in step S2, the harmonic impedance matrix is ​​calculated through Norton equivalent circuit, and then the weighted contribution method is used to calculate the responsibility ratio of each area.

[0013] S5. Compensation optimization strategy: Targeted compensation based on responsibility ratio: According to the responsibility ratio calculated in step S4, intra-regional compensation, near-regional cross-regional compensation, and cross-regional dynamic compensation are executed respectively. When there is no compensation capacity in adjacent areas, priority compensation is executed.

[0014] Preferably, in step S1, the main grid layer is divided with 500kV / 220kV hub substations as the core, covering the provincial power grid area; the zoning layer is divided with 110kV / 35kV substations as units, and sub-regions are divided according to power supply radius ≤15km, with each zone corresponding to an independent power supply area; the distribution network layer is based on 10kV feeders as basic units, and each distribution network sub-region contains 1-3 distribution transformers.

[0015] Preferably, the harmonic monitoring and data acquisition in step S2 are as follows, and the output data of this step directly provides input for steps S3 and S4:

[0016] Monitoring equipment deployment: PMUs are configured in hub substations at the main grid layer, FTUs / DTUs are deployed in regional substations at the zoning layer, and smart meters and harmonic monitoring terminals are installed on the user side at the distribution network layer.

[0017] Data preprocessing: The PAA dimensionality reduction algorithm is used to reduce the length of the data. time series Compressed to a length of sequence The compression formula is: , in, To compress the window length;

[0018] The Shape Dynamic Time Warping (ShapeDTW) algorithm is used to compute two time series. and similarity distance To solve the problem of asynchronous sampling: , in, This represents the optimal alignment path for the time series. This represents the path length.

[0019] Preferably, the harmonic source tracing and location in step S3 is as follows, and the location result of this step provides a location basis for the responsibility determination in step S4:

[0020] The FastICA algorithm is used to separate mixed harmonic signals: Let... for Data collected from each monitoring point 3D mixed signal matrix, ,in It is a mixed matrix. For independent source signal matrices;

[0021] Initialize the separation matrix Through iterative updates: ;in, It is a nonlinear function. It is a constant. For expectation operation; when Upon convergence, the estimated source signal is obtained. ;

[0022] Using graph convolutional networks to locate harmonic sources: Modeling the power grid topology as a graph Node features Includes node voltage and current harmonic components, edge characteristics Represents line impedance; feature propagation is performed using a graph convolutional network. ,in, For nodes The set of neighboring nodes, The normalization constant is This is the weight matrix. For bias vectors, The activation function is used to predict the location of harmonic sources using a node classifier.

[0023] The location result is combined with the monitoring data from step S2 and used in step S4 to calculate the responsibility percentage of the corresponding area.

[0024] Preferably, the harmonic source tracing and location in step S3 is as follows, and the location result of this step provides a location basis for the responsibility determination in step S4:

[0025] Based on the Norton equivalent circuit, harmonic voltage is measured at the PCC point. Harmonic currents of each feeder ;

[0026] Calculate the harmonic impedance matrix ,in, The weighted contribution method was used to calculate the proportion of responsibility. ;

[0027] The calculated liability percentage is directly used as the basis for selecting a compensation strategy in step S5.

[0028] Preferably, the compensation optimization strategy in step S5 is as follows, and this step depends entirely on the responsibility ratio result in step S4:

[0029] Compensation within the area of ​​minor responsibility (percentage of liability) ):

[0030] An LC single-tuned filter was selected and designed for the dominant harmonic frequency. The filter parameters, including its capacitance value, were calculated based on the harmonic source capacity and grid parameters. The calculation formula is: ,in, For harmonic frequencies, Inductance value; Inductance value Must meet ,and , ;

[0031] Cross-regional compensation for similar liabilities (proportion of liability) ):

[0032] The state of each regional agent includes its own harmonic distortion rate, remaining APF capacity, THD and APF states of neighboring regions; APF compensation current adjustment range; reward function: ,in Harmonic distortion weight, Cost weighting The energy consumption cost of APF operation; the agent adjusts the APF compensation strategy based on reward feedback through interaction with the environment, and gradually optimizes the collaborative scheme;

[0033] Cross-regional dynamic compensation for major liabilities (responsibility ratio) ):

[0034] The power grid includes Each region For each harmonic frequency, the compensation capability matrix is... middle Indicates the region harmonic frequencies The compensation capacity; calculate the compensation requirements for each frequency of the harmonic source. Prioritize locations close to harmonic sources and Compensation will be provided to the selected region. If multiple regions meet the criteria, then the selected region will be chosen. The region with the largest value;

[0035] Priority compensation (responsibility percentage) ):

[0036] When adjacent areas are unable to compensate, they are sorted by their distance from the harmonic source, with closer areas having higher priority, and compensation is carried out in order of priority; if the distances are the same, areas with larger compensation capacity are selected first.

