Control method and system of reactive power compensation cabinet capacitor switch protection device

Abstract

The invention discloses a control method and system for a capacitor switch protection device of a reactive power compensation cabinet, and relates to the field of switch protection device control, and the method comprises the steps: obtaining voltage data of two ends of an electric reactor in the reactive power compensation cabinet, predicting the load condition of the reactive power compensation cabinet according to the voltage data, the demand degree of capacitor switching is evaluated based on the prediction result; analyzing the starting time and sequence of the capacitor switch protection device corresponding to each reactor according to the demand degree and the voltage data, and adjusting the protection state of the capacitor switch protection device in combination with a local intelligent control algorithm; and obtaining the overvoltage multiple of the reactive compensation cabinet based on the protection state, evaluating the limit effectiveness of the control of the capacitor switch protection device, and adjusting the installation state of the capacitor switch protection device according to the limit effectiveness. According to the method, the overvoltage multiple of the reactive compensation cabinet is obtained through reverse modeling of the protection state and serves as a core index for evaluating the control effectiveness, and the control process has the self-verification capacity.

Description

Technical Field

[0001] This invention relates to the field of switch protection device control, and in particular to a control method and system for a capacitor switch protection device in a reactive power compensation cabinet. Background Technology

[0002] A reactive power compensation cabinet is a device used in power systems. Its main purpose is to provide reactive power compensation to improve the power factor of the power grid, reduce power loss, and improve the stability and efficiency of the power grid. The capacitor switch protection device of the reactive power compensation cabinet is used to protect and control the capacitors and other key electrical components in the reactive power compensation cabinet. It mainly uses automation to protect the capacitors from switching, status monitoring, overvoltage, and overcurrent, ensuring the normal and stable operation of the reactive power compensation function in the power system.

[0003] However, existing control methods for capacitor switch protection devices in reactive power compensation cabinets often rely on fixed switching voltage thresholds or schedules, lacking the ability to model complex load response paths. This leads to drawbacks such as delayed response or over-switching, especially in intermittent loads or frequent fluctuation scenarios.

[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a control method and system for a capacitor switch protection device in a reactive power compensation cabinet, thereby enabling the control process to possess self-verification capabilities.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a control method for a capacitor switch protection device in a reactive power compensation cabinet, the control method comprising: Obtain voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and assess the demand for capacitor switching based on the prediction results and quantification techniques. Based on demand and voltage data analysis, the activation time and sequence of the capacitor switch protection device corresponding to each reactor are determined, and the protection status of the capacitor switch protection device is adjusted in conjunction with the local intelligent control algorithm. Based on the protection status, the overvoltage multiple of the reactive power compensation cabinet is obtained, the effectiveness of the limiting control of the capacitor switch protection device is evaluated, and the installation status of the capacitor switch protection device is adjusted according to the limiting effectiveness.

[0007] Preferably, the process involves acquiring voltage data across the reactors in the reactive power compensation cabinet, predicting the load condition of the cabinet based on the voltage data, and assessing the demand for capacitor switching based on the prediction results and quantitative techniques. The voltage data across the reactor during the operation of the reactive power compensation cabinet is collected by sensor equipment and used as a voltage data sequence. The voltage data sequence is then converted into an image format using the Gram matrix angle field. The voltage data sequence in image format is processed by a convolutional neural network to output long-time and short-time features, and combined with dual-channel fusion technology to predict the load status of the reactive power compensation cabinet. Based on the analysis of the load status of the reactive power compensation cabinet, the dynamic characteristics of the voltage data sequence on the time axis are analyzed, and the reactive power change pattern is identified by combining the weighted dynamic time programming analysis of the load status. Based on the identification results and fuzzy clustering technology, demand response characteristic parameters are constructed, and a quantitative assessment model of capacitor switching demand is established to evaluate the demand for capacitor switching in the face of reactive power changes.

[0008] Preferably, converting voltage data sequences into image formats using Gram matrix angular field includes: The voltage data sequence is normalized to a preset interval, and the normalized voltage data sequence is converted from Cartesian coordinates to polar coordinates. A multi-scale sliding window framework is defined based on the coordinate information. The voltage data sequence is divided into segments of target length at each scale using a multi-scale sliding window framework, which serve as basis vectors for the local vector subspace. The basis vectors at the corresponding scales are then multiplied together. Construct a local Gram matrix based on the multiplication result, convert the local Gram matrix into an angular matrix using the inverse cosine function, and then weight the angular matrix by introducing volatility weights to output a weighted angular matrix. By superimposing weighted angle matrices corresponding to each scale to form a multi-channel tensor, the coupling strength of the angular structure and fluctuation amplitude between local time blocks is analyzed to determine the pixel position of the multi-channel tensor and complete the image format conversion of the voltage data sequence.

