Power distribution area voltage treatment method for high-proportion new energy access

Abstract

The invention relates to the technical field of power system operation control, and discloses a power distribution area voltage treatment method for high-proportion new energy access, which comprises the following steps of: acquiring multi-source electrical quantity data of a power distribution transformer and a user side in a power distribution area; on the basis of the multi-source electrical quantity data, a typical voltage behavior mode, caused by new energy fluctuation, of the power distribution unit area is identified; generating a targeted prospective reactive power optimization control instruction based on the identified voltage behavior mode and output prediction data of the new energy unit; and sending the control instruction to a voltage regulation device in a transformer area so as to control the voltage regulation device to carry out voltage regulation and suppress an expected voltage out-of-limit risk in a future preset time period. According to the method, different execution layer intelligent mechanisms (master-slave cooperation, sensitivity feedforward and plan-feedback check) are designed for different modes, efficient and accurate execution of a governance instruction is ensured, and the characteristics of different adjustment resources are fully utilized.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a method for voltage management in distribution substations with a high proportion of new energy access. Background Technology

[0002] With the increasing penetration of distributed photovoltaic and wind power in distribution substations, the randomness, intermittency, and volatility of their output have significantly impacted the voltage stability of these substations. Traditional voltage mitigation methods mainly rely on techniques such as fixed capacitor switching and adjusting transformer taps. These methods have significant limitations: First, their response is delayed, and they are mostly reactive, unable to cope with the rapid fluctuations of new energy sources. Second, their mitigation strategies are singular and lack specificity, failing to address the diverse voltage problems caused by differences in grid structure, load characteristics, and new energy access in different substations. Third, traditional data acquisition and processing methods struggle to efficiently utilize massive amounts of multi-source time-series data, thus failing to accurately characterize the true voltage behavior of substations.

[0003] Therefore, current technologies suffer from insufficient data processing capabilities, low accuracy in locating voltage problems, poor adaptability of mitigation solutions, delayed prediction and response, and limited device functionality. There is an urgent need for an intelligent voltage mitigation method capable of accurate profiling, proactive prediction, and adaptive mitigation. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for voltage management in distribution transformer areas with a high proportion of renewable energy integration. This method achieves accurate identification of voltage behavior patterns in distribution transformer areas by constructing a unique fusion data set in the smart grid. Based on these patterns and predictive data, it generates forward-looking and differentiated management instructions, and finally executes the management through an intelligent collaborative and secure closed-loop mechanism, thereby effectively suppressing the risk of voltage exceeding limits.

[0005] This invention is achieved through the following technical solution:

[0006] A method for voltage management in distribution transformer areas with a high proportion of renewable energy integration includes the following steps:

[0007] Collect multi-source electrical quantity data from distribution transformers and user sides in the distribution area;

[0008] Based on the multi-source electrical quantity data, the typical voltage behavior patterns of the distribution substation caused by new energy fluctuations are identified.

[0009] Based on the identified voltage behavior patterns and the output prediction data of the new energy units, targeted and forward-looking reactive power optimization control commands are generated.

[0010] The reactive power optimization control command is sent to the voltage regulation device in the transformer area to control the voltage regulation device to perform voltage regulation and suppress the expected risk of voltage exceeding the limit in the future preset period.

[0011] As an optimization, the multi-source electrical quantity data is a fusion data set used to characterize the dual characteristics of new energy volatility and voltage sensitivity. The fusion data set includes first data for tracking the root causes of new energy volatility and second data for evaluating the sensitivity of transformer area voltage to power injection. The first data includes instantaneous power pairs on the DC and AC sides of the new energy power generation unit. The difference between the instantaneous power pairs is calculated to capture transient fluctuations in inverter energy conversion efficiency in real time, and the transient fluctuations are used as a forward-looking criterion for ultra-short-term power output mutations of the new energy power generation unit.

[0012] The second data includes minute voltage change events generated by daily load switching or capacitor bank operation within the transformer area. By analyzing the voltage and power changes at the time of the event, the equivalent Thevenin impedance parameter of the transformer area is identified and updated online, and the equivalent Thevenin impedance parameter is used as a key weighting factor in the control command to determine the strength of the governance strategy.

[0013] As an optimization, the specific process for identifying the typical voltage behavior pattern based on the fused data set is as follows:

[0014] A multi-dimensional voltage behavior feature vector with clear physical meaning is constructed. The voltage behavior feature vector includes time-series fluctuation features, sensitivity features, and source-load time-series correlation features. The time-series fluctuation features are based on the first data, extracting the standard deviation and mutation frequency of the transient fluctuations. The sensitivity features are based on the second data, using the equivalent Thevenin impedance parameter as the core indicator characterizing the inherent vulnerability of the transformer area voltage. The source-load time-series correlation features are calculated by measuring the dynamic time warping distance between the new energy daily power generation curve and the typical load curve, in order to quantify the severity of source-load mismatch.

[0015] The voltage behavior feature vector is input into an unsupervised clustering model to identify data clusters;

[0016] By combining knowledge of power system operation, each data cluster is assigned an engineering label with direct governance guidance significance, thereby identifying the typical voltage behavior patterns.

[0017] As an optimization, the typical voltage behavior modes include efficiency-driven sudden change type, high impedance-sensitive type and source-load timing mismatch type. The efficiency-driven sudden change type is characterized by the timing fluctuation characteristics being higher than the fluctuation threshold, indicating that the voltage fluctuation is mainly driven by the violent and rapid fluctuation of the output of the new energy unit itself.