[0037] The intelligent harmonic responsibility allocation and compensation optimization system is used to execute the aforementioned intelligent harmonic responsibility allocation and compensation optimization method. It includes a data acquisition module, a data analysis module, and an execution module, with data flow and method steps corresponding one-to-one between the three modules.

[0038] The data acquisition module, corresponding to steps S1 and S2, is used to realize the three-level regional division of the power grid and deploy monitoring equipment to collect harmonic data. After PAA dimensionality reduction and ShapeDTW alignment preprocessing, high-quality data is output.

[0039] The data analysis module, corresponding to steps S3 and S4, includes a source tracing unit and a responsibility determination unit: the source tracing unit uses the FastICA algorithm and graph neural network to separate and locate harmonic sources, and the responsibility determination unit calculates the responsibility ratio based on the source tracing results and preprocessed data using the harmonic impedance matrix and weighted contribution model.

[0040] The execution module, corresponding to step S5, receives the responsibility ratio result output by the data analysis module, and executes the hierarchical compensation strategy and sequential compensation scheduling based on the result to complete the harmonic compensation optimization.

[0041] The output of the data acquisition module serves as the input of the data analysis module, and the responsibility determination results of the data analysis module serve as the input of the execution module. The modules work together to achieve a complete process of harmonic responsibility allocation and compensation optimization.

[0042] Beneficial effects compared to existing technologies:

[0043] 1. This invention achieves accurate acquisition and asynchronous sampling alignment of power grid harmonic data by using a three-level regional layering (main grid layer, regional layer, distribution network layer) and multi-device monitoring (PMU, FTU / DTU, etc.), combined with PAA and ShapeDTW algorithm preprocessing, thus providing reliable data support for subsequent analysis;

[0044] 2. This invention separates mixed harmonic signals using the FastICA algorithm and combines GNN to model the power grid topology and locate harmonic sources, achieving efficient source tracing and localization of harmonic sources. When the GCN is stacked with 3 layers, the source tracing accuracy reaches 92%, improving the accuracy of responsibility determination.

[0045] 3. This invention uses a hierarchical compensation strategy (different strategies for small, medium, and large responsibility areas), combined with multi-agent reinforcement learning and sequential compensation scheduling, to achieve dynamic optimization compensation of harmonics, reducing harmonic distortion rate while reducing energy consumption costs by 18% compared to traditional solutions. Attached Figure Description

[0046] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0047] Figure 1 This is a schematic diagram of the steps of the present invention;

[0048] Figure 2 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0049] Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention can also be implemented in various different forms, and therefore the present invention is not limited to the embodiments described below.

[0050] The technical solution in this application embodiment is to solve the problems mentioned in the background art, and the overall idea is as follows:

[0051] The intelligent allocation and compensation optimization method for harmonic responsibility includes the following steps:

[0052] S1. The division of power grid areas provides a basic framework for subsequent harmonic monitoring and responsibility allocation;

[0053] The main grid layer is divided into core sections based on 500kV / 220kV hub substations, covering the provincial power grid area; the sub-regional layer is divided into sub-regions based on 110kV / 35kV substations, with each sub-region corresponding to an independent power supply area; the distribution network layer is based on 10kV feeders as basic units, with each distribution network sub-region containing 1-3 distribution transformers.

[0054] S2. Harmonic monitoring and data acquisition provide reliable data support for subsequent analysis: Based on the regional division results of step S1, corresponding monitoring equipment is deployed at different levels to collect harmonic data. After preprocessing, high-quality data is output for subsequent steps.

[0055] Monitoring equipment deployment: PMUs are configured in hub substations at the main grid layer, FTUs / DTUs are deployed in regional substations at the zoning layer, and smart meters and harmonic monitoring terminals are installed on the user side at the distribution network layer.

