Copper smelting harmonic source positioning and self-adaptive suppression system
By deploying high-precision sensors in the power distribution system of a copper smelter and performing wavelet packet transform and time-frequency analysis, combined with support vector machine and dynamic equivalent circuit model, harmonic sources in copper smelting are identified and filter deployment is optimized. This solves the problems of low efficiency in harmonic source identification and incomplete mitigation measures in traditional methods, and achieves high-precision harmonic source localization and adaptive suppression.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YUNNAN COPPER CO LTD
- Filing Date
- 2025-11-20
- Publication Date
- 2026-04-14
Smart Images
Figure CN121863403A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of harmonic localization technology in copper smelting, specifically relating to a harmonic source localization and adaptive suppression system and method in copper smelting. Background Technology
[0002] Harmonic location technology in copper smelting refers to the technique of accurately locating the main equipment or area generating harmonics by collecting and analyzing voltage and current signals from various equipment nodes in the power distribution system of a copper smelter. Therefore, how to utilize advanced technologies to improve the intelligence and safety of harmonic location in copper smelting has become one of the urgent problems to be solved.
[0003] In the field of harmonic localization in copper smelting, traditional methods often require global scanning or rely on experience to identify harmonic sources in complex power grid topologies, resulting in low efficiency and inaccurate results. Furthermore, existing technologies only focus on the management of local nodes, ignoring the propagation path of harmonics throughout the entire power grid, leading to a lack of systematic and targeted management measures. At the same time, traditional fixed-parameter controllers cannot adjust control strategies in real time when dealing with complex and ever-changing power grid operating conditions, resulting in poor harmonic suppression effects. Summary of the Invention
[0004] To address the aforementioned issues, this invention provides a copper smelting harmonic source localization and adaptive suppression system and method. When identifying harmonic sources in complex power grid topologies, global scanning or reliance on empirical judgment is often required, leading to low efficiency and inaccurate results. Furthermore, existing technologies only focus on the management of local nodes, neglecting the propagation path of harmonics throughout the entire power grid, resulting in a lack of systematic and targeted management measures. At the same time, traditional fixed-parameter controllers cannot adjust control strategies in real time when dealing with complex and ever-changing power grid operating conditions, resulting in poor harmonic suppression effects. To achieve the above technology, the specific steps are as follows: S1. Deploy high-precision voltage sensors and high-precision current sensors at key nodes of the power distribution system in the copper smelter, and synchronously collect instantaneous voltage and current signals through the IEEE 1588 protocol at a preset sampling frequency. In this invention, the key nodes of the power distribution system of the copper smelter include: the main substation, the workshop incoming line cabinet, and the nonlinear load equipment terminal; In this embodiment, the high-precision voltage sensor used has an accuracy of 0.5%-1%, and the high-precision current sensor has an accuracy of 0.5%-1%. Instantaneous voltage and current signals are collected at a sampling frequency of 100kHz, and key node time synchronization is achieved through the IEEE 1588 protocol. S2. Using the acquired instantaneous voltage signal as input, wavelet packet transform and time-frequency analysis are first performed to construct a complex spatial disturbance feature vector. Then, a spatial weighted fusion algorithm is used to calculate the comprehensive harmonic intensity distribution vector. The specific steps are as follows: S2.1 Input instantaneous voltage signal, first perform wavelet packet transform and time-frequency analysis, extract local disturbance features, and calculate the voltage change rate; The wavelet packet transform and time-frequency analysis method of this invention is as follows: The acquired instantaneous voltage signal is input into the wavelet packet transform module, the signal is decomposed into frequency bands, local disturbance characteristics are extracted, the phase angle is extracted through Hilbert transform, and the voltage change rate is calculated. The expression is: In the formula, Indicates the sampling time interval. This represents the voltage signal at the first instant of sampling. This represents the second instantaneous voltage signal sampled, and and These are two consecutive instantaneous voltage signals; S2.2 Input: Extract local disturbance characteristics and voltage change rate, and obtain the comprehensive harmonic intensity distribution vector by constructing a complex spatial disturbance feature vector; Based on the local disturbance characteristics and voltage change rate, a complex form spatial disturbance feature vector is constructed. The expression for the spatial disturbance feature vector is as follows: in, Indicates the first The rate of change of node voltage over time Indicates the first Phase angle of the voltage signal at each node. Indicates the rate of change of voltage. t Indicates time, Represents the imaginary unit. Indicates key nodes; All key nodes in complex form spatial perturbation eigenvectors According to its weighting coefficient By performing weighted summation, we obtain the comprehensive harmonic intensity distribution vector, which reflects the harmonic energy distribution at key nodes in the power grid. ; Calculate the weighting coefficients The expression is: in, Indicates the first The electrical distance from each node to the system reference node An index representing the total number of nodes in the power grid. Indicates the first Apparent power connected to each node Indicates the total number of nodes in the power grid; The spatial disturbance eigenvectors are weighted and summed according to weighting coefficients to obtain the comprehensive harmonic intensity distribution vector, expressed as: in, The comprehensive harmonic intensity distribution vector reflects the harmonic energy distribution at each node in the power grid; S3. After decomposing the comprehensive harmonic intensity distribution vector into real and imaginary information, construct a four-dimensional feature vector, input it into a support vector machine (SVM) for training, obtain the set of main harmonic source nodes, and calculate the harmonic source contribution. S3.1 Input the comprehensive harmonic intensity distribution vector, decompose it into real part information and imaginary part information, and construct a four-dimensional feature vector; In this invention, the comprehensive harmonic intensity distribution vector is used as input to extract the real and imaginary parts of each node under typical harmonic orders in copper smelting, constructing a four-dimensional feature vector containing process characteristics. The expression of the four-dimensional feature vector is as follows: In the formula, This represents the real part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. The