[0009] Preferably, the process of obtaining the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluating the limiting effectiveness of the capacitor switch protection device control, and adjusting the installation status of the capacitor switch protection device according to the limiting effectiveness includes: By using a sampler and an electromagnetic interference disturbance identifier deployed at the capacitor switch protection device, the voltage response curves generated on the bus before and after the switching operation of the capacitor switch protection device are monitored and identified. Based on the voltage response curve, an overvoltage multiple current spectrum is constructed. Combined with the control strategy effectiveness evaluation network, the switching action of the capacitor switch protection device is analyzed to obtain the coupling degree of the switching action. The effectiveness assessment results of the control of the capacitor switch protection device are obtained based on the degree of coupling. When the effectiveness assessment results are less than the target threshold, the installation status of the capacitor switch protection device is adjusted.

[0010] Secondly, the present invention also provides a control system for a capacitor switch protection device for a reactive power compensation cabinet, the system comprising: The capacitor switching demand assessment module is used to obtain voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and assess the demand for capacitor switching based on the prediction results and quantification technology. The switch protection device status adjustment module is used to analyze the opening time and sequence of the capacitor switch protection device corresponding to each reactor based on demand and voltage data, and adjust the protection status of the capacitor switch protection device in combination with local intelligent control algorithm. The switch protection device performance evaluation module is used to obtain the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluate the effectiveness of the limiting control of the capacitor switch protection device, and adjust the installation status of the capacitor switch protection device according to the limiting effectiveness.

[0011] The beneficial effects of this invention are as follows: This invention uses voltage data from both ends of the reactor to predict load conditions and obtain switching demand, enabling the control strategy to have forward-looking and trend-judgment capabilities. Combined with the analysis of start-up time and sequence, it introduces a local intelligent control algorithm, allowing each protection device to have autonomous evaluation and collaborative judgment capabilities without increasing the global computational burden. At the same time, it obtains the overvoltage multiple of the reactive power compensation cabinet through reverse modeling of the protection state and uses it as the core indicator for evaluating the effectiveness of the control, giving the entire control process the ability to self-verify and self-correct. This breaks through the limitation of traditional control devices lacking a feedback mechanism for aftereffects, thereby significantly improving the service life and stability margin of the protection device equipment. Attached Figure Description

[0012] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0013] Figure 1 This is a flowchart of a control method for a capacitor switch protection device for a reactive power compensation cabinet according to an embodiment of the present invention; Figure 2 This is a schematic block diagram of the control system of a capacitor switch protection device for a reactive power compensation cabinet according to an embodiment of the present invention.

[0014] In the picture: 1. Capacitor switching demand assessment module; 2. Switch protection device status adjustment module; 3. Switch protection device performance assessment module. Detailed Implementation

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0017] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0019] Please see Figure 1 This invention provides a control method for a capacitor switch protection device in a reactive power compensation cabinet, the control method comprising: Step S1: Obtain the voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and evaluate the demand for capacitor switching based on the prediction results and quantification techniques.

[0020] In one embodiment, during the process of acquiring voltage data across the reactors in the reactive power compensation cabinet, predicting the load status of the cabinet based on the voltage data, and assessing the demand for capacitor switching based on the prediction results and quantification techniques, the voltage data across the reactors during the operation of the cabinet can be collected using sensor devices as a voltage data sequence. The voltage data sequence is then converted into an image format using a Gram matrix angle field. A convolutional neural network is used to process the image-formatted voltage data sequence, outputting long-time and short-time features. Combined with dual-channel fusion technology, the load status of the reactive power compensation cabinet is predicted. The dynamic characteristics of the voltage data sequence on the time axis are analyzed based on the load status of the cabinet, and the reactive power change pattern is identified by combining weighted dynamic time programming analysis. Based on the identification results and fuzzy clustering techniques, demand response feature parameters are constructed, and a quantitative assessment model for capacitor switching demand in the face of reactive power changes is established to evaluate the demand for capacitor switching.

[0021] The process of converting voltage data sequences into image format using Gram matrix angular fields includes: normalizing the voltage data sequence to a preset interval and converting the normalized voltage data sequence from Cartesian coordinates to polar coordinates; defining a multi-scale sliding window framework based on the coordinate information; dividing the voltage data sequence into segments of target length at each scale using the multi-scale sliding window framework, using these segments as basis vectors for the local vector subspace, and multiplying the basis vectors within the corresponding scales; constructing a local Gram matrix based on the multiplication result; converting the local Gram matrix into an angular matrix using the inverse cosine function; weighting the angular matrix by introducing volatility weights; and outputting a weighted angular matrix; superimposing the weighted angular matrices corresponding to each scale to form a multi-channel tensor; analyzing the coupling strength of the angular structure and volatility amplitude between local time blocks; determining the pixel positions of the multi-channel tensor; and completing the image format conversion of the voltage data sequence.

[0022] It needs to be explained that, in the process of assessing the demand for capacitor switching, voltage data across the reactors in the reactive power compensation cabinet is collected in real time using sensor equipment, forming a raw voltage data sequence that represents the electrical state fluctuations of the power grid in the microscopic time domain. To improve the expressiveness of the data in the deep model, this sequence is normalized to a specific interval, specifically 0-1 or -1 to 1, thus avoiding feature bias caused by different magnitudes. It is also mapped from the Cartesian coordinate system to the polar coordinate system, giving it derivable directional and periodic information, laying the foundation for subsequent angle field construction. At the same time, a multi-scale sliding window framework is constructed to slice the normalized data at different time scales. Each segment constitutes a local vector subspace, serving as a segment representation of the voltage waveform at the current scale. Based on these local vectors, a Gram matrix is ​​constructed, and the inner product matrix between each pair of local subvectors is calculated. Then, the arccosine function is introduced to transform the inner product relationship into angle information, thereby forming a local angle matrix, realizing the encoding from numerical similarity to geometric directional structure.