[0018] The high impedance sensitive type is characterized by a sensitivity characteristic value higher than the sensitivity threshold, indicating that the grid structure of the transformer area is weak, and even if the power fluctuation is not large, it is easy to cause voltage over-limit.

[0019] The source-load timing mismatch is characterized by the source-load timing correlation value being higher than the correlation threshold, indicating that the voltage problem is mainly caused by the time-varying misalignment between the peak output of new energy sources and the peak load.

[0020] As an optimization, the specific process of generating targeted, forward-looking reactive power optimization control commands based on the identified voltage behavior patterns and the output prediction data of the new energy units is as follows:

[0021] The system invokes a pre-established library of differentiated governance strategies strongly correlated with different voltage behavior patterns, and uses the output prediction data of the new energy unit as the core input, wherein:

[0022] If the problem is identified as being dominated by abrupt efficiency changes, an inertial smoothing management strategy is initiated: based on the power output prediction data, the gradient of new energy power output change within the ultra-short-term time window is calculated, and the gradient of change is defined as the predicted fluctuation trend; a first reactive power optimization model is constructed with the goal of suppressing the expected voltage change rate caused by the predicted fluctuation trend; control commands are generated through the first reactive power optimization model to control the inverter to output dynamic reactive current opposite to the predicted fluctuation trend in advance.

[0023] If identified as a high impedance sensitive type, a forward-looking prevention and control strategy is initiated: Based on the power output prediction data, the direction and magnitude of the continuous change in new energy power output within the future medium-term time window are determined, and the direction and magnitude of the continuous change in new energy power output are defined as the predicted power trend; a second reactive power optimization model is constructed with the goal of eliminating the expected voltage deviation generated by the predicted power trend on the equivalent Thevenin impedance; control commands are generated through the second reactive power optimization model to control the static reactive power compensation device to output reactive power opposite to the predicted power trend in a small step manner before the voltage over-limit point, so as to maintain voltage stability;

[0024] If the source-load time-series mismatch is identified, a spatiotemporal shift mitigation strategy is initiated: Based on the power output forecast data and load forecast data, a source-load power net difference curve is plotted on a future daily time scale, and the shape of the net difference curve is defined as the predicted time-series imbalance trend; a third reactive power optimization model is constructed with the goal of offsetting the time-series imbalance trend and flattening the daily voltage curve; control commands are generated through the third reactive power optimization model to plan and control the adjustable transformer taps and capacitor reactor groups, performing voltage reduction operations before the predicted overvoltage period and voltage boosting and compensation operations before the predicted undervoltage period.

[0025] As an optimization, the specific process of sending the reactive power optimization control command to the voltage regulation device in the transformer area to control the voltage regulation device to perform voltage regulation and suppress the expected risk of voltage exceeding the limit in the future preset period is as follows:

[0026] When the target of governance is the efficiency-sudden change-dominant mode, a master-slave collaborative mechanism is adopted: the inverter with the fastest response is set as the master execution unit, responsible for quickly tracking and smoothing fluctuations; at the same time, the static reactive power compensation device is set as the slave execution unit to provide background reactive power support to compensate for the capacity limitation of the inverter. During the execution of the master-slave collaborative mechanism, the master execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its own remaining regulation capacity and governance effect, so that the output of the slave execution unit is coordinated with the action of the master execution unit to jointly complete the fluctuation smoothing task.

[0027] When the target of the governance is the high impedance sensitive mode, a voltage sensitivity feedforward mechanism is adopted: before the control command is issued, the equivalent Thevenin impedance parameter updated in real time is used as the feedforward quantity to dynamically correct the amplitude of the control command, so as to ensure that the same control command can produce the expected voltage regulation effect when the grid impedance changes, and avoid under-regulation or over-regulation.

[0028] When the target of governance is the aforementioned source-load time-series mismatch mode, a plan-feedback verification mechanism is adopted: the real-time collected source-load short-term forecast data is compared with the forecast reference curve on which the original control command was generated, and the absolute deviation or root mean square error of the source-load short-term forecast data and the forecast reference curve at key time points is calculated. The source-load short-term forecast data includes the new energy output forecast curve and the load forecast curve within a preset time period. If the deviation or error exceeds the reconstruction threshold, it is determined that the actual time-series imbalance trend has changed significantly, and the rapid recalculation of the third reactive power optimization model is triggered. Based on the new time-series imbalance trend obtained from the recalculation, a new set of control commands is generated to replan and control the adjustable transformer tap and capacitor reactor group.

[0029] As an optimization, the main execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its remaining adjustment capacity and governance effect as follows:

[0030] A1. The main execution unit monitors its output current in real time and calculates the difference between the current reactive power output and the rated capacity of the main execution unit as the real-time residual regulation capacity; at the same time, it monitors the voltage fluctuation rate of local key nodes. If the voltage fluctuation rate exceeds the smoothing target threshold, it is determined that there is a fluctuation power gap that has not been completely smoothed.

[0031] A2. Based on the fluctuating power gap and the real-time remaining adjustment capability, the initial output reference value of the slave execution unit is calculated through a first-order hysteresis filter function; the time constant of the filter function is set to be greater than the response period of the master execution unit but less than the response period of the slave execution unit to ensure the stability of the output reference command.

[0032] A3. Limit the initial output reference value, wherein the upper limit of the initial output reference value does not exceed the rated capacity of the slave execution unit, and the lower limit of the initial output reference value is not lower than the minimum reactive power allowed by the system; and issue the final output reference command after limiting to the slave execution unit.