table represents the imaginary part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. This represents the real part of the composite harmonic intensity distribution vector at the 7th harmonic frequency. The imaginary part of the composite harmonic intensity distribution vector at the 7th harmonic frequency is represented by the vector. In this invention, we consider that in the harmonic series, the higher the harmonic order, the smaller its amplitude. Therefore, the 5th and 7th harmonics are among the harmonics with the largest amplitude and the most serious harmonics in low-voltage and medium-voltage power distribution systems. They contribute the most to the pollution of the power grid and are the primary targets of harmonic control. Therefore, this invention sets up harmonic leveling for the 5th and 7th harmonics. S3.2 Based on the constructed four-dimensional feature vector and the historical operating condition dataset of the copper smelter, the data is input into the support vector machine (SVM) for training to obtain the classification results. After obtaining the set of main harmonic source nodes and their corresponding position numbers, the harmonic source contribution is calculated. This invention utilizes a historical operating condition dataset of a copper smelter and four-dimensional feature vectors to train a Support Vector Machine (SVM) model. The training data is categorized and labeled according to three stages: smelting, oxidation, and reduction. A process-adaptive radial basis function kernel is employed, expressed as follows: in, Represents a four-dimensional eigenvector. Sample dataset of historical operating conditions of copper smelters. The preset distribution characteristic range value, Based on the harmonic distribution characteristics of copper smelting, a value of 0.5~1.2 is selected. The four-dimensional feature vectors of all nodes in the current power grid are input into a support vector machine (SVM) model for training, and the classification results are obtained. The SVM classification function expression is as follows: in, For the smelting stage correction factor (S=1 represents the smelting period, S=2 represents the oxidation period of the smelting period and S=3 represents the reduction period), For SVM bias terms, This is the coefficient corresponding to the correction factor for the smelting stage. w These are the weights of the SVM; This invention is achieved through The value is used to determine whether a node belongs to an abnormal disturbance zone. It is determined to be an abnormal disturbance area, when Nodes identified as normal are marked as harmonic source nodes and assigned corresponding location numbers, thus obtaining the set of main harmonic source nodes and their corresponding location numbers. The set of major harmonic source nodes will be obtained. A process weighting coefficient will be introduced to calculate the contribution of each harmonic source to the overall power grid harmonic pollution. The expression for calculating the harmonic source contribution is as follows: in, This represents the total number of nodes identified as harmonic sources. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first 'Contribution of each harmonic source' Indicates the first The importance weight of each harmonic source equipment process. Indicates the index of the harmonic source node. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first The importance weight of each harmonic source equipment process; the calculation process must meet the following requirements. ; The contribution level of this invention is used for subsequent governance priority ranking, including: prioritizing the deployment of filtering equipment on high contribution nodes, dynamically adjusting the APF / SVC control strategy, and prioritizing the response to disturbance signals from high contribution nodes; S4. Based on the collected instantaneous voltage signal, instantaneous current signal, and harmonic source set, after constructing a dynamic equivalent circuit model and fitting and verifying the model parameters, a directed graph of the power grid is constructed and edge weights are defined to obtain the harmonic injection function model. The specific steps are as follows: S4.1 Based on the acquired instantaneous current signal, a time-varying nonlinear admittance model is used to establish a dynamic effect circuit model for the nonlinear load equipment electric arc furnace and frequency converter. This invention models the nonlinear loads unique to copper smelters, such as electric arc furnaces and rectifier units, and establishes dynamic equivalent circuit models with process characteristics based on their actual operating characteristics, including a three-stage model of the electric arc furnace and a 12-pulse model of the rectifier unit. A time-varying nonlinear admittance model is adopted, whose parameters are dynamically adjusted with the smelting stage. The expression of the time-varying nonlinear admittance model is as follows: in, for , for , This is the recovery period; S4.2 Input the collected instantaneous current signal and the set of harmonic sources into the established dynamic equivalent circuit model, calculate the harmonic current output by the model and obtain the harmonic injection function model, then construct a directed graph and define the edge weights. In the bridge rectifier circuit model, a 12-pulse characteristic harmonic component is explicitly added, and the output current of the dynamic equivalent circuit model is calculated. The calculation expression is: In the formula, This represents the base current waveform of a 6-pulse rectifier. Indicates the harmonic order. Indicates the fundamental angular frequency of the power grid. Indicates time, Represents positive integers. Indicates the first The initial phase angle of the subharmonic current. Indicates the first The amplitude of the second harmonic current; Compare the output current of the dynamic equivalent circuit model with the actual measured current. For comparison, a process-weighted error function is used for parameter fitting, and the error function is defined as follows: in, This indicates the actual measured current. Indicates the total number of sampling points. Indicates the index of the sampling point; The fitted equivalent circuit model is converted into a frequency domain expression, and the harmonic current components are extracted using Fourier transform to form a harmonic injection function model, as shown in the following expression: in, Indicates the device at frequency Harmonic currents injected into the system at the location Represents frequency variables. This indicates the voltage amplitude at that frequency. Represents the set of model parameters. It is a nonlinear function derived from the equivalent circuit. Indicates the harmonic order. Indicates the multiplication symbol; Convert the entire power distribution network structure into a directed graph. ,in, This represents a set of nodes, indicating the electrical connection points in the power grid, such as busbars, transformers, and load connection points. This represents a set of branches, indicating the electrical connection between two nodes; For each branch road Define edge weights The edge weight is used to measure the ability of a branch to transmit harmonic current at a specific frequency. The edge weight expression is defined as follows: in, This represents the effective value of a certain harmonic current flowing through the branch. This represents the effective value of the