[0023] Meanwhile, to enhance the physical meaning of the angular field image, the angular matrix is ​​combined with voltage fluctuation rate for weighting. The standard deviation or short-term power change amplitude of each local window is calculated as the fluctuation rate weight. The influence of each angle value is adjusted through multiplication operation to generate a weighted angular matrix at multiple scales. A multi-channel image tensor is constructed by stacking tensors, with each channel representing a time scale. Each pixel encodes the relative angle and fluctuation coupling strength between the corresponding time segments, so that the final image retains the temporal correlation, scale hierarchy and perturbation pattern differences of the original sequence, and thus can understand the deep features of time series changes from a visual structure perspective.

[0024] A multi-channel graph is input into a convolutional neural network (CNN) to extract local structural features by leveraging its ability to extract spatial patterns. Long-term (e.g., trend changes) and short-term (e.g., spike disturbances) features are extracted through multi-layer convolution, pooling, and activation operations. Simultaneously, a dual-channel fusion structure is used to fuse the two types of features at the feature level, enhancing the model's ability to discriminate states under complex load change backgrounds. The output feature vector is further fed into dynamic time warping. Weighted dynamic time warping (WDTW) is used to perform nonlinear matching of voltage responses in different time periods, aligning the temporal evolution paths under different load conditions, thereby achieving cross-cycle state analogy and recognition. The recognition results are transformed into fuzzy sets through fuzzy clustering technology, forming multiple load state subspaces. Each subspace constructs demand response feature parameters, including reactive power demand level, voltage fluctuation tolerance, and disturbance response inertia. Finally, a demand measurement model for capacitor switching is established, which outputs a continuous demand index to guide subsequent capacitor switching actions.

[0025] Based on the above implementation process, a complete closed loop is achieved, from raw voltage sampling, time-series imaging, visual deep feature extraction, dynamic pattern alignment, fuzzy demand quantification, and control intent output. This not only improves response resolution and load discrimination capability, but also provides a solid data and model foundation for the intelligent scheduling of subsequent capacitor switch protection strategies.

[0026] Specifically, in constructing demand response feature parameters based on identification results and fuzzy clustering technology, and establishing a quantitative assessment model for capacitor switching demand under reactive power variation, the process involves evaluating the demand for capacitor switching. This is achieved by establishing a membership matrix based on the number of clusters in the reactive power variation pattern, determining the distance between the reactive power variation pattern and the cluster center, and defining the distance result as the loss function of the fuzzy clustering layer. Based on the loss function and the identification results of the reactive power variation pattern, demand feature classification variables are analyzed. After processing these variables using one-hot encoding, a reverse correction is implemented. Demand response feature parameters for reactive power variation patterns are constructed based on the reverse correction results, and a mapping relationship between these parameters and capacitor switching labels is established using a collaborative regression algorithm. The mapping relationship is then used to divide the demand response feature parameters into labeled and unlabeled sample sets, constructing a quantitative assessment model for capacitor switching demand under reactive power variation conditions. This quantitatively assesses the switching potential of voltage data sequences and evaluates the demand for capacitor switching.

[0027] It needs to be explained that in the entire chain modeling process from reactive power change pattern recognition to quantitative assessment of capacitor switching demand, an innovative mechanism integrating data-driven, fuzzy logic, and semi-supervised learning is used. This transforms the originally static "to switch or not to switch" problem into a refined high-resolution control problem of "when to switch, how much to switch, and based on what to switch." In the clustering layer, the system performs fuzzy clustering on the identified reactive power change patterns using a preset or adaptively determined number of clusters (e.g., k=5 or k=7). The Euclidean distance between the power pattern and the cluster center in each time period is used as the basis for its membership degree, and a membership matrix U=[uij] is constructed, where uij represents the membership degree of the i-th sample to the j-th cluster. Furthermore, minimizing the membership matrix multiplied by the sum of squared distances is used as the loss function of the fuzzy clustering layer. This allows for the clustering centers to fit the sample distribution to a certain degree, serving as an important input for subsequent reverse correction of demand characteristics. The clustering results are then structurally encoded, treating the clustering outcome of each reactive power change pattern as a categorical variable. This variable is converted into a binary vector (e.g., [0, 0, 1, 0, 0]) via one-hot encoding. Combined with the clustering residuals and historical switching feedback, reverse correction is performed. Therefore, the reverse correction step is no longer a simple error feedback, but rather a dynamic remapping of the label space for each type of demand characteristic, combining historical switching success rates, switch temperature rise, and malfunction records. This results in a corrected multidimensional demand state characteristic R={r1, r2, ..., rn}, which integrates indicators such as pattern morphology, fluctuation frequency, power transition amplitude, and system response stability.