[0033] As an optimization, the process of dynamically correcting the amplitude of the control command by using the real-time updated equivalent Thevenin impedance parameter as a feedforward is as follows:

[0034] B1. Based on the equivalent Thevenin impedance parameters, establish a linear relationship model between the change in the control command and the change in the voltage of the key node, and define the linear relationship coefficient of the linear relationship model as the real-time command-voltage sensitivity.

[0035] B2. Based on the control command, the theoretical voltage regulation amount required to eliminate the expected voltage deviation is determined;

[0036] B3. Divide the theoretical voltage regulation amount by the real-time command-voltage sensitivity to calculate the reactive power control command amplitude after feedforward correction; the calculation formula is: corrected command amplitude = theoretical voltage regulation amount / real-time command-voltage sensitivity.

[0037] B4. The control command using the modified command amplitude is sent to the static reactive power compensation device.

[0038] As an optimization, the key timing points include power extreme points or zero-crossing points.

[0039] As an optimization, while controlling the voltage regulating device to perform voltage regulation, it also includes:

[0040] C1. Within a preset time window after the control command is issued, monitor the actual voltage change of key nodes in real time and calculate the correlation coefficient between the actual voltage change curve and the expected voltage change curve; if the correlation coefficient is consistently lower than the reliability threshold, it is determined that the treatment effect has not met expectations.

[0041] C2. Establish a dual safety protection mechanism with the voltage acceptable range as the hard boundary and the voltage change rate as the soft boundary; inside the voltage acceptable range, a stricter safety action boundary is preset; once the actual voltage is detected to exceed this safety action boundary, it is determined to be approaching the hard boundary, and an emergency correction command is immediately generated and executed to restore the voltage to the safe target range within the voltage acceptable range; once the voltage change rate is detected to exceed the soft boundary, a reverse damping command is immediately superimposed on the current control command, and the magnitude of the reverse damping command is proportional to the degree of voltage change rate exceeding the limit.

[0042] C3. Whenever the dual-boundary security protection mechanism is triggered, the type of the event, the control command before the trigger, and the parameters of the protection command are recorded as a security event. Based on the statistical pattern of security events over a period of time, with the optimization goal of reducing false activation and refusal to activate the security protection mechanism, the threshold of the soft boundary, the position of the security action boundary, and the proportional coefficient of the reverse damping command are automatically adjusted to match the security protection strategy with the dynamic characteristics of the transformer area.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] This invention defines and collects a fusion dataset, focusing not only on conventional electrical quantities, but also on the fundamental aspects such as efficiency fluctuations within new energy units and grid impedance characteristics, laying a solid data foundation for accurate analysis.

[0045] This invention achieves accurate diagnosis and classification of complex voltage behavior patterns by constructing feature vectors with clear physical meaning and using a combination of unsupervised clustering and engineered labels, thus clarifying the causes of voltage problems.

[0046] This invention combines the identified patterns with power output prediction data to tailor forward-looking management strategies for different types of voltage problems, achieving a fundamental shift from passive response to proactive prevention and control.

[0047] This invention designs different execution layer intelligent mechanisms (master-slave collaboration, sensitivity feedforward, and plan-feedback verification) for different modes, ensuring the efficient and accurate execution of governance instructions and making full use of the characteristics of different regulatory resources.

[0048] This invention introduces a real-time security protection mechanism that integrates monitoring, evaluation, intervention, and self-learning, providing a reliable security safety net for the entire governance process and enabling the system to continuously optimize, thus significantly enhancing its robustness. Attached Figure Description

[0049] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:

[0050] Figure 1 This is a flowchart of a distribution transformer area voltage management method for high-proportion renewable energy access, as described in this invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0052] This embodiment 1 provides a method for voltage management in distribution transformer areas with a high proportion of new energy access, such as... Figure 1 As shown, it includes the following steps:

[0053] S1. Collect multi-source electrical quantity data of distribution transformers and user side in the distribution area.

[0054] In this step, sensors and acquisition terminals deployed on the distribution transformer and user side systematically collect multi-source electrical quantity data to form a fused data set. This set aims to characterize the dual characteristics of renewable energy volatility and voltage sensitivity, specifically including:

[0055] First, the data is used to track the root causes of fluctuations in renewable energy output. Its core is the acquisition of instantaneous power pairs on the DC and AC sides of renewable energy generation units (such as photovoltaic inverters). By calculating the instantaneous difference between DC and AC power, transient fluctuations in the inverter's energy conversion efficiency can be captured in real time. This fluctuation signal directly reflects the internal state of the equipment and can serve as a forward-looking criterion for ultra-short-term changes in renewable energy output on the order of seconds or minutes, predicting fluctuations much earlier than monitoring only the AC power side.

[0056] This involves collecting instantaneous voltage and current on both the DC and AC sides of the new energy power generation unit, and calculating the instantaneous power on the DC and AC sides respectively. By calculating the difference between the instantaneous power on the DC and AC sides, the transient power loss value, which characterizes the inverter's energy conversion process, is obtained in real time. Compared to stable AC power, fluctuations in this transient power loss value can reflect the impact of changes in sunlight, wind speed, or the equipment's own condition on output earlier and more sensitively. Therefore, it serves as a forward-looking criterion for ultra-short-term power output changes on the order of seconds or minutes in the new energy power generation unit.