[0056] Data preprocessing: The PAA dimensionality reduction algorithm is used to reduce the length of the data. time series Compressed to a length of sequence The compression formula is: , in, To compress the window length,

[0057] The Shape Dynamic Time Warping (ShapeDTW) algorithm is used to compute two time series. and similarity distance To solve the problem of asynchronous sampling:

[0058] ,in, This represents the optimal alignment path for the time series. This represents the path length.

[0059] S3. Harmonic source tracing and location, providing locational basis for liability determination: Based on the preprocessed data collected in step S2, the mixed harmonic signals are separated and the harmonic source locations are located, clarifying the source area of ​​harmonic pollution;

[0060] The FastICA algorithm is used to separate mixed harmonic signals: Let... for Data collected from each monitoring point 3D mixed signal matrix, ,in It is a mixed matrix. For independent source signal matrices;

[0061] Initialize the separation matrix Through iterative updates: ;in, It is a nonlinear function. It is a constant. For expectation operation; when Upon convergence, the estimated source signal is obtained. ;

[0062] Using graph convolutional networks to locate harmonic sources: Modeling the power grid topology as a graph Node features Includes node voltage and current harmonic components, edge characteristics Represents line impedance; feature propagation is performed using a graph convolutional network. ,in, For nodes The set of neighboring nodes, The normalization constant is This is the weight matrix. For bias vectors, The activation function is used to predict the location of harmonic sources using a node classifier.

[0063] The location result is combined with the monitoring data from step S2 and used in step S4 to calculate the responsibility percentage of the corresponding area.

[0064] S4. Harmonic liability determination provides a quantitative basis for compensation strategies: Based on the location of the harmonic source determined in step S3, combined with the monitoring data in step S2, the harmonic impedance matrix is ​​calculated through the Norton equivalent circuit, and then the weighted contribution method is used to calculate the liability ratio of each region.

[0065] Based on the Norton equivalent circuit, harmonic voltage is measured at the PCC point. Harmonic currents of each feeder ;

[0066] Calculate the harmonic impedance matrix ,in, The weighted contribution method was used to calculate the proportion of responsibility. ;

[0067] The calculated liability percentage is directly used as the basis for selecting a compensation strategy in step S5.

[0068] S5. Compensation optimization strategy: Targeted compensation based on responsibility ratio: According to the responsibility ratio calculated in step S4, intra-regional compensation, near-regional cross-regional compensation, and cross-regional dynamic compensation are executed respectively. When there is no compensation capacity in adjacent regions, priority compensation is executed.

[0069] implement:

[0070] Compensation within the area of ​​minor responsibility:

[0071] An LC single-tuned filter was selected and designed for the dominant harmonic frequency. The filter parameters, including its capacitance value, were calculated based on the harmonic source capacity and grid parameters. The calculation formula is: ,in, For harmonic frequencies, Inductance value; Inductance value Must meet ,and , ;

[0072] Cross-regional compensation for similar responsibilities:

[0073] The state of each regional agent includes its own harmonic distortion rate, remaining APF capacity, THD and APF states of neighboring regions; APF compensation current adjustment range; reward function: ,in Harmonic distortion weight, Cost weighting The energy consumption cost of APF operation; the agent adjusts the APF compensation strategy based on reward feedback through interaction with the environment, and gradually optimizes the collaborative scheme;

[0074] Dynamic compensation for cross-regional major liabilities:

[0075] The power grid includes Each region For each harmonic frequency, the compensation capability matrix is... middle Indicates the region harmonic frequencies The compensation capacity; calculate the compensation requirements for each frequency of the harmonic source. Prioritize locations close to harmonic sources and Compensation will be provided to the selected region. If multiple regions meet the criteria, then the selected region will be chosen. The region with the largest value;

[0076] Rank compensation:

[0077] When adjacent areas are unable to compensate, they are sorted by their distance from the harmonic source, with closer areas having higher priority, and compensation is carried out in order of priority; if the distances are the same, areas with larger compensation capacity are selected first.

[0078] The intelligent harmonic responsibility allocation and compensation optimization system, used to execute the intelligent harmonic responsibility allocation and compensation optimization method according to any one of claims 1-5, includes a data acquisition module, a data analysis module, and an execution module, with data flow and method steps corresponding one-to-one between the three modules:

[0079] The data acquisition module, corresponding to steps S1 and S2, is used to realize the three-level regional division of the power grid and deploy monitoring equipment to collect harmonic data. After PAA dimensionality reduction and ShapeDTW alignment preprocessing, high-quality data is output.