harmonic voltage at both ends of the branch, which can be extracted from the original voltage and current signals by discrete Fourier transform. The above modeling and graph theory analysis outputs the harmonic injection function model of each nonlinear load device, the directed graph representation of the distribution network structure, and the harmonic propagation capability weight of each branch. S5. Based on the set of main harmonic source nodes and their corresponding location numbers, directed graph, edge weights and harmonic source contributions, the Dijkstra algorithm for shortest path search is used to identify the main harmonic propagation paths, determine key governance areas and filter deployment recommendations. Based on the harmonic source identification results of the SVM model, this invention obtains a set of node location numbers that are identified as the main harmonic sources; Harmonic source nodes are used as the starting point set for subsequent path searches in the power grid diagram. Find the main propagation path from each harmonic source node to each load node; Starting from a point, Dijkstra's shortest path search algorithm is used to calculate the path to all load nodes. Given the sum of path weights, find the set of paths with the smallest weights. The expression is: in, Indicates from node To the node The set of all possible paths, It is one of the paths. Representing a path The set of branches included; Based on the above process, the main propagation paths between each major harmonic source and the critical load are identified; The total weight of each propagation path is compared with the harmonic source contribution of its starting node. Multiply to calculate the path importance index: in, Indicates the first 'Contribution of each harmonic source' The larger the value, the greater the harmonic propagation influence of the path in the entire system; Path importance index based on set thresholds Sort and filter out the top... These paths are considered as the set of critical propagation paths; (in this embodiment, the threshold is set to all paths). (twice the average) This invention performs statistical analysis on branches in all critical propagation path sets, counts the frequency of each branch in different paths, and forms a branch usage frequency distribution table; for branches with usage frequencies higher than a preset threshold, they are marked as critical governance areas. Branches are key channels for harmonic energy transmission and should be prioritized for the deployment of filtering devices. Based on information about key governance areas, filter deployment recommendations are generated, including: The deployment locations of the nodes at both ends of the branch road, marked as key governance areas; The filtering frequency range is determined based on the main harmonic order propagating on the branch. The capacity configuration is set according to the average harmonic current amplitude of the branch and the filtering target. We recommend using either an active power filter or a passive filter, and dynamically adjust the settings according to the filtering requirements. It should be noted that by combining the shortest path search algorithm with the harmonic source contribution index to identify the main propagation path, the full path tracking function from the source to the load end is realized, which makes up for the shortcomings of traditional methods that only focus on local node governance and ignore the overall propagation characteristics. By ranking and statistically analyzing the path importance index, key governance areas are accurately located, which significantly improves the targeting and economy of filter deployment. S6. Based on instantaneous current signals, instantaneous voltage signals, dynamic equivalent circuit models, and filter deployment suggestions, a control strategy is constructed with total harmonic distortion (THD), reactive power error, voltage fluctuation amplitude, and smelting stage as state variables and APF / SVC control parameters as action variables. The reward function and loss function are set for the intelligent controller trained by the PPO algorithm to dynamically adjust the APF / SVC output parameters, thus completing the construction of the system and method. The APF is an active power filter, and the SVC is a static var compensator. Total harmonic distortion, reactive power error, voltage fluctuation amplitude, and smelting stage identifiers are collected from the power distribution system of the copper smelter as a set of state variables. The total harmonic distortion rate (THD) is calculated using discrete Fourier transform to determine the total harmonic content of the current voltage signal. The reactive power error The calculation expression is: in, This represents the reactive power value measured at the current moment. This indicates the set target reactive power reference value; Define the motion space based on the actual output capability of the adjustable device. This includes the APF injection current magnitude, APF cutoff frequency, SVC conduction angle, and controller gain; Construct a reward function to measure the performance of the control policy in the current state. The reward function expression is as follows: in, and These are preset weighting coefficients, representing the priority of THD, reactive power error, and control input, respectively. Represents action vector Control input variables in; The PPO algorithm is used to train the intelligent controller, and the power grid operation state under different working conditions is simulated in the simulation environment to generate a large number of state-action-reward samples. By calculating the KL divergence constraint term between the old and new strategies, the expression for the loss function of the PPO algorithm is: in, It is the probability ratio of the new strategy to the old strategy. It is the dominant function. The shearing range is used to minimize the mean square error loss between the predicted value and the actual return; the above process is repeated until the strategy converges to obtain the optimal control strategy model. Beneficial effects of this invention: This invention effectively captures the local disturbance characteristics of voltage signals through wavelet packet transform and time-frequency analysis, extracting key voltage change rate and phase angle information. This information is used to construct a complex spatial disturbance feature vector, and then a comprehensive harmonic intensity distribution vector is calculated through a spatial weighted fusion algorithm. This not only improves the accuracy of harmonic detection but also provides high-quality data support for subsequent clustering analysis, helping to more accurately identify the location of harmonic sources and their contribution. By abstracting the power grid structure into a directed graph and defining edge weights, the harmonic propagation capability of each branch can be intuitively displayed. The shortest path search algorithm is used to identify the main harmonic propagation paths, clarifying the key treatment areas. The filter deployment suggestions generated based on the information not only consider the importance of the harmonic propagation paths but also combine the contribution of the harmonic sources, making the treatment scheme more scientific and reasonable. Attached Figure Description
[0005] Figure 1 This is a flowchart of the copper smelting harmonic source localization and adaptive suppression method of the present invention.