[0057] The second data point is used to assess the voltage sensitivity of a distribution area to power injection. Its core is to utilize small voltage fluctuations (e.g., voltage changes within 0.5% to 5% of rated voltage) caused by routine events such as daily load switching or capacitor bank operations within the distribution area. By analyzing the voltage and power changes corresponding to these events, the equivalent Thevenin impedance parameter of the distribution area can be identified and updated online. This parameter (Thevenin impedance parameter) is a key indicator characterizing the inherent vulnerability of the distribution area's grid structure and directly determines the magnitude of voltage changes under the same power disturbance.

[0058] S2. Based on the multi-source electrical quantity data, identify the typical voltage behavior patterns of the distribution substation caused by new energy fluctuations. This step performs in-depth analysis of the fused data collected in S1 to achieve a precise profile of the voltage behavior patterns.

[0059] In some embodiments, the specific process of S2 is as follows:

[0060] S2.1 Construct a multi-dimensional voltage behavior feature vector with clear physical meaning. Extract three dimensions with clear physical meaning from the fused data to construct the feature vector:

[0061] Temporal fluctuation characteristics: Based on the first data (efficiency transient fluctuation), its standard deviation and mutation frequency are calculated to quantify the instability of the power output of the new energy unit itself.

[0062] Sensitivity characteristics: The equivalent Thevenin impedance parameter identified online by the second data is directly used as this characteristic, which directly reflects the sensitivity of the transformer area voltage to external power injection and is the core indicator characterizing the inherent vulnerability of the transformer area voltage.

[0063] Source-load time-series correlation characteristics: A dynamic time warping algorithm is used to calculate the distance between the daily power generation curve of new energy sources and the typical load curve. This distance value is used to quantify the severity of the mismatch between the source and load on the time axis.

[0064] S2.2 Input the voltage behavior feature vector into an unsupervised clustering model to identify data clusters.

[0065] The constructed multidimensional feature vectors are then input into unsupervised clustering models such as HDBSCAN (a hierarchical density-based clustering algorithm). This algorithm can automatically discover natural clusters in the data, with each cluster representing a common voltage behavior.

[0066] However, different clustering algorithms will produce different clustering results. In order to ensure that the most accurate pattern division can be obtained for different transformer areas and different data characteristics, the clustering algorithms are screened as follows:

[0067] 1. Construction of candidate algorithm set:

[0068] First, establish a candidate set containing multiple complementary clustering algorithms, for example:

[0069] The density-based algorithm, HDBSCAN, excels at discovering clusters of arbitrary shapes and identifying noise points.

[0070] Prototype-based algorithm: Gaussian mixture model, which is good at finding ellipsoidal clusters and can provide probabilistic attribution.

[0071] Partition-based algorithms: K-Means, computationally efficient, suitable for convex datasets.

[0072] This ensures a broad range of optimization options and avoids falling into the limitations of a single algorithm.

[0073] 2. Selection of Quantitative Evaluation Indicators: A comprehensive evaluation index is used instead of a single index to assess clustering quality from different dimensions. Key indicators include:

[0074] Silhouette coefficient: Used to measure the density of clusters within clusters and the separation between clusters. The closer the value is to 1, the better the clustering quality.

[0075] Calinski-Harabasz index: assesses clustering quality by the ratio of inter-cluster dispersion to intra-cluster dispersion; a higher value is better.

[0076] Davidson-Bauerdin index: In contrast to the CH index, a smaller value indicates a better clustering effect.

[0077] 3. Automated optimization process:

[0078] Using the aforementioned quantitative evaluation metrics, an automated grid search and evaluation process is executed:

[0079] Step a (Parameter Scan): Scan the key parameter space for each candidate algorithm. For example, scan the parameter space for HDBSCAN. (Minimum cluster size) and (Minimum number of samples); refers to the number of cluster components scanned by the GMM. ; Scanning for K-Means (Number of clusters).

[0080] Step b (clustering and evaluation): For each combination of algorithms and parameters, cluster the data on the voltage behavior feature vector dataset and calculate all the evaluation metrics mentioned above.

[0081] Step c (Comprehensive Score): Normalize the evaluation indicators (profile coefficient, CH index, etc.) of different dimensions and assign weights (e.g., profile coefficient weight 0.5, CH index weight 0.3, DB index weight 0.2), and calculate the weighted comprehensive score of each clustering result.

[0082] Step d (Optimal Selection): Select the algorithm and its parameter combination with the highest weighted comprehensive score as the optimal clustering model for the historical data of the current transformer area.

[0083] 4. Model solidification and application:

[0084] The optimal clustering model and its parameters obtained through optimization are solidified and used for online voltage behavior pattern identification in the corresponding distribution area over a subsequent period (e.g., one quarter). This optimization process is periodically (e.g., the following quarter) re-executed to adapt to changes in grid structure or load characteristics.

[0085] S2.3. Combining knowledge of power system operation, assign engineering labels with direct governance guidance significance to each data cluster, thereby identifying the typical voltage behavior patterns.

[0086] Based on the knowledge of power system operation experts, each data cluster identified in S202 is assigned an engineering label with direct governance guidance significance, namely, a typical voltage behavior pattern. This embodiment mainly defines three categories:

[0087] Efficiency-driven fluctuation type: This mode is characterized by time-series fluctuations exceeding a preset fluctuation threshold. This indicates that the voltage problem in this distribution area is mainly driven by the drastic and rapid fluctuations in the output of the renewable energy units themselves.

[0088] High impedance sensitive type: This mode is characterized by a sensitivity characteristic value (i.e., equivalent Thevenin impedance) that is higher than a preset sensitivity threshold. This indicates that the power grid structure of the transformer area is relatively weak, and even with small power fluctuations, it is very easy to cause voltage over-limit.