[0080] The data analysis module, corresponding to steps S3 and S4, includes a source tracing unit and a responsibility determination unit: the source tracing unit uses the FastICA algorithm and graph neural network to separate and locate harmonic sources, and the responsibility determination unit calculates the responsibility ratio based on the source tracing results and preprocessed data using the harmonic impedance matrix and weighted contribution model.

[0081] The execution module, corresponding to step S5, receives the responsibility ratio result output by the data analysis module, and executes the hierarchical compensation strategy and sequential compensation scheduling based on the result to complete the harmonic compensation optimization.

[0082] The output of the data acquisition module serves as the input of the data analysis module, and the responsibility determination results of the data analysis module serve as the input of the execution module. The modules work together to achieve a complete process of harmonic responsibility allocation and compensation optimization.

[0083] Example 1:

[0084] Please refer to Figure 1 As shown in the figure, this embodiment introduces a method for intelligent allocation and compensation optimization of harmonic responsibility. The specific steps of the method are as follows:

[0085] I. Power Grid Regional Division

[0086] 1.1 Regional Hierarchy Rules

[0087] Main grid layer: Covers 500kV and 220kV hub substations, divided by provincial power grid, with each main grid area containing 3-5 hub nodes;

[0088] Zoning: Based on the coverage of 110kV and 35kV substations, the main grid area is subdivided into 5-10 zones, each zone corresponding to an independent power supply area;

[0089] Distribution network layer: Based on 10kV feeders and 0.4kV distribution areas, the zone is further divided into several sub-regions, each containing 1-3 distribution transformers;

[0090] 1.2 Algorithm for Determining Region Boundaries

[0091] The minimum spanning tree (MST) algorithm combined with the power grid topology is used to construct the region boundary:

[0092] Let the set of power grid nodes be The route set is Each line weight This refers to the line length or impedance value.

[0093] Calculate the minimum spanning tree using Kruskal's algorithm. To ensure optimal connectivity among nodes within the region;

[0094] Based on voltage level and power supply range constraints, Divided into different hierarchical areas;

[0095] Example: A municipal power grid consists of 100 nodes and 150 lines. Using the MST algorithm, it is divided into 2 main grid areas, 8 sub-grid areas, and 30 distribution grid areas, with a boundary node error rate of <3%.

[0096] II. Harmonic Monitoring and Data Acquisition

[0097] 2.1 Deployment of Monitoring Equipment

[0098] Main grid layer: PMUs are configured in hub substations, with a sampling frequency of 10kHz, to monitor harmonics from the 2nd to the 50th order;

[0099] Zonal layer: FTU / DTU is deployed in regional substations to collect harmonic data every 5 minutes;

[0100] Distribution network layer: Install smart meters and harmonic monitoring terminals on the user side to upload current and voltage waveform data in real time;

[0101] 2.2 Data Preprocessing

[0102] The Piecewise Aggregate Approximation (PAA) algorithm and the Shape Dynamic Time Warping (ShapeDTW) algorithm are employed.

[0103] PAA dimensionality reduction: Dimensionality reduction of length... time series Compressed to a length of sequence The formula is: ,in, To compress the window length;

[0104] ShapeDTW Alignment: Calculating Two Time Series and similarity distance To solve the problem of asynchronous sampling: ,in, This represents the optimal alignment path for the time series. This represents the path length.

[0105] III. Harmonic Source Tracing and Location

[0106] 3.1 Blind Source Separation (BSS) Algorithm

[0107] The FastICA algorithm is used to separate mixed harmonic signals:

[0108] set up for Data collected from each monitoring point 3D mixed signal matrix, ,in It is a mixed matrix. For independent source signal matrices;

[0109] Initialize the separation matrix Through iterative updates: ,in, It is a nonlinear function. It is a constant. For expectation calculation;

[0110] when Upon convergence, the estimated source signal is obtained. ;

[0111] Example: In an industrial park, the FastICA algorithm was used to separate three main harmonic sources from the mixed signal at the PCC point, with contribution rates of 45%, 30%, and 25%, respectively.

[0112] 3.2 Graph Neural Network (GNN) Localization

[0113] Model the power grid topology as a graph. Node features Includes node voltage and current harmonic components, edge characteristics Indicates line impedance;

[0114] Feature propagation is performed using a graph convolutional network (GCN): ,in, For nodes The set of neighboring nodes, The normalization constant is This is the weight matrix. For bias vectors, For activation functions;

[0115] The accuracy of predicting the location of harmonic sources using a node classifier can reach 92%.