[0006] Figure 2 This is a schematic diagram of the copper smelting harmonic source localization and adaptive suppression system of the present invention. Detailed Implementation
[0007] The present invention will be further described in detail below with reference to specific embodiments; like Figure 1 and Figure 2 As shown, this embodiment provides a copper smelting harmonic source localization and adaptive suppression system and method, including the following steps: S1. Deploy high-precision voltage sensors and high-precision current sensors at key nodes of the power distribution system in the copper smelter, and synchronously collect instantaneous voltage and current signals through the IEEE 1588 protocol at a preset sampling frequency. In this invention, the key nodes of the power distribution system of the copper smelter include: the main substation, the workshop incoming line cabinet, and the nonlinear load equipment terminal; In this embodiment, the high-precision voltage sensor used has an accuracy of 0.5%-1%, and the high-precision current sensor has an accuracy of 0.5%-1%. Instantaneous voltage and current signals are collected at a sampling frequency of 100kHz, and key node time synchronization is achieved through the IEEE 1588 protocol. It should be noted that by deploying high-precision sensors at key nodes and employing high-speed sampling and time synchronization technology, the signal distortion problem caused by insufficient sampling rate and timing asynchrony in traditional harmonic detection systems has been solved, thereby improving the accuracy and reliability of subsequent analysis results and providing a high-quality data foundation for achieving high-precision harmonic source identification and propagation path modeling.
[0008] S2. Using the acquired instantaneous voltage signal as input, wavelet packet transform and time-frequency analysis are first performed to construct a complex spatial disturbance feature vector. Then, a spatial weighted fusion algorithm is used to calculate the comprehensive harmonic intensity distribution vector. The specific steps are as follows: S2.1 Input instantaneous voltage signal, first perform wavelet packet transform and time-frequency analysis, extract local disturbance features, and calculate the voltage change rate; The wavelet packet transform and time-frequency analysis method of this invention is as follows: The acquired instantaneous voltage signal is input into the wavelet packet transform module, the signal is decomposed into frequency bands, local disturbance characteristics are extracted, the phase angle is extracted through Hilbert transform, and the voltage change rate is calculated. The expression is: In the formula, Indicates the sampling time interval. This represents the voltage signal at the first instant of sampling. This represents the second instantaneous voltage signal sampled, and and These are two consecutive instantaneous voltage signals; S2.2 Input: Extract local disturbance characteristics and voltage change rate, and obtain the comprehensive harmonic intensity distribution vector by constructing a complex spatial disturbance feature vector; Based on the local disturbance characteristics and voltage change rate, a complex form spatial disturbance feature vector is constructed. The expression for the spatial disturbance feature vector is as follows: in, Indicates the first The rate of change of node voltage over time Indicates the first Phase angle of the voltage signal at each node. Indicates the rate of change of voltage. t Indicates time, Represents the imaginary unit. Indicates key nodes; All key nodes in complex form spatial perturbation eigenvectors According to its weighting coefficient By performing weighted summation, we obtain the comprehensive harmonic intensity distribution vector, which reflects the harmonic energy distribution at key nodes in the power grid. ; Calculate the weighting coefficients The expression is: in, Indicates the first The electrical distance from each node to the system reference node An index representing the total number of nodes in the power grid. Indicates the first Apparent power connected to each node Indicates the total number of nodes in the power grid; The spatial disturbance eigenvectors are weighted and summed according to weighting coefficients to obtain the comprehensive harmonic intensity distribution vector, expressed as: in, The comprehensive harmonic intensity distribution vector reflects the harmonic energy distribution at each node in the power grid; It should be noted that the wavelet packet transform and time-frequency analysis method can effectively capture the local disturbance characteristics of voltage signals, overcoming the problem of poor adaptability of traditional Fourier transform to non-stationary signals. At the same time, a complex spatial disturbance feature vector is constructed, and a weighted fusion algorithm based on electrical distance and apparent power is introduced to further improve the ability to characterize the harmonic energy distribution of each node in the power grid, providing more representative input features for subsequent harmonic source identification.