[0028] By utilizing a collaborative regression algorithm to establish a mapping relationship between demand response characteristic parameters and capacitor switching labels, a highly robust prediction function f(R)=y is formed through multi-model collaborative training. Here, y represents a specific capacitor switching scheme, such as switching group number, switching capacity, or switching delay. Finally, a demand quantification assessment model is constructed to quickly determine the "potential for switching" corresponding to the voltage data sequence when future reactive power changes occur, i.e., whether there is a clear necessity for switching. Its numerical output is a demand index D∈[0,1] in a continuous domain, which is used as the basis for subsequent capacitor switching action planning. This significantly improves the expression granularity and prediction reliability, and realizes intelligent identification of switching demand based on multi-dimensional coupling of time, structure, intensity, and behavior feedback. This greatly improves the system's response speed, switching accuracy, and overvoltage control capability, providing a powerful and flexible control foundation for the entire reactive power compensation system.

[0029] Step S2: Analyze the activation time and sequence of the capacitor switch protection device corresponding to each reactor based on the demand and voltage data, and adjust the protection status of the capacitor switch protection device in conjunction with the local intelligent control algorithm.

[0030] In one embodiment, during the process of analyzing the activation time and sequence of the capacitor switch protection devices corresponding to each reactor based on demand and voltage data, and adjusting the protection state of the capacitor switch protection devices using a local intelligent control algorithm, a mapping function between load state and switching priority can be established based on voltage data and capacitor switching demand. This mapping function is then converted into a multi-objective switching sequence for the capacitor switch protection devices based on the reactor topology. A game theory-based weighting method is used to correct the influence of subjective factors on the multi-objective switching sequence, and a multi-criteria compromise solution ranking algorithm is used to quantify and rank the switching sequence and the influence of bus fluctuation gain. A ranking diagram is dynamically generated based on the quantified ranking, and the activation time and sequence of the capacitor switch protection devices corresponding to each reactor are determined by combining the state lifetime factor and response delay of the capacitor switch protection devices. Based on the activation time and sequence, the control logic is modified using a local intelligent control algorithm to determine the number of actions of the capacitor switch protection devices, generate control commands, and adjust the protection state of the capacitor switch protection devices.

[0031] Specifically, the method utilizes game theory-based combinatorial weighting to correct the influence of subjective factors on multi-objective switching sequences, and combines a multi-criteria compromise solution ranking algorithm to quantify and rank the impact of switching sequences and bus fluctuation gain. This includes: constructing a decision matrix based on the switching demand indicators of the capacitor switch protection device and the multi-objective switching sequences; calculating the characteristic weight of each multi-objective switching sequence relative to the corresponding indicator under any switching demand indicator; analyzing the information entropy of any switching demand indicator based on the characteristic weight, and determining the weight vector based on the coefficient of variation of any switching demand indicator obtained from the information entropy; performing arbitrary linear combinations on the weight vector; analyzing the first derivative of the decision matrix based on the combination process, determining the optimal linear coefficients to obtain the optimal combination weights, analyzing the positive and negative ideals of any switching demand indicator, and completing the correction of subjective factors; analyzing the group utility value and individual regret value of any switching demand indicator based on the optimal combination weights and positive and negative ideals, analyzing the tendency of switching demands on multi-objective switching sequences, and combining the impact of bus fluctuation gain as a boundary condition with the tendency to determine the quantified ranking of multi-objective switching sequences.

[0032] It should be explained that in the process of adjusting the protection status of the capacitor switch protection device, a "load status - switching priority" mapping function can be established through historical training data and real-time acquired data. This function not only considers the voltage fluctuation trend and capacitor switching demand, but also combines the topological position of the reactor in the compensation cabinet and the load partitioning influence factor, and introduces physical structural attributes, so that the switching priority is not only a logical order, but also has an electrical system stability orientation. After obtaining the mapping function, a preliminary multi-objective switching sequence is generated, and each candidate in the sequence represents a possible capacitor switch triggering path.

[0033] Furthermore, a game theory-based combinatorial weighting method is introduced to correct subjective preferences among multiple objectives. Specifically, a decision matrix can be constructed based on the mapping relationship between each switching demand indicator (such as voltage stability, equipment wear minimization, compensation response rate, etc.) and the switching sequence. The standardized feature weight of each column is calculated to measure the contribution of different switching schemes to a certain indicator. At the same time, the information value and distinguishing ability of each indicator are evaluated based on the weight information entropy and coefficient of variation. The initial subjective weight vector set by expert experience is linearly combined to construct a weighted vector. To ensure the optimal combination weight, the first derivative analysis of the decision matrix is ​​performed to obtain the optimal combination weight, thus avoiding human bias from dominating the decision results.