[0116] IV. Harmonic Liability Determination

[0117] 4.1 Calculation of Harmonic Impedance Matrix

[0118] Based on the Norton equivalent circuit, the harmonic voltage is measured at the PCC point. Harmonic currents of each feeder ;

[0119] Calculate the harmonic impedance matrix ,in, ;

[0120] 4.2 Responsibility Quantification Model

[0121] The weighted contribution method is used to calculate the proportion of responsibility: ;

[0122] Example: At point PCC of a substation, the harmonic currents of users A, B, and C are 20A, 15A, and 10A respectively, with corresponding self-impedances of 5Ω, 3Ω, and 2Ω. What is the percentage of responsibility for user A? ;

[0123] V. Compensation and Optimization Strategies

[0124] 5.1 Compensation within the area of ​​minor liability (liability percentage < 30%)

[0125] (1) Selection and parameter design of compensation equipment

[0126] Equipment selection: LC single-tuned filters should be given priority and designed for the dominant harmonic frequencies (such as the 5th and 7th harmonics); for example, if the 5th harmonic current accounts for the highest proportion in the region, then a 5th harmonic LC single-tuned filter should be configured.

[0127] Parameter calculation: Calculate filter parameters based on harmonic source capacity and power grid parameters.

[0128] capacitance value Calculation formula: ,in, For harmonic frequencies, This is the inductance value;

[0129] Inductance value Must meet (Resonance condition), and , ;

[0130] (2) Implementation Case

[0131] For example, in a small commercial park, the harmonic liability rate was determined to be 25%, and the dominant harmonic was the 5th order (amplitude 15A).

[0132] Filter design: Assume the fundamental frequency of the power grid 5th harmonic frequency ,according to Calculated , ;

[0133] Configure a 5th order LC single-tuned filter with a rated voltage of 10kV and a rated capacity of 300kvar;

[0134] Compensation effect: After commissioning, the fifth harmonic current in the area was reduced to 3A, and the harmonic distortion rate was reduced from 8% to 3%, meeting the national standard requirements;

[0135] 5.2 Cross-regional compensation for similar liabilities (responsibility ratio 30-70%)

[0136] (1) Multi-agent reinforcement learning collaborative compensation

[0137] System modeling:

[0138] State space: The state of each agent in a region includes its own harmonic distortion rate (THD), APF remaining capacity (...). ), THD and APF status of adjacent areas;

[0139] Operating range: APF compensation current adjustment range, such as ±50A;

[0140] Reward function: ,in (Harmonic distortion weight) (Cost weighting) Energy consumption cost of APF operation;

[0141] Learning process: The agent interacts with the environment (changes in power grid harmonics) and adjusts the APF compensation strategy based on reward feedback, gradually optimizing the collaborative scheme;

[0142] (2) Implementation Case

[0143] The harmonic liability rate of a certain industrial park is 55%, and each of the three adjacent areas is equipped with a 500A APF;

[0144] Strategy execution:

[0145] Initial state: Industrial park THD=12%, adjacent areas have THDs of 8%, 6%, and 7% respectively, and APF remaining capacity is 80% in all areas;

[0146] Intelligent agent decision-making: The APF compensation current in the industrial park is increased by 30A, and the compensation current in adjacent areas 1, 2, and 3 is increased by 20A, 15A, and 18A respectively;

[0147] Compensation effect: After 30 minutes, the THD in the industrial park dropped to 5%, and the THD in the adjacent area remained within the national standard range. The overall energy consumption cost was reduced by 18% compared with the traditional compensation scheme.

[0148] 5.3 Dynamic compensation for major liabilities across regions (liability ratio > 70%)

[0149] (1) Task allocation based on compensation capability matrix

[0150] Construction of the compensation capability matrix:

[0151] Assume the power grid includes Each region For each harmonic frequency, the compensation capability matrix is... middle Indicates the region harmonic frequencies Compensation capacity (unit: A); for example, This indicates that the compensation capacity for the 5th harmonic in region 3 is 200A;

[0152] Allocation strategy:

[0153] Calculate the compensation requirements for each frequency of the harmonic source. Prioritize locations close to harmonic sources and Compensation will be provided to the selected region. If multiple regions meet the criteria, then the selected region will be chosen. The region with the largest value;