[0089] Source-load timing mismatch type: This mode is characterized by a source-load timing correlation characteristic value (dynamic time warping distance) that is higher than a preset correlation threshold. This indicates that the voltage problem is mainly caused by the time misalignment between the peak output of new energy sources and the peak load.

[0090] S3. Based on the identified voltage behavior patterns and the output prediction data of the new energy units, generate targeted and forward-looking reactive power optimization control commands.

[0091] This step, based on the S2 model diagnostic results and combined with predictive data, generates targeted treatment instructions.

[0092] In some embodiments, the specific process of S3 is as follows:

[0093] The system invokes a pre-established library of differentiated governance strategies that are strongly correlated with different voltage behavior patterns, and uses the output prediction data of the new energy unit as the core input.

[0094] If the problem is identified as being dominated by sudden efficiency changes, an inertial smoothing management strategy is initiated. Specifically, based on the output forecast data of the renewable energy units, the output change gradient within a future ultra-short-term time window (e.g., 5-15 minutes) is calculated, and this gradient is defined as the predicted fluctuation trend. Subsequently, a first reactive power optimization model is constructed with the objective of suppressing the expected voltage change rate caused by this trend. The control commands output by this model are used to control the inverter to output dynamic reactive current opposite to the predicted fluctuation trend in advance, thereby achieving rapid smoothing.

[0095] If the system is identified as high-impedance sensitive, a proactive preventative management strategy is initiated. Specifically, based on power output forecast data, the direction and magnitude of continuous changes in renewable energy output within a medium-term time window (e.g., 0.5-2 hours in the future) are determined and defined as the predicted power trend. A second reactive power optimization model is constructed with the goal of eliminating the expected voltage deviation generated by this trend on the high-impedance network. The instructions generated by this model control the static reactive power compensation device to output reactive power opposite to the predicted power trend in small steps, prior to the voltage overshoot point, thus achieving gentle, preventative voltage support.

[0096] If the problem is identified as a source-load timing mismatch, a spatiotemporal shift mitigation strategy is initiated. Specifically, based on output and load forecast data, a source-load power net difference curve is plotted on a future daily timescale, and the shape of this curve is defined as the predicted timing imbalance trend. A third reactive power optimization model is constructed to offset this timing imbalance trend and flatten the daily voltage curve. The instructions generated by this model are used to plan and control the operation sequence of adjustable transformer taps and capacitor-reactor groups throughout the day, such as performing voltage reduction operations before predicted overvoltage periods and voltage boosting and compensation operations before predicted undervoltage periods.

[0097] S4. Send the reactive power optimization control command to the voltage regulation device in the transformer area to control the voltage regulation device to perform voltage regulation and suppress the expected risk of voltage exceeding the limit in the future preset period.

[0098] This step sends the control commands generated by S3 to specific devices for execution, and adopts an intelligent execution mechanism for different modes.

[0099] In some embodiments, the specific process of S4 is as follows:

[0100] When the target of the governance is the efficiency-sudden change-dominant mode, a master-slave collaborative mechanism is adopted. The inverter with the fastest response is set as the master execution unit, responsible for quickly tracking and smoothing fluctuations; at the same time, the static reactive power compensation device is set as the slave execution unit to provide background reactive power support to compensate for the capacity limitation of the inverter. During this process, the master execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its own remaining regulation capacity and governance effect, so that the output of the slave execution unit is coordinated with the action of the master execution unit to jointly complete the fluctuation smoothing task.

[0101] More specifically, the process by which the main execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its remaining adjustment capacity and governance effect is as follows:

[0102] A1. The main execution unit monitors its output current in real time and calculates the difference between the current reactive power output and the rated capacity of the main execution unit as the real-time residual regulation capacity; at the same time, it monitors the voltage fluctuation rate of local key nodes. If the voltage fluctuation rate exceeds the smoothing target threshold, it is determined that there is a fluctuation power gap that has not been completely smoothed.

[0103] A2. Based on the fluctuating power gap and the real-time remaining adjustment capability, the initial output reference value of the slave execution unit is calculated through a first-order hysteresis filter function; the time constant of the filter function is set to be greater than the response period of the master execution unit but less than the response period of the slave execution unit to ensure the stability of the output reference command.

[0104] A3. Limit the initial output reference value, wherein the upper limit of the initial output reference value does not exceed the rated capacity of the slave execution unit, and the lower limit of the initial output reference value is not lower than the minimum reactive power allowed by the system; and issue the final output reference command after limiting to the slave execution unit.

[0105] When the target of the control is the high-impedance-sensitive mode, a voltage sensitivity feedforward mechanism is adopted. Before the control command is issued, the real-time updated equivalent Thevenin impedance parameter is used as a feedforward quantity to dynamically correct the amplitude of the control command. First, a linear relationship model between the command change and the voltage change is established, and its coefficient is defined as real-time command - voltage sensitivity. Then, the theoretically required voltage regulation is divided by this sensitivity to obtain the corrected command amplitude. This ensures that the command can produce the expected regulation effect when the grid impedance changes, avoiding under-adjustment or over-adjustment.

[0106] The more specific process is as follows:

[0107] B1. Based on the equivalent Thevenin impedance parameters, establish a linear relationship model between the change in the control command and the change in the voltage of the key node, and define the linear relationship coefficient of the linear relationship model as the real-time command-voltage sensitivity.