[0009] S3. After decomposing the comprehensive harmonic intensity distribution vector into real and imaginary information, construct a four-dimensional feature vector, input it into a support vector machine (SVM) for training, obtain the set of main harmonic source nodes, and calculate the harmonic source contribution. S3.1 Input the comprehensive harmonic intensity distribution vector, decompose it into real part information and imaginary part information, and construct a four-dimensional feature vector; In this invention, the comprehensive harmonic intensity distribution vector is used as input to extract the real and imaginary parts of each node under typical harmonic orders in copper smelting, constructing a four-dimensional feature vector containing process characteristics. The expression of the four-dimensional feature vector is as follows: In the formula, This represents the real part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. The table represents the imaginary part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. This represents the real part of the composite harmonic intensity distribution vector at the 7th harmonic frequency. The imaginary part of the composite harmonic intensity distribution vector at the 7th harmonic frequency is represented by the vector. In this invention, we consider that in the harmonic series, the higher the harmonic order, the smaller its amplitude. Therefore, the 5th and 7th harmonics are among the harmonics with the largest amplitude and the most serious harmonics in low-voltage and medium-voltage power distribution systems. They contribute the most to the pollution of the power grid and are the primary targets of harmonic control. Therefore, this invention sets up harmonic leveling for the 5th and 7th harmonics. S3.2 Based on the constructed four-dimensional feature vector and the historical operating condition dataset of the copper smelter, the data is input into the support vector machine (SVM) for training to obtain the classification results. After obtaining the set of main harmonic source nodes and their corresponding position numbers, the harmonic source contribution is calculated. This invention utilizes a historical operating condition dataset of a copper smelter and four-dimensional feature vectors to train a Support Vector Machine (SVM) model. The training data is categorized and labeled according to three stages: smelting, oxidation, and reduction. A process-adaptive radial basis function kernel is employed, expressed as follows: in, Represents a four-dimensional eigenvector. Sample dataset of historical operating conditions of copper smelters. The preset distribution characteristic range value, Based on the harmonic distribution characteristics of copper smelting, a value of 0.5~1.2 is selected. The four-dimensional feature vectors of all nodes in the current power grid are input into a support vector machine (SVM) model for training, and the classification result is obtained. The classification function expression is: in, For the smelting stage correction factor (S=1 represents the smelting period, S=2 represents the oxidation period of the smelting period and S=3 represents the reduction period), For SVM bias terms, This is the coefficient corresponding to the correction factor for the smelting stage. w These are the weights of the SVM; This invention is achieved through The value is used to determine whether a node belongs to an abnormal disturbance zone. It is determined to be an abnormal disturbance area, when Nodes identified as normal are marked as harmonic source nodes and assigned corresponding location numbers, thus obtaining the set of main harmonic source nodes and their corresponding location numbers. The set of major harmonic source nodes will be obtained. A process weighting coefficient will be introduced to calculate the contribution of each harmonic source to the overall power grid harmonic pollution. The expression for calculating the harmonic source contribution is as follows: in, This represents the total number of nodes identified as harmonic sources. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first 'Contribution of each harmonic source' Indicates the first The importance weight of each harmonic source equipment process. Indicates the index of the harmonic source node. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first The importance weight of each harmonic source equipment process; the calculation process must meet the following requirements. ; The contribution level of this invention is used for subsequent governance priority ranking, including: prioritizing the deployment of filtering equipment on high contribution nodes, dynamically adjusting the APF / SVC control strategy, and prioritizing the response to disturbance signals from high contribution nodes; It should be noted that by inputting the comprehensive harmonic intensity distribution vector into the Support Vector Machine (SVM) model for clustering and classification analysis, constructing a four-dimensional feature vector based on copper smelting process characteristics, and introducing smelting stage correction factors and process weight coefficients, the accuracy and adaptability of harmonic source identification can be significantly improved. Compared with traditional fixed threshold judgment or single-band analysis, this method has stronger operating condition adaptability and classification robustness. By training the SVM model with historical operating condition data and using a process-adaptive radial basis function kernel, the model can effectively distinguish the harmonic characteristics of different smelting stages such as smelting, oxidation, and reduction, thereby more accurately identifying abnormal disturbance areas. Furthermore, by combining the power grid topology to mark potential harmonic source candidate points, not only is the positioning accuracy improved, but a clear node basis is also provided for subsequent mitigation measures. Introducing process weight coefficients to calculate the contribution index of each harmonic source makes the priority ranking of mitigation more scientific and reasonable, enabling efficient coordination between the deployment of filtering equipment and the adjustment of dynamic control strategies, and significantly improving the power grid power quality management level and mitigation efficiency.
[0010] S4. Based on the collected instantaneous voltage signal, instantaneous current signal, and harmonic source set, after constructing a dynamic equivalent circuit model and fitting and verifying the model parameters, a directed graph of the power grid is constructed and edge weights are defined to obtain the harmonic injection function model. The specific steps are as follows: S4.1 Based on the acquired instantaneous current signal, a time-varying nonlinear admittance model is used to establish a dynamic effect circuit model for the nonlinear load equipment electric arc furnace and frequency converter. This invention models the nonlinear loads unique to copper smelters, such as electric arc furnaces and rectifier units, and establishes dynamic equivalent circuit models with process characteristics based on their actual operating characteristics, including a three-stage model of the electric arc furnace and a 12-pulse model of the rectifier unit. A time-varying nonlinear admittance model is adopted, whose parameters are dynamically adjusted with the smelting stage. The expression of the time-varying nonlinear admittance model is as follows: in, for , for , This is the recovery period; S4.2 Input the collected instantaneous current signal and the set of harmonic sources into the established dynamic equivalent circuit model, calculate the harmonic current output by the model and obtain the harmonic injection function model, then construct a directed graph and define the edge weights. In the bridge rectifier circuit model, a 12-pulse characteristic harmonic component is explicitly added, and the output current of the dynamic equivalent circuit model is calculated. The calculation expression is: In the formula, This represents the base current waveform of a 6-pulse rectifier. Indicates the harmonic order. Indicates the fundamental angular frequency of the power grid. Indicates time, Represents positive integers. Indicates the first The initial phase angle of the subharmonic current. Indicates the first The amplitude of the second harmonic current; Compare the output current of the dynamic equivalent circuit model with the actual measured current. For comparison, a process-weighted error function is used for parameter fitting, and the error function is defined as follows: in, This indicates the actual measured current. Indicates the total number of sampling points. Indicates the index of the sampling point; The fitted equivalent circuit model is converted into a frequency domain expression, and the harmonic current components are extracted using Fourier transform to form a harmonic injection function model, as shown in the following expression: in, Indicates the device at frequency Harmonic currents injected into the system at the location Represents frequency variables. This indicates the voltage amplitude at that frequency. Represents the