[0154] (2) Implementation Case

[0155] A large steel plant is responsible for 80% of its harmonic emissions, with the dominant harmonics being the 5th (requiring a compensation current of 300A) and the 7th (requiring a compensation current of 200A). The compensation capacity matrix for the surrounding area is as follows:

[0156] area 5th harmonic compensation capacity (A) 7th harmonic compensation capacity (A) Distance (km) Area 1 150 100 2 Area 2 200 120 3 Area 3 350 250 5

[0157] Task assignment:

[0158] 5th harmonic: Prioritized allocation to region 3 (although far away, the capacity is sufficient);

[0159] 7th harmonic: allocated to region 3 (remaining capacity can cover demand);

[0160] Compensation effect: After joint compensation, the 5th harmonic current of the steel plant decreased from 350A to 50A, the 7th harmonic current decreased from 230A to 30A, and the harmonic distortion rate of the entire network met the standard.

[0161] 5.4 Priority compensation in case of inability to compensate

[0162] (1) Construction of the compensation priority queue

[0163] Sort regions by distance from harmonic sources, with closer regions having higher priority; if the distances are the same, prioritize regions with larger compensation capacity.

[0164] (2) Implementation Case

[0165] A certain new energy power plant is responsible for 85% of the harmonic load, requiring compensation for a 400A fifth harmonic current. The surrounding area is as follows:

[0166] Area A: 1km away, compensation capacity 300A (insufficient);

[0167] Area B: Distance 2km, compensation capacity 450A;

[0168] Area C: 3km away, compensation capacity 500A;

[0169] The process of priority compensation:

[0170] First try region A; if the capacity is insufficient, select region B from the queue to compensate.

[0171] Effect: After compensation in Region B, the fifth harmonic current of the new energy power station was reduced to a safe range, avoiding the need to apply for support from the higher-level region.

[0172] Example 2:

[0173] Please refer to Figure 2 As shown, the harmonic responsibility intelligent division and compensation optimization system implements the method described in Example 1. The system includes a data acquisition module, a data analysis module, and an execution module.

[0174] I. Data Acquisition Module

[0175] 1.1 Regional Division Units

[0176] (1) Hardware composition

[0177] Deployed on the dispatch center server, it carries a power grid topology database and stores node parameters (voltage level, impedance value) and line connection relationships.

[0178] (2) Core functions

[0179] Three-level regional stratification:

[0180] Main grid layer: Based on 500kV / 220kV hub substations, 3-5 main grid areas are automatically generated through the provincial power grid topology;

[0181] Zoning: Within the coverage area of ​​110kV / 35kV substations, sub-regions are divided according to a power supply radius of ≤15km;

[0182] Distribution network layer: Each sub-area consists of 1-3 distribution transformers, with 10kV feeders as the unit.

[0183] 1.2 Harmonic Monitoring Unit

[0184] (1) Hardware deployment

[0185] Main grid layer: Hub substations are equipped with PMU (phasor measurement unit) with a sampling rate of 10kHz, supporting 2nd-50th harmonic monitoring;

[0186] Zonal layer: FTU / DTU is deployed in regional substations to collect harmonic voltage / current data every 5 minutes;

[0187] Distribution network layer: Smart meters (0.2S level) and harmonic monitoring terminals are installed on the user side to upload waveform data in real time.

[0188] (2) Data preprocessing

[0189] Piecewise Aggregate Approximation (PAA) algorithm: Compresses a time series of length N into an n-point sequence. Dimensionality reduction formula: ( );

[0190] Shape Dynamic Time Warping (ShapeDTW): Solves the problem of asynchronous sampling alignment; distance calculation formula: It is applied to clock skew correction for different PMUs, with an alignment error of <10μs.

[0191] II. Data Analysis Module

[0192] 2.1 Traceability Unit

[0193] (1) Blind Source Separation (BSS) Subunit

[0194] FastICA Algorithm: Mixed Signal Model ,in For monitoring matrix, It is a mixed matrix. Independent source signals;

[0195] Iterative update of the separation matrix: In the formula It is a nonlinear function. The convergence speed is fastest at this time.

[0196] (2) Graph Neural Network (GNN) Subunit

[0197] Power grid topology modeling: constructing a graph structure Node features Includes harmonic voltage / current amplitude and phase, side characteristics Line impedance;

[0198] Feature propagation formula: ;in For neighboring nodes, As a normalized constant, the source tracing accuracy reaches 92% when the GCN stack has 3 layers.