[0108] B2. Based on the control command, the theoretical voltage regulation amount required to eliminate the expected voltage deviation is determined;

[0109] B3. Divide the theoretical voltage regulation amount by the real-time command-voltage sensitivity to calculate the reactive power control command amplitude after feedforward correction; the calculation formula is: corrected command amplitude = theoretical voltage regulation amount / real-time command-voltage sensitivity.

[0110] B4. The control command using the modified command amplitude is sent to the static reactive power compensation device.

[0111] When the target of the governance is the source-load time-series mismatch mode, a plan-feedback verification mechanism is adopted. The real-time collected source-load short-term forecast data (including the power output and load forecast curves of new energy sources) is compared with the forecast reference curve on which the original control command was based. The absolute deviation or root mean square error of the source-load short-term forecast data and the forecast reference curve at key time-series points (such as power extreme points, zero crossing points) is calculated. If the deviation or error exceeds the reconstruction threshold, it is determined that the actual time-series imbalance trend has changed significantly, and the rapid recalculation of the third reactive power optimization model is triggered. Based on the new time-series imbalance trend obtained by recalculation, a new set of control commands is generated to replan and control the adjustable transformer tap and capacitor reactor group.

[0112] S5 provides real-time security protection and adaptive operation.

[0113] This step is executed in parallel with step S4, providing full-process security monitoring for the governance process.

[0114] C1. Command Tracking and Effect Evaluation: Within a preset time window T1 (e.g., 30 seconds to 5 minutes) after the control command is issued, the actual voltage change of the key node is monitored in real time at a fixed sampling period (e.g., 1 second), and the correlation coefficient R between the actual voltage change curve and the expected voltage change curve is calculated; a judgment time window T2 (e.g., 10 consecutive sampling periods, i.e., 10 seconds) is set. If the correlation coefficient R is lower than the reliability threshold for a continuous period of T2, the judgment is made accordingly. (For example, 0.7) indicates that the governance effect did not meet expectations and the system's effectiveness was not fully realized.

[0115] C2. Establish a dual safety protection mechanism with the voltage qualification range as the hard boundary and the voltage change rate as the soft boundary;

[0116] Hard boundary protection: Inside the acceptable voltage range, a stricter safety action boundary is preset (e.g., ±3% of the rated voltage). Once the actual voltage is detected to exceed this safety action boundary, it is determined to be approaching the hard boundary, and an emergency correction command is immediately generated and executed to restore the voltage to the safe target range within the acceptable voltage range.

[0117] Soft boundary protection: Once the voltage change rate is detected to exceed the soft boundary, a reverse damping command is immediately superimposed on the current control command. The magnitude of the reverse damping command is proportional to the degree to which the voltage change rate exceeds the limit, in order to suppress excessively rapid voltage changes.

[0118] For example:

[0119] Soft boundary threshold: Set the soft boundary of the voltage change rate dv / dt to ±1 V / s (that is, a change of more than 1 volt per second is considered too fast).

[0120] Proportional coefficient Kp: The proportional coefficient for the reverse damping command is set to 3 kVar / (V / s). This means that for every 1 V / s increase in the voltage change rate beyond the limit, 3 kVar of reverse reactive power is added. At time t0, due to a thick cloud suddenly blocking the sun, the voltage at a certain node begins to drop rapidly. The system monitors this in real time. The voltage dropped from 230.0V to 228.8V within seconds.

[0121] Calculate the instantaneous voltage change rate: dv / dt = (228.8 - 230.0)V / 0.5s = -2.4V / s.

[0122] The absolute value of the current rate of change, |dv / dt|, is 2.4V / s, which has exceeded the soft boundary threshold of 1V / s.

[0123] Calculate the degree of exceeding the limit: Exceeding limit value = |dv / dt| - soft boundary threshold = 2.4 - 1.0 = 1.4V / s.

[0124] Generate reverse damping command amplitude: damping reactive power kVar.

[0125] Since the voltage is decreasing (dv / dt is negative), to suppress this decrease, the voltage needs to be increased. Therefore, this reverse damping command should be capacitive reactive power (positive value). So, kVar.

[0126] Assume that the conventional forward-looking control command generated by the intelligent decision-making module at this time is kVar. The soft boundary protection mechanism immediately superimposes this inverse damping instruction onto the current instruction. Final instruction executed: The system urgently issues this +6.2kVar command to the Static Var Compensator (SVG) for execution.

[0127] C3. Whenever the dual-boundary security protection mechanism is triggered, the type of the event, the control command before the trigger, and the parameters of the protection command are recorded as a security event. Based on the statistical pattern of security events over a period of time, with the optimization goal of reducing false activation and refusal to activate the security protection mechanism, the threshold of the soft boundary, the position of the security action boundary, and the proportional coefficient of the reverse damping command are automatically adjusted to match the security protection strategy with the dynamic characteristics of the transformer area.

[0128] It should be noted that the specific process of generating and executing emergency correction instructions is as follows:

[0129] D1. Calculate the voltage deviation between the current actual voltage value and the median of the safety target range;

[0130] D2. Input the voltage deviation to a proportional-integral controller, the output of which is the reactive power value of the emergency correction command;

[0131] D3. Mark the emergency correction command as the highest priority and, breaking through the logic of conventional governance strategies, directly issue it to the reactive power regulation device with the fastest response speed in the current distribution area for execution.

[0132] Next, the method of the present invention will be described using specific numerical values.