[0034] Simultaneously, based on the multi-criteria compromise solution ranking algorithm, the distance between each switching sequence and the "positive ideal solution" and "negative ideal solution" is calculated, quantifying the optimal proximity of individuals under various indicators. This is used to calculate the "group utility value" (reflecting overall performance) and the "individual regret value" (reflecting extreme gaps). Furthermore, to further align with actual power grid stability requirements, the influence of bus fluctuation gain (such as the amplification factor of short-term voltage fluctuations on system power response sensitivity) is combined with the ranking tendency as a boundary constraint, ultimately determining the comprehensive ranking value of the switching sequences. This results in the construction of a ranking graph where each node represents a capacitor bank protection device, and the edge weights reflect its response priority in the current sequence. After introducing state lifetime factors (such as the ratio of action count to rated life) and response delay parameters, the ranking graph is further adjusted to make the action arrangement more closely match the actual capability and current health status of the switches, thereby determining the optimal activation time and sequence strategy.

[0035] The process involves modifying the control logic based on the opening time and sequence using a local intelligent control algorithm to determine the number of actions of the capacitor switch protection device, generate control commands, and adjust the protection state of the capacitor switch protection device. This includes: building a mathematical model of the capacitor switch protection device; collecting the operating parameters of the protection device in real time; using a convolutional network to spatiotemporally encode the operating parameters to generate an electrical situation projection with predictive capabilities; treating the capacitor switch protection device as an intelligent agent, and generating a dynamic control demand field by incorporating the time distribution of switch actions, action frequency statistics, and aging coefficient, combined with the electrical situation projection mapping; embedding the dynamic control demand field, opening time, and sequence into the control logic of the local intelligent control algorithm, integrating the optimal switching path for the target time period, and generating a command package containing predictions of the number of actions; encoding the command package into logic code to obtain the control commands for the capacitor switch protection device, and using the control commands to adjust the protection state of the capacitor switch protection device.

[0036] It needs to be explained that in the process of adjusting the protection state of the capacitor switch protection device, a mathematical model of the capacitor switch protection device is constructed to simulate its dynamic behavior and response characteristics. The model is based on the electrical characteristics of the device (such as parameters such as voltage, current, inductance, and capacitance) and changes in the external power grid environment. The state equation of the device is established through appropriate physical laws and electrical principles (such as Ohm's law and Kirchhoff's laws). Specifically, differential equations can be used to describe the voltage changes and current flow of the capacitor. Furthermore, the state transition and electrical response of the capacitor switch are represented by a state-space model. The operating parameters of the protection device (such as the number of switching operations, current and voltage changes, temperature, load conditions, etc.) are collected in real time and input into the model. The collected operating parameters are spatiotemporally encoded using a convolutional neural network. In this process, the convolutional neural network acts as a feature extractor, which can effectively extract spatial features (such as voltage change patterns at different time points) and temporal features (such as the trends of voltage and current changes over time) from high-dimensional electrical parameter data. This generates an electrical situation projection with predictive capabilities. This projection not only reveals the real-time status of electrical equipment, but also infers the behavior trend of capacitor switches in the near future, such as current fluctuations, the possibility of overvoltage, and the instantaneous response of the power grid, based on the spatiotemporal characteristics of historical data.

[0037] Simultaneously, the capacitor switch protection device is treated as an intelligent agent. Combining the time distribution of switch actions, action frequency statistics, and aging coefficient, these factors are further integrated into the electrical situation projection to form a dynamic control demand field. This dynamic control demand field is a multi-dimensional dynamic environment mapping that considers the switch's usage history, frequency, and aging effects. It helps assess the response capability required for each capacitor switch in future switching decisions. For example, a higher switch action frequency and a larger aging coefficient may weaken its response capability, leading to a corresponding decrease in the predicted value of future switching requirements. This ensures that the system does not cause equipment damage or premature failure due to frequent actions. The dynamic control demand field generates a dynamic environment mapping under different conditions. The system generates a map of switching demand at different times under different conditions, enabling dynamic and precise decision-making. The dynamic control demand field, activation time, and sequence information are further embedded into the local intelligent control algorithm. The local intelligent control algorithm adjusts the control logic according to the optimal switching path for the target time period, generating an instruction package containing predictions of the number of actions. For example, a capacitor switch may need to wait for a period of time before switching to avoid the impact of overvoltage, or the system may select to switch the capacitor bank with the heavier load first according to the load situation of the time period to maintain the stability of the power grid. By predicting the number of actions of each switch, the system can determine the frequency of use of the switch in advance and allocate the best action time for each switch within the predetermined time window.

[0038] The generated instruction package is ultimately encoded into logic code to adjust the working state of the capacitor switch protection device. Through the encoded control instructions, the switching action and number of actions of each protection device are precisely controlled, avoiding electrical oscillations caused by frequent start-stop, while minimizing the risks caused by equipment aging. This enables timely capacitor switching when the load fluctuates drastically, and reduces unnecessary operations when the load is stable, thereby optimizing the overall performance of the system. It not only realizes the intelligent transformation from electrical status to switching decision, but also achieves refined and personalized control of the capacitor switch protection device by introducing local intelligent control algorithms.

[0039] Step S3: Obtain the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluate the effectiveness of the limiting control of the capacitor switch protection device, and adjust the installation status of the capacitor switch protection device according to the limiting effectiveness.