[0199] 2.2 Responsibility Determination Unit

[0200] (1) Calculation of harmonic impedance matrix

[0201] Based on the Norton equivalent circuit, harmonic voltage is measured at the PCC point. With feeder current Construct the impedance matrix: , ;

[0202] (2) Weighted contribution model

[0203] Formula for calculating liability ratio: ;

[0204] III. Execution Module

[0205] 3.1 Compensation Execution Unit

[0206] (1) Hierarchical compensation strategy engine

[0207] Small liability area (<30%): Harmonic impedance matrix calculation, and measurement of harmonic voltage at PCC point based on Norton equivalent circuit. Harmonic currents of each feeder ; Calculate the harmonic impedance matrix ,in, The weighted contribution method was used to calculate the proportion of responsibility. ;

[0208] Mid-responsibility region (30%-70%): Multi-agent reinforcement learning collaborative compensation, the state of each agent in the region includes its own harmonic distortion rate (THD), APF remaining capacity (... ), adjacent area THD and APF status; APF compensation current adjustment range, such as ±50A; ,in (Harmonic distortion weight) (Cost weighting) The energy consumption cost of APF operation; the agent adjusts the APF compensation strategy based on reward feedback by interacting with the environment (changes in power grid harmonics), and gradually optimizes the collaborative scheme.

[0209] Large responsibility area (>70%): Calculate the compensation requirements for each frequency of the harmonic source. Prioritize locations close to harmonic sources and Compensation will be provided to the selected region. If multiple regions meet the criteria, then the selected region will be chosen. The region with the largest value;

[0210] (2) Sequential compensation scheduler

[0211] Priority queue rules: sort by distance from the harmonic source from nearest to farthest; when the distances are the same, prioritize the region with the larger compensation capacity.

[0212] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for intelligent allocation and compensation optimization of harmonic responsibility, characterized in that, Includes the following steps: S1. The division of power grid areas provides a basic framework for subsequent harmonic monitoring and responsibility allocation; S2. Harmonic monitoring and data acquisition provide reliable data support for subsequent analysis: Based on the regional division results of step S1, corresponding monitoring equipment is deployed at different levels to collect harmonic data. After preprocessing, high-quality data is output for subsequent steps. S3. Harmonic source tracing and location, providing locational basis for liability determination: Based on the preprocessed data collected in step S2, the mixed harmonic signals are separated and the harmonic source locations are located, clarifying the source area of ​​harmonic pollution; S4. Harmonic liability determination provides a quantitative basis for compensation strategy: Based on the location of the harmonic source determined in step S3, combined with the monitoring data in step S2, the harmonic impedance matrix is ​​calculated through Norton equivalent circuit, and then the weighted contribution method is used to calculate the responsibility ratio of each area. S5. Compensation optimization strategy: Targeted compensation based on responsibility ratio: According to the responsibility ratio calculated in step S4, intra-regional compensation, near-regional cross-regional compensation, and cross-regional dynamic compensation are executed respectively. When there is no compensation capacity in adjacent areas, priority compensation is executed.

2. The intelligent harmonic responsibility allocation and compensation optimization method as described in claim 1, characterized in that, In step S1, the main grid layer is divided with 500kV / 220kV hub substations as the core, covering the provincial power grid area; the zoning layer is divided with 110kV / 35kV substations as units, and sub-regions are divided according to power supply radius ≤15km, with each zone corresponding to an independent power supply area; the distribution network layer is based on 10kV feeders as basic units, and each distribution network sub-region contains 1-3 distribution transformers.

3. The intelligent harmonic responsibility allocation and compensation optimization method as described in claim 1, characterized in that, The harmonic monitoring and data acquisition in step S2 are as follows, and the output data of this step directly provides input for steps S3 and S4: Monitoring equipment deployment: PMUs are configured in hub substations at the main grid layer, FTUs / DTUs are deployed in regional substations at the zoning layer, and smart meters and harmonic monitoring terminals are installed on the user side at the distribution network layer. Data preprocessing: The PAA dimensionality reduction algorithm is used to reduce the length of the data. time series Compressed to a length of sequence The compression formula is: ,in, To compress the window length; The Shape Dynamic Time Warping (ShapeDTW) algorithm is used to compute two time series. and similarity distance Solving the problem of asynchronous sampling: ,in, This represents the optimal alignment path for the time series. This represents the path length.