[0133] Overview of the Taiwan area:

[0134] Name: Distribution station area in a residential area;

[0135] Load: 150 households, peak load in the evening;

[0136] New energy: Total capacity of distributed photovoltaic power generation is 400kW;

[0137] Problem: Voltage is too high at midday and too low at night, with a voltage qualification rate of only 89%.

[0138] Step S1: Collect multi-source electrical quantity data

[0139] 1. Collect initial data (tracking the root cause of fluctuations):

[0140] At 12:00:00 noon on a clear but cloudy day, data was collected from a photovoltaic inverter.

[0141] Measured DC side voltage V, DC side current A.

[0142] DC side instantaneous power kW.

[0143] Meanwhile, the output power was measured on the AC side. kW.

[0144] Calculate the transient value of power loss: kW.

[0145] Analysis: This instantaneous power loss of 1.0kW, compared to the usual stable 0.8kW, suddenly increases, indicating that a cloud layer is passing by, causing reduced sunlight, a change in the inverter's MPPT tracking point, and transient fluctuations in conversion efficiency. This is compared to the AC power... The drop was detected about 2-3 seconds earlier, becoming a prospective criterion for ultra-short-term mutations.

[0146] 2. Collect second data (to assess voltage sensitivity):

[0147] At 14:30:00, a high-power air conditioner (approximately 5kW) in the transformer area automatically started.

[0148] Before the event occurred, the voltage at node A was... V, load power kW.

[0149] After the event occurred, the voltage at node A... V, load power kW.

[0150] Calculate voltage change V.

[0151] Calculate the change in power kW.

[0152] Online identification of equivalent Thevenin impedance Ω. (Simplified calculation, phase angle ignored)

[0153] analyze: The relatively large value of Ω indicates that the grid structure of this transformer area is relatively weak and the voltage sensitivity is high.

[0154] Step S2: Identify typical voltage behavior patterns

[0155] Constructing feature vectors:

[0156] Temporal fluctuation characteristics: ΔP data from 10:00 AM to 12:00 PM were statistically analyzed, and its standard deviation was calculated to be 0.25 kW. The frequency of sudden changes (events exceeding 50% of the average loss) was 15 times / hour.

[0157] Sensitivity characteristics: directly using online identification Ω.

[0158] Source-load time series correlation characteristics: Taking typical daily data, the dynamic time warping distance between the photovoltaic output curve and the load curve is calculated to be 85 (dimensionless, the larger the value, the greater the difference in shape).

[0159] Clustering and Engineering Tags:

[0160] (Standard deviation = 0.25, (DTW distance = 85) This feature vector is input into the trained HDBSCAN model.

[0161] Based on historical data, the model determines that this vector belongs to a composite pattern of source-load time-series mismatch and efficiency-driven sudden change. Specifically, this manifests as a high risk of overvoltage due to high photovoltaic power generation and source-load mismatch at midday, coupled with the inherent volatility of photovoltaic power output.

[0162] Step S3: Generate forward-looking reactive power optimization control instructions

[0163] The pattern has been identified, and the system invokes the spatiotemporal translation governance strategy and the inertial smoothing governance strategy.

[0164] Based on forecast data: The weather forecast indicates sunny to partly cloudy skies from 13:00 to 15:00 in the afternoon, while the photovoltaic output forecast curve shows several rapid fluctuations. Load forecasts indicate peak electricity consumption from 19:00 to 21:00 in the evening.

[0165] Generate instructions:

[0166] For timing mismatch (spatiotemporal shift): The third reactive power optimization model outputs instructions to control the transformer tap changer to the step-down position (e.g., from step I to step II) when the predicted voltage exceeds 235V during the period from 13:00 to 15:00; and to control the capacitor bank to engage capacitive reactive power compensation during the period from 19:00 to 21:00.

[0167] For sudden efficiency changes (inertial smoothing): The first reactive power optimization model calculates the output change gradient in real time over the next minute. Assuming the calculated gradient is -5kW / min, a command is generated to control the inverter to output a dynamic reactive current of +3 kVar over the next minute to offset the voltage drop caused by the decrease in active power.

[0168] Steps S4 & S5: Perform adjustments and security safeguards

[0169] 1. Plan-Feedback Verification:

[0170] At 14:00, the original forecast was for sunny weather, but the actual source-load short-term forecast data (with thicker cloud cover) showed that the photovoltaic output was 15% lower than the original baseline curve.

[0171] The system calculated that the root mean square error of the two at the power extreme point was 12%, which exceeded the reconstruction threshold (10%).

[0172] Action: Immediately trigger the third reactive power optimization model for rapid recalculation. The new model determined that the risk of overvoltage at midday had decreased, and dynamically canceled the transformer tap-down operation originally scheduled for 14:10, avoiding unnecessary adjustments.

[0173] 2. Master-slave collaboration:

[0174] A thick cloud at 12:05 caused a sharp drop in photovoltaic power, and the inverter (main unit) had exhausted its own reactive power capacity.

[0175] The main unit calculates its remaining regulation capacity to be 0 and detects that the voltage fluctuation rate dv / dt still exceeds the smoothing target threshold.

[0176] It calculates using a first-order hysteresis filter (time constant τ=3s) and sends an output reference command of +10 kVar to the SVG (slave unit).

[0177] After receiving the command, the SVG outputs reactive power smoothly, thus smoothing out the voltage fluctuation.

[0178] 3. Security Protection:

[0179] After a certain treatment, within a 3-minute evaluation window, the correlation coefficient between the actual voltage and the expected voltage curve calculated by the system was only 0.5, and remained below the reliability threshold of 0.7 for 10 seconds, indicating that the effect did not meet expectations.