[0040] In one embodiment, during the process of obtaining the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluating the limiting effectiveness of the capacitor switch protection device control, and adjusting the installation status of the capacitor switch protection device according to the limiting effectiveness, a sampler and an electromagnetic interference disturbance identifier deployed at the capacitor switch protection device can be used to monitor and identify the voltage response curves generated on the bus before and after the switching operation of the capacitor switch protection device; an overvoltage multiple flow graph is constructed based on the voltage response curve, and combined with the control strategy effectiveness evaluation network, the switching action of the capacitor switch protection device is analyzed to obtain the coupling degree of the switching action; the limiting effectiveness evaluation result of the capacitor switch protection device control is obtained based on the coupling degree, and the installation status of the capacitor switch protection device is adjusted when the limiting effectiveness evaluation result is less than the target threshold.

[0041] Specifically, based on the voltage response curve, an overvoltage multiple current spectrum is constructed. Combined with a control strategy effectiveness evaluation network, the switching action of the capacitor switch protection device is analyzed to obtain the coupling degree of the switching action. This includes: combining high-frequency Fourier transform with an adaptive wavelet network to extract the ratio of the overvoltage peak value to the steady-state reference voltage from the voltage response curve, constructing an overvoltage multiple current spectrum to reflect the dynamic excitation capability of the capacitor switch protection device during the switching process; using the overvoltage multiple current spectrum as a constraint function input to the control strategy effectiveness evaluation network to compress and encode the overvoltage accumulation situation generated by the switching action of the capacitor switch protection device; mapping the encoding result to the strategy response space, defining the action of the capacitor switch protection device as an excitation function using the strategy response space, and obtaining the disturbance weight of the stability of the capacitor switch protection device; comparing the disturbance weight with the actual voltage response deviation, generating a constraint residual, and constructing a graph neural structure through backpropagation to analyze the resonant coupling degree of the corresponding switching action of the capacitor switch protection device.

[0042] It should be explained that, during the effectiveness assessment of the limitation, a high-precision voltage sampler and an electromagnetic interference disturbance identifier deployed at the capacitor switch protection device are used to continuously monitor the bus voltage waveform before and after the switching operation. This captures the voltage response disturbances caused by the switching action, including short-time overshoot, resonance excitation, and transient waveform jitter, thereby obtaining a high-time-domain resolution voltage response curve. Frequency domain analysis is then performed using the response curve, and key voltage features are extracted by combining high-frequency Fourier transform and adaptive wavelet network, especially the ratio between the overvoltage peak and the steady-state reference voltage, i.e., the overvoltage multiple. An overvoltage multiple flow graph is constructed, which can then show the changing trend of the voltage response over multiple time periods, as well as the dynamic relationship between the capacitor switching moment and the grid excitation behavior, quantifying the transient excitation intensity caused by the switch switching.

[0043] After constructing the graph, it is injected as an input function into the control strategy effectiveness evaluation network (usually a deep neural network or graph neural network structure). This compresses redundant oscillation information in the voltage response curve, extracts representative features, and encodes them as a strategy response vector for capacitor switching actions. This vector is then mapped to a strategy response space, where each capacitor switching action is represented as an excitation function. The output of this function is used to measure the potential impact of the action on system stability disturbances. Furthermore, the disturbance weights generated by this excitation function are compared with the actual observed voltage response deviations (e.g., steady-state settling time, fluctuation range) to obtain the constraint residual, which is the numerical difference between the actual response and the ideal strategy response.

[0044] Based on the constrained residuals, a graph neural network structure is constructed using backpropagation. Nodes in the graph represent different capacitor bank protection devices, and edges represent the electrical coupling relationships between these devices. Through graph neural network modeling, the path information that triggers resonance or superposition effects in the switching behavior of different devices is learned. The system analyzes whether resonance coupling effects exist in the switching actions. If a protection device frequently triggers high-multiple overvoltages, generates nonlinear cumulative disturbances, or forms resonance peaks with other devices after switching, its "switching coupling degree" index is calculated, reflecting the device's interference potential in the network. When the evaluated constraint effectiveness index (such as device interference coefficient or control margin) is lower than the system-set threshold, the device's control effect is deemed insufficient. This triggers an adjustment mechanism for the installation status, which may include: changing the matching relationship between the capacitor bank and the switching device, adjusting the device's connection location (such as busbar replacement), replacing physical hardware, or changing the connection branch topology.

[0045] Therefore, it not only realizes a closed-loop control mechanism for capacitor switch protection devices from "real-time response - disturbance measurement - strategy evaluation - structural optimization", but also dynamically captures transient characteristics caused by switching behavior, effectively ensuring that the reactive power compensation system still has high stability and dispatchability under high frequency and strong load fluctuation conditions, reducing resonance risk, and improving overall robustness and fine control capability.

[0046] Please see Figure 2 The present invention also provides a control system for a capacitor switch protection device for a reactive power compensation cabinet, the system comprising: The capacitor switching demand assessment module 1 is used to acquire voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and assess the demand for capacitor switching based on the prediction results and quantification technology. The switch protection device status adjustment module 2 is used to analyze the opening time and sequence of the capacitor switch protection device corresponding to each reactor based on the demand and voltage data, and adjust the protection status of the capacitor switch protection device in combination with the local intelligent control algorithm. The switch protection device performance evaluation module 3 is used to obtain the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluate the effectiveness of the limiting control of the capacitor switch protection device, and adjust the installation status of the capacitor switch protection device according to the limiting effectiveness.