4. The intelligent harmonic responsibility allocation and compensation optimization method as described in claim 1, characterized in that, The harmonic source tracing and location in step S3 are as follows, and the location result of this step provides the location basis for the responsibility determination in step S4: The FastICA algorithm is used to separate mixed harmonic signals: Let... for Data collected from each monitoring point 3D mixed signal matrix, ,in It is a mixed matrix. For independent source signal matrices; Initialize the separation matrix Through iterative updates: ;in, It is a nonlinear function. It is a constant. For expectation operation; when Upon convergence, the estimated source signal is obtained. ; Using graph convolutional networks to locate harmonic sources: Modeling the power grid topology as a graph Node features Includes node voltage and current harmonic components, edge characteristics Represents line impedance; feature propagation is performed using a graph convolutional network. ,in, For nodes The set of neighboring nodes, The normalization constant is This is the weight matrix. For bias vectors, The activation function is used to predict the location of harmonic sources using a node classifier. The location result is combined with the monitoring data from step S2 and used in step S4 to calculate the responsibility percentage of the corresponding area.

5. The intelligent harmonic responsibility allocation and compensation optimization method as described in claim 1, characterized in that, The harmonic source tracing and location in step S3 are as follows, and the location result of this step provides the location basis for the responsibility determination in step S4: Based on the Norton equivalent circuit, harmonic voltage is measured at the PCC point. Harmonic currents of each feeder ; Calculate the harmonic impedance matrix ,in, The weighted contribution method was used to calculate the proportion of responsibility. ; The calculated liability percentage is directly used as the basis for selecting a compensation strategy in step S5.

6. The intelligent harmonic responsibility allocation and compensation optimization method as described in claim 1, characterized in that, The compensation optimization strategy for step S5 is as follows, and this step depends entirely on the responsibility ratio result of step S4: Compensation within the area of ​​minor responsibility: An LC single-tuned filter was selected and designed for the dominant harmonic frequency. The filter parameters, including its capacitance value, were calculated based on the harmonic source capacity and grid parameters. The calculation formula is: ,in, For harmonic frequencies, Inductance value; Inductance value Must meet ,and , ; Cross-regional compensation for similar responsibilities: The state of each regional agent includes its own harmonic distortion rate, remaining APF capacity, THD and APF states of neighboring regions; APF compensation current adjustment range; reward function: ,in Harmonic distortion weighting Cost weighting The energy consumption cost of APF operation; the agent adjusts the APF compensation strategy based on reward feedback through interaction with the environment, and gradually optimizes the collaborative scheme; Dynamic compensation for cross-regional major liabilities: The power grid includes Each region For each harmonic frequency, the compensation capability matrix is... middle Indicates the region harmonic frequencies The compensation capacity; calculate the compensation requirements for each frequency of the harmonic source. Prioritize locations close to harmonic sources and Compensation will be provided to the selected region. If multiple regions meet the criteria, then the selected region will be chosen. The region with the largest value; Rank compensation: When adjacent areas are unable to compensate, they are sorted by their distance from the harmonic source, with closer areas having higher priority, and compensation is carried out in order of priority; if the distances are the same, areas with larger compensation capacity are selected first.

7. A harmonic responsibility intelligent allocation and compensation optimization system, characterized in that, The method for intelligent allocation and compensation optimization of harmonic responsibility as described in any one of claims 1-5 includes a data acquisition module, a data analysis module, and an execution module, with data flow and method steps corresponding one-to-one among the three modules: The data acquisition module is used to realize the three-level regional division of the power grid and deploy monitoring equipment to collect harmonic data. After PAA dimensionality reduction and ShapeDTW alignment preprocessing, high-quality data is output. The data analysis module includes a source tracing unit and a responsibility determination unit: the source tracing unit uses the FastICA algorithm and graph neural network to separate and locate harmonic sources, and the responsibility determination unit calculates the responsibility ratio based on the source tracing results and preprocessed data using the harmonic impedance matrix and weighted contribution model. The execution module receives the responsibility ratio results output by the data analysis module, executes the hierarchical compensation strategy and sequential compensation scheduling based on the results, and completes harmonic compensation optimization. The output of the data acquisition module serves as the input of the data analysis module, and the responsibility determination results of the data analysis module serve as the input of the execution module. The modules work together to achieve a complete process of harmonic responsibility allocation and compensation optimization.

Images