set of model parameters. It is a nonlinear function derived from the equivalent circuit. Indicates the harmonic order. Indicates the multiplication symbol; Convert the entire power distribution network structure into a directed graph. ,in, This represents a set of nodes, indicating the electrical connection points in the power grid, such as busbars, transformers, and load connection points. This represents a set of branches, indicating the electrical connection between two nodes; For each branch road Define edge weights The edge weight is used to measure the ability of a branch to transmit harmonic current at a specific frequency. The edge weight expression is defined as follows: in, This represents the effective value of a certain harmonic current flowing through the branch. This represents the effective value of the harmonic voltage at both ends of the branch, which can be extracted from the original voltage and current signals by discrete Fourier transform. The above modeling and graph theory analysis outputs the harmonic injection function model of each nonlinear load device, the directed graph representation of the distribution network structure, and the harmonic propagation capability weight of each branch. It should be noted that by constructing a dynamic equivalent circuit model of nonlinear load equipment (such as electric arc furnaces and rectifier units) and fitting parameters with measured data, the harmonic injection characteristics of the equipment at different smelting stages can be accurately reflected. Compared with traditional static models, this method considers the time-varying and nonlinear nature of the equipment's operating state, significantly improving the simulation accuracy and engineering applicability of the model. By converting the power grid structure into a directed graph and defining the branch harmonic propagation capability weights, visualization and quantitative analysis of the harmonic propagation path are achieved, which helps to identify key nodes and weak links in harmonic propagation. In addition, the use of a process-weighted error function for parameter fitting can highlight the impact of key smelting stages on the model's accuracy, thereby improving the model's adaptability to actual operating conditions. The output harmonic injection function model and directed graph structure provide a solid theoretical foundation and data support for subsequent harmonic source tracing, propagation path optimization, and the layout of control equipment, significantly enhancing the systematicness and pertinence of power grid harmonic pollution control.
[0011] S5. Based on the set of main harmonic source nodes and their corresponding location numbers, directed graph, edge weights and harmonic source contributions, the Dijkstra algorithm for shortest path search is used to identify the main harmonic propagation paths, determine key governance areas and filter deployment recommendations. Based on the harmonic source identification results of the SVM model, this invention obtains a set of node location numbers that are identified as the main harmonic sources; Harmonic source nodes are used as the starting point set for subsequent path searches in the power grid diagram. Find the main propagation path from each harmonic source node to each load node; Starting from a point, Dijkstra's shortest path search algorithm is used to calculate the path to all load nodes. Given the sum of path weights, find the set of paths with the smallest weights. The expression is: in, Indicates from node To the node The set of all possible paths, It is one of the paths. Representing a path The set of branches included; Based on the above process, the main propagation paths between each major harmonic source and the critical load are identified; The total weight of each propagation path is compared with the harmonic source contribution of its starting node. Multiply to calculate the path importance index: in, Indicates the first 'Contribution of each harmonic source' The larger the value, the greater the harmonic propagation influence of the path in the entire system; Path importance index based on set thresholds Sort and filter out the top... These paths are considered as the set of critical propagation paths; (in this embodiment, the threshold is set to all paths). (twice the average) This invention performs statistical analysis on branches in all critical propagation path sets, counts the frequency of each branch in different paths, and forms a branch usage frequency distribution table; for branches with usage frequencies higher than a preset threshold, they are marked as critical governance areas. Branches are key channels for harmonic energy transmission and should be prioritized for the deployment of filtering devices. Based on information about key governance areas, filter deployment recommendations are generated, including: The deployment locations of the nodes at both ends of the branch road, marked as key governance areas; The filtering frequency range is determined based on the main harmonic order propagating on the branch. The capacity configuration is set according to the average harmonic current amplitude of the branch and the filtering target. We recommend using either an active power filter or a passive filter, and dynamically adjust the settings according to the filtering requirements. It should be noted that by combining the shortest path search algorithm with the harmonic source contribution index to identify the main propagation path, the full path tracking function from the source to the load end is realized, which makes up for the shortcomings of traditional methods that only focus on local node governance and ignore the overall propagation characteristics. By ranking and statistically analyzing the path importance index, key governance areas are accurately located, which significantly improves the targeting and economy of filter deployment. S6. Based on instantaneous current signals, instantaneous voltage signals, dynamic equivalent circuit models, and filter deployment suggestions, a control strategy is constructed with total harmonic distortion (THD), reactive power error, voltage fluctuation amplitude, and smelting stage as state variables and APF / SVC control parameters as action variables. The reward function and loss function are set for the intelligent controller trained by the PPO algorithm to dynamically adjust the APF / SVC output parameters, thus completing the construction of the system and method. The APF is an active power filter, and the SVC is a static var compensator. Total harmonic distortion, reactive power error, voltage fluctuation amplitude, and smelting stage identifiers are collected from the power distribution system of the copper smelter as a set of state variables. The total harmonic distortion rate (THD) is calculated using discrete Fourier transform to determine the total harmonic content of the current voltage signal. The reactive power error The calculation expression is: in, This represents the reactive power value measured at the current moment. This indicates the set target reactive power reference value; Define the motion space based on the actual output capability of the adjustable device. This includes the APF injection current magnitude, APF cutoff frequency, SVC conduction angle, and controller gain; Construct a reward function to measure the performance of the control policy in the current state. The reward function expression is as follows: in, and These are preset weighting coefficients, representing the priority of THD, reactive power error, and control input, respectively. Represents action vector Control input variables in; The PPO algorithm is used to train the intelligent controller, and the power grid operation state under different working conditions is simulated in the simulation environment to generate a large number of state-action-reward samples. By calculating the KL divergence constraint term between the old and new strategies, the expression for the loss function of the PPO algorithm is: in, It is the probability ratio of the new strategy to the old strategy. It is the dominant function. It is the shearing range, minimizing the mean squared error loss between the predicted value and the actual return; Repeat the above process until the policy converges to obtain the optimal control policy model; The trained intelligent controller is deployed to edge computing nodes and connected to sensors and execution terminals; The controller receives state vectors from the sensors and generates control actions through forward propagation. It should be noted that the intelligent control strategy based on deep reinforcement learning, which couples multiple state variables and action variables, breaks through the adaptation bottleneck of traditional fixed parameter controllers in complex dynamic environments. The intelligent controller trained using the PPO algorithm has online adaptive adjustment capability and can optimize the APF / SVC output parameters in real time according to the current smelting stage and power grid operation status, thereby achieving efficient closed-loop suppression of harmonic pollution and improving the system's intelligence level and response speed.