[0047] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0048] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A control method for a capacitor switch protection device in a reactive power compensation cabinet, characterized in that, The control method includes: Obtain voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and assess the demand for capacitor switching based on the prediction results and quantification techniques. Based on demand and voltage data analysis, the activation time and sequence of the capacitor switch protection device corresponding to each reactor are determined, and the protection status of the capacitor switch protection device is adjusted in conjunction with the local intelligent control algorithm. Based on the protection status, the overvoltage multiple of the reactive power compensation cabinet is obtained, the effectiveness of the limiting control of the capacitor switch protection device is evaluated, and the installation status of the capacitor switch protection device is adjusted according to the limiting effectiveness.

2. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 1, characterized in that, The process of acquiring voltage data across the reactors in the reactive power compensation cabinet, predicting the load condition of the reactive power compensation cabinet based on the voltage data, and assessing the demand for capacitor switching based on the prediction results and quantification techniques includes: The voltage data across the reactor during the operation of the reactive power compensation cabinet is collected by sensor equipment and used as a voltage data sequence. The voltage data sequence is then converted into an image format using the Gram matrix angle field. The voltage data sequence in image format is processed by a convolutional neural network to output long-time and short-time features, and combined with dual-channel fusion technology to predict the load status of the reactive power compensation cabinet. Based on the analysis of the load status of the reactive power compensation cabinet, the dynamic characteristics of the voltage data sequence on the time axis are analyzed, and the reactive power change pattern is identified by combining the weighted dynamic time programming analysis of the load status. Based on the identification results and fuzzy clustering technology, demand response characteristic parameters are constructed, and a quantitative assessment model of capacitor switching demand is established to evaluate the demand for capacitor switching in the face of reactive power changes.

3. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 2, characterized in that, The method of converting voltage data sequences into image format using Gram matrix angular field includes: The voltage data sequence is normalized to a preset interval, and the normalized voltage data sequence is converted from Cartesian coordinates to polar coordinates. A multi-scale sliding window framework is defined based on the coordinate information. The voltage data sequence is divided into segments of target length at each scale using a multi-scale sliding window framework, which serve as basis vectors for the local vector subspace. The basis vectors at the corresponding scales are then multiplied together. Construct a local Gram matrix based on the multiplication result, convert the local Gram matrix into an angular matrix using the inverse cosine function, and then weight the angular matrix by introducing volatility weights to output a weighted angular matrix. By superimposing weighted angle matrices corresponding to each scale to form a multi-channel tensor, the coupling strength of the angular structure and fluctuation amplitude between local time blocks is analyzed to determine the pixel position of the multi-channel tensor and complete the image format conversion of the voltage data sequence.

4. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 3, characterized in that, The demand response feature parameters are constructed based on the identification results and fuzzy clustering technology. A quantitative assessment model for capacitor switching demand is established to evaluate the demand for capacitor switching in the face of reactive power changes. The assessment of the demand for capacitor switching includes: Based on the number of clusters of reactive power change patterns, a membership matrix is ​​established to determine the distance between reactive power change patterns and cluster centers, and the distance result is defined as the loss function of the fuzzy clustering layer. Based on the identification results of loss function and reactive power change mode, the demand characteristic classification variables are analyzed, and after processing the demand characteristic classification variables using one-hot coding, a reverse correction is implemented. Based on the reverse correction results, demand response characteristic parameters of reactive power change mode are constructed, and the mapping relationship between demand response characteristic parameters and capacitor input labels is established through collaborative regression algorithm; By mapping the characteristic parameters of demand response into sets of labeled and unlabeled samples, a quantitative assessment model for capacitor input demand under reactive power variation is constructed to quantitatively assess the potential for switching of voltage data sequences and evaluate the demand for capacitor switching.

5. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 1, characterized in that, The process of analyzing the activation time and sequence of the capacitor switch protection devices corresponding to each reactor based on demand and voltage data, and adjusting the protection status of the capacitor switch protection devices in conjunction with a local intelligent control algorithm, includes: A mapping function between load state and switching priority is established based on voltage data and the demand for capacitor switching. The mapping function is then converted into a multi-objective switching sequence for capacitor switch protection devices by combining the reactor topology location. The influence of subjective factors on multi-objective switching sequences is corrected by using game theory combinatorial weighting method, and the influence of switching sequences and bus fluctuation gain is quantified and ranked by multi-criteria compromise solution ranking algorithm. The sorting diagram is dynamically generated based on the quantitative sorting, and the opening time and sequence of the capacitor switch protection device corresponding to each reactor are determined by combining the state lifetime factor and response delay of the capacitor switch protection device. Based on the start-up time and sequence, the control logic is modified using a local intelligent control algorithm to determine the number of times the capacitor switch protection device will operate, generate control commands, and adjust the protection status of the capacitor switch protection device.

6. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 5, characterized in that, The method of using game theory combined weighting to correct the influence of subjective factors on multi-objective switching sequences, and combining a multi-criteria compromise solution ranking algorithm to quantify and rank the influence of switching sequences and bus fluctuation gain, includes: A decision matrix is ​​constructed based on the switching demand index and multi-objective switching sequence of the capacitor switch protection device, and the characteristic weight of each multi-objective switching sequence to the corresponding index is calculated under any switching demand index. Based on the feature weight analysis, the information entropy of any target appeal indicator is analyzed, and the coefficient of variation of any target appeal indicator is obtained based on the information entropy to determine the weight vector. The weight vector is then arbitrarily linearly combined. Based on the analysis of the first derivative of the decision matrix in the combination process, the optimal linear coefficients are determined to obtain the optimal combination weights. The positive and negative ideals of any target demand index are analyzed to correct subjective factors. Based on the optimal combination of demands and positive and negative ideals, the group utility value and individual regret value of arbitrary switching demand indicators are analyzed. The degree of tendency of switching demands on multi-objective switching sequences is analyzed. The influence of bus fluctuation gain is used as a boundary condition and combined with the degree of tendency to determine the quantitative ranking of multi-objective switching sequences.

7. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 5, characterized in that, The process of modifying the control logic based on the opening time and sequence using a local intelligent control algorithm, determining the number of times the capacitor switch protection device will operate, generating control commands, and adjusting the protection state of the capacitor switch protection device includes: A mathematical model of a capacitor switch protection device is built, the operating parameters of the protection device are collected in real time, and the operating parameters are spatiotemporally encoded using a convolutional network to generate an electrical situation projection with predictive capabilities. By treating the capacitor switch protection device as an intelligent agent, and by introducing the time distribution of switch action, action frequency statistics and aging coefficient, dynamic control demand field is generated by combining electrical situation projection mapping. The dynamic control demand field, activation time and sequence are embedded into the local intelligent control algorithm control logic, and the optimal switching path of the target time period is integrated to generate an instruction package containing the prediction of the number of actions. The instruction packet is encoded into logic code to obtain the control instructions for the capacitor switch protection device, and the protection status of the capacitor switch protection device is adjusted using the control instructions.

8. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 1, characterized in that, The process of obtaining the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluating the effectiveness of the limiting control of the capacitor switch protection device, and adjusting the installation status of the capacitor switch protection device according to the limiting effectiveness includes: By using a sampler and an electromagnetic interference disturbance identifier deployed at the capacitor switch protection device, the voltage response curves generated on the bus before and after the switching operation of the capacitor switch protection device are monitored and identified. Based on the voltage response curve, an overvoltage multiple current spectrum is constructed. Combined with the control strategy effectiveness evaluation network, the switching action of the capacitor switch protection device is analyzed to obtain the coupling degree of the switching action. The effectiveness assessment results of the control of the capacitor switch protection device are obtained based on the degree of coupling. When the effectiveness assessment results are less than the target threshold, the installation status of the capacitor switch protection device is adjusted.

9. The control method for a capacitor switch protection device in a reactive power compensation cabinet according to claim 8, characterized in that, The overvoltage multiple current spectrum is constructed based on the voltage response curve. Combined with the control strategy effectiveness evaluation network, the switching action of the capacitor switch protection device is analyzed to obtain the coupling degree of the switching action, including: By combining high-frequency Fourier transform with adaptive wavelet network, the ratio of overvoltage peak value to steady-state reference voltage is extracted from the voltage response curve to construct an overvoltage multiple current spectrum, which reflects the dynamic excitation capability of the capacitor switch protection device during the switching process. The overvoltage multiple current spectrum is used as a constraint function input to the control strategy effectiveness evaluation network to compress and encode the overvoltage accumulation situation generated by the switching action of the capacitor switch protection device. The encoding result is mapped to the policy response space, and the action of the capacitor switch protection device is defined as the excitation function using the policy response space to obtain the disturbance weights for the stability of the capacitor switch protection device. The perturbation weights are compared with the actual voltage response deviation to generate a restricted residual. The reverse propagation is then used to construct a neural network structure to analyze the resonant coupling degree of the corresponding switching action of the capacitor switch protection device.

10. A control system for a capacitor switch protection device in a reactive power compensation cabinet, used to implement the control method for the capacitor switch protection device in a reactive power compensation cabinet as described in any one of claims 1-9, characterized in that, The system includes: The capacitor switching demand assessment module is used to obtain voltage data across the reactor in the reactive power compensation cabinet, predict the load status of the reactive power compensation cabinet based on the voltage data, and assess the demand for capacitor switching based on the prediction results and quantification technology. The switch protection device status adjustment module is used to analyze the opening time and sequence of the capacitor switch protection device corresponding to each reactor based on demand and voltage data, and adjust the protection status of the capacitor switch protection device in combination with local intelligent control algorithm. The switch protection device performance evaluation module is used to obtain the overvoltage multiple of the reactive power compensation cabinet based on the protection status, evaluate the effectiveness of the limiting control of the capacitor switch protection device, and adjust the installation status of the capacitor switch protection device according to the limiting effectiveness.

Images