[0012] This embodiment also provides a copper smelting harmonic source localization and adaptive suppression system, including: The system includes a signal acquisition module, a disturbance modeling module, a source identification module, a path analysis module, a filtering strategy module, and an intelligent control module. The signal acquisition module is used to deploy high-precision voltage and current sensors in the main substation, workshop incoming line cabinet and nonlinear load equipment of the copper smelter's power distribution system to acquire instantaneous voltage and instantaneous current signals. The disturbance modeling module is used to perform wavelet packet transform and time-frequency analysis on the acquired voltage signal, extract the voltage change rate and phase angle, construct a complex spatial disturbance feature vector, and use a spatial weighted fusion algorithm to calculate the comprehensive harmonic intensity distribution vector. The source identification module is used to perform support vector machine clustering analysis on the comprehensive harmonic intensity distribution vector, identify the location number of the main harmonic sources in the power grid topology and their corresponding harmonic contribution, and output the harmonic source set and contribution ranking results. The path analysis module is used to abstract the power grid structure into a directed graph, define the edge weight as the ratio of a certain harmonic current to voltage, and combine the harmonic source location information with the shortest path search algorithm to identify the main propagation path and determine the key treatment area. The filtering strategy module is used to generate filter deployment suggestions based on information about key governance areas, including deployment location, filtering frequency range, capacity configuration and equipment type selection, and to prioritize responding to disturbance signals from high-contribution nodes; The intelligent control module is used to construct a deep reinforcement learning control strategy with total harmonic distortion rate, reactive power error, voltage fluctuation amplitude and smelting stage as state variables and APF / SVC control parameters as action variables. The intelligent controller is trained using the PPO algorithm, and the APF / SVC output parameters are dynamically adjusted based on the real-time status to generate control commands.
[0013] This embodiment also provides a computer device applicable to the copper smelting harmonic source localization and adaptive suppression method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the copper smelting harmonic source localization and adaptive suppression method proposed in the above embodiment. The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0014] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for locating and adaptively suppressing harmonic sources in copper smelting as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0015] In summary, this invention, through wavelet packet transform and time-frequency analysis, can effectively capture the local disturbance characteristics of voltage signals, extract key voltage change rate and phase angle information, and use this information to construct a complex spatial disturbance feature vector. Then, a spatial weighted fusion algorithm is used to calculate the comprehensive harmonic intensity distribution vector, which not only improves the accuracy of harmonic detection but also provides high-quality data support for subsequent clustering analysis, helping to more accurately identify the location of harmonic sources and their contribution. By abstracting the power grid structure into a directed graph and defining edge weights, the harmonic propagation capability of each branch can be intuitively displayed. The shortest path search algorithm is used to identify the main harmonic propagation paths, clarifying key treatment areas. The filter deployment suggestions generated based on the information not only consider the importance of harmonic propagation paths but also combine the contribution of harmonic sources, making the treatment scheme more scientific and reasonable.
[0016] 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 it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A system and method for locating and adaptively suppressing harmonic sources in copper smelting, characterized in that, Includes the following steps: S1. Deploy high-precision voltage sensors and high-precision current sensors at key nodes of the power distribution system in the copper smelter, and synchronously collect instantaneous voltage and current signals through the IEEE 1588 protocol at a preset sampling frequency. S2. Using the acquired instantaneous voltage signal as input, wavelet packet transform and time-frequency analysis are first performed to construct a complex spatial disturbance feature vector. Then, a spatial weighted fusion algorithm is used to calculate the comprehensive harmonic intensity distribution vector. S3. After decomposing the comprehensive harmonic intensity distribution vector into real and imaginary information, construct a four-dimensional feature vector, input it into a support vector machine (SVM) for training, obtain the set of main harmonic source nodes, and calculate the harmonic source contribution. S4. Based on the collected instantaneous voltage signal, instantaneous current signal, and harmonic source set, after constructing a dynamic equivalent circuit model and fitting and verifying the model parameters, a directed graph of the power grid is constructed and edge weights are defined to obtain the harmonic injection function model. S5. Based on the set of main harmonic source nodes and their corresponding location numbers, directed graph, edge weights and harmonic source contributions, the Dijkstra algorithm for shortest path search is used to identify the main harmonic propagation paths, determine key governance areas and filter deployment recommendations. S6. Based on instantaneous current signals, instantaneous voltage signals, dynamic equivalent circuit models, and filter deployment suggestions, a control strategy is constructed with total harmonic distortion (THD), reactive power error, voltage fluctuation amplitude, and smelting stage as state variables and APF / SVC control parameters as action variables. The reward function and loss function are set for the intelligent controller trained by the PPO algorithm to dynamically adjust the APF / SVC output parameters, thus completing the construction of the system and method.
2. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, The process involves inputting the acquired instantaneous voltage signal, first performing wavelet packet transform and time-frequency analysis to construct a complex spatial disturbance feature vector, and then using a spatial weighted fusion algorithm to calculate the comprehensive harmonic intensity distribution vector. The specific steps include the following: S2.1 Input instantaneous voltage signal, first perform wavelet packet transform and time-frequency analysis, extract local disturbance features, and calculate the voltage change rate; The expression for the rate of change of voltage is: In the formula, Indicates the sampling time interval. This represents the voltage signal at the first instant of sampling. This represents the second instantaneous voltage signal sampled; S2.2 Input: Extract local disturbance characteristics and voltage change rate, and obtain the comprehensive harmonic intensity distribution vector by constructing a complex spatial disturbance feature vector; The expression for the spatial perturbation eigenvector is: in, Indicates the first The rate of change of node voltage over time Indicates the first Phase angle of the voltage signal at each node. Indicates the rate of change of voltage. t Indicates time, Represents the imaginary unit. Indicates key nodes; Calculate the weighting coefficients The expression is: in, Indicates the first The electrical distance from each node to the system reference node An index representing the total number of nodes in the power grid. Indicates the first Apparent power connected to each node Indicates the total number of nodes in the power grid; The expression for the combined harmonic intensity distribution vector is as follows: 。 3. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, The process of decomposing the comprehensive harmonic intensity distribution vector into real and imaginary parts to construct a four-dimensional feature vector, inputting it into a support vector machine (SVM) for training, obtaining the set of major harmonic source nodes, and then calculating the harmonic source contribution is as follows: S3.1 Input the comprehensive harmonic intensity distribution vector, decompose it into real part information and imaginary part information, and construct a four-dimensional feature vector; The expression for the four-dimensional feature vector is: In the formula, This represents the real part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. The table represents the imaginary part of the composite harmonic intensity distribution vector at the 5th harmonic frequency. This represents the real part of the composite harmonic intensity distribution vector at the 7th harmonic frequency. The imaginary part of the composite harmonic intensity distribution vector at the 7th harmonic frequency is represented by the vector. S3.2 Based on the constructed four-dimensional feature vector and the historical operating condition dataset of the copper smelter, the data is input into the support vector machine (SVM) for training to obtain the classification results. After obtaining the set of main harmonic source nodes and their corresponding position numbers, the harmonic source contribution is calculated. The kernel expression for the process-adaptive radial basis function is as follows: in, Represents a four-dimensional eigenvector. Sample dataset of historical operating conditions of copper smelters. The preset distribution characteristic range value, Based on the harmonic distribution characteristics of copper smelting, a value of 0.5~1.2 is selected. The four-dimensional feature vectors of all nodes in the current power grid are input into a support vector machine (SVM) model for training, and the classification results are obtained. The SVM classification function expression is as follows: in, This is a correction factor for the smelting stage. For SVM bias terms, This is the coefficient corresponding to the correction factor for the smelting stage. w These are the weights of the SVM.
4. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 3, characterized in that, The expression for the harmonic source contribution obtained from the calculation is as follows: in, This represents the total number of nodes identified as harmonic sources. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first 'Contribution of each harmonic source' Indicates the first The importance weight of each harmonic source equipment process. Indicates the index of the harmonic source node. Indicates the first The absolute value of the combined disturbance intensity of each harmonic source, Indicates the first The importance weight of each harmonic source equipment process; the calculation process must meet the following requirements. .
5. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, The harmonic injection function model is obtained by constructing a dynamic equivalent circuit model, fitting and verifying the model parameters, building a directed graph of the power grid and defining edge weights, based on the collected instantaneous voltage signal, instantaneous current signal, and harmonic source set. The specific steps include: S4.1 Based on the acquired instantaneous current signal, a time-varying nonlinear admittance model is used to establish a dynamic effect circuit model for the nonlinear load equipment electric arc furnace and frequency converter. S4.2 Input the collected instantaneous current signal and the set of harmonic sources into the established dynamic equivalent circuit model, calculate the harmonic current output by the model and obtain the harmonic injection function model, then construct a directed graph and define the edge weights. The edge weight expression is defined as follows: in, This represents the effective value of a certain harmonic current flowing through the branch. This represents the effective value of the harmonic voltage at both ends of the branch.
6. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 5, characterized in that, Based on the acquired instantaneous current signal, a time-varying nonlinear admittance model is used to establish a dynamic effect circuit model for the nonlinear load equipment, electric arc furnace and frequency converter. The expression of the time-varying nonlinear admittance model is as follows: in, for , for , This is the recovery period.
7. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, The method described above, which uses the set of main harmonic source nodes and their corresponding location numbers, directed graph, edge weights, and harmonic source contributions, employs the Dijkstra algorithm for shortest path search to identify main harmonic propagation paths, determine key governance areas, and provide filter deployment recommendations. The expression for the Dijkstra algorithm is as follows: in, Indicates from node To the node The set of all possible paths, It is one of the paths. Representing a path The set of branches included.
8. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, In step S6, the reward function calculation expression is as follows: in, and For preset weighting coefficients, The control input variables that represent the action vector, This represents the reactive power value measured at the current moment. This represents the total harmonic distortion rate.
9. The copper smelting harmonic source localization and adaptive suppression system and method according to claim 1, characterized in that, In step S6, the loss function calculation expression is as follows: in, This represents the ratio of the probability of the new strategy to the probability of the old strategy. Represents the dominance function. Indicates the cutting range. E t This indicates that the average value of empirical data is calculated.