[0180] Meanwhile, the voltage at a certain node was monitored to rise to 244V (the safe operating boundary is set at 245V), approaching the hard boundary (250V).

[0181] Action: The safety protection mechanism was immediately activated, exceeding the conventional control logic, and an emergency correction command was forcibly issued, instantly boosting the SVG output to full capacitive reactive power, successfully pulling the voltage back to 238V, and avoiding voltage over-limit.

[0182] The incident was recorded, and the system automatically adjusted the reverse damping command ratio of that area to make it respond faster in the future.

[0183] Through the comprehensive management process in this embodiment, the problems of midday overvoltage and nighttime undervoltage in this transformer area have been significantly improved, with the voltage qualification rate increasing from 89% to 96%, and the accuracy and safety of the management actions greatly enhanced.

[0184] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for voltage management in distribution transformer areas with a high proportion of renewable energy access, characterized in that, Includes the following steps: Collect multi-source electrical quantity data from distribution transformers and user sides in the distribution area; Based on the multi-source electrical quantity data, the typical voltage behavior patterns of the distribution substation caused by new energy fluctuations are identified. Based on the identified voltage behavior patterns and the output prediction data of the new energy units, targeted and forward-looking reactive power optimization control commands are generated. The control command is sent to the voltage regulation device in the transformer area to control the voltage regulation device to perform voltage regulation and suppress the expected risk of voltage exceeding the limit in the future preset period.

2. The method for voltage management of distribution transformer areas with a high proportion of new energy access according to claim 1, characterized in that, The multi-source electrical quantity data is a fusion data set used to characterize the dual characteristics of new energy volatility and voltage sensitivity. The fusion data set includes first data for tracking the root causes of new energy volatility and second data for evaluating the sensitivity of transformer area voltage to power injection. The first data includes instantaneous power pairs on the DC and AC sides of the new energy power generation unit. The difference between the instantaneous power pairs is calculated to capture transient fluctuations in inverter energy conversion efficiency in real time, and the transient fluctuations are used as a forward-looking criterion for ultra-short-term power output mutations of the new energy power generation unit. The second data includes minute voltage change events generated by daily load switching or capacitor bank operation within the transformer area. By analyzing the voltage and power changes at the time of the event, the equivalent Thevenin impedance parameter of the transformer area is identified and updated online, and the equivalent Thevenin impedance parameter is used as a key weighting factor in the control command to determine the strength of the governance strategy.

3. The method for voltage management of distribution transformer areas with a high proportion of new energy access according to claim 2, characterized in that, The specific process for identifying the typical voltage behavior pattern based on the fused data set is as follows: A multi-dimensional voltage behavior feature vector with clear physical meaning is constructed. The voltage behavior feature vector includes time-series fluctuation features, sensitivity features, and source-load time-series correlation features. The time-series fluctuation features are based on the first data, extracting the standard deviation and mutation frequency of the transient fluctuations. The sensitivity features are based on the second data, using the equivalent Thevenin impedance parameter as the core indicator characterizing the inherent vulnerability of the transformer area voltage. The source-load time-series correlation features are calculated by measuring the dynamic time warping distance between the new energy daily power generation curve and the typical load curve, in order to quantify the severity of source-load mismatch. The voltage behavior feature vector is input into an unsupervised clustering model to identify data clusters; By combining knowledge of power system operation, each data cluster is assigned an engineering label with direct governance guidance significance, thereby identifying the typical voltage behavior patterns.

4. The method for voltage management of distribution transformer areas with a high proportion of new energy access according to claim 3, characterized in that, The typical voltage behavior modes include efficiency-driven, high-impedance-sensitive, and source-load timing mismatch, wherein the efficiency-driven mode is characterized by timing fluctuations exceeding a fluctuation threshold. The high impedance sensitive type is characterized in that the sensitivity characteristic value is higher than the sensitivity threshold. The source-load time mismatch type is characterized by the source-load time correlation feature value being higher than the correlation threshold.

5. A method for voltage management in distribution transformer areas with a high proportion of new energy access according to claim 1, characterized in that, The specific process for generating targeted, forward-looking reactive power optimization control commands based on the identified voltage behavior patterns and the output prediction data of the new energy units is as follows: The system invokes a pre-established library of differentiated governance strategies strongly correlated with different voltage behavior patterns, and uses the output prediction data of the new energy unit as the core input, wherein: If the problem is identified as being dominated by abrupt efficiency changes, an inertial smoothing management strategy is initiated: based on the power output prediction data, the gradient of new energy power output change within the ultra-short-term time window is calculated, and the gradient of change is defined as the predicted fluctuation trend; a first reactive power optimization model is constructed with the goal of suppressing the expected voltage change rate caused by the predicted fluctuation trend; control commands are generated through the first reactive power optimization model to control the inverter to output dynamic reactive current opposite to the predicted fluctuation trend in advance. If identified as a high impedance sensitive type, a forward-looking prevention and control strategy is initiated: Based on the power output prediction data, the direction and magnitude of the continuous change in new energy power output within the future medium-term time window are determined, and the direction and magnitude of the continuous change in new energy power output are defined as the predicted power trend; a second reactive power optimization model is constructed with the goal of eliminating the expected voltage deviation generated by the predicted power trend on the equivalent Thevenin impedance; control commands are generated through the second reactive power optimization model to control the static reactive power compensation device to output reactive power opposite to the predicted power trend in a small step manner before the voltage over-limit point, so as to maintain voltage stability; If the source-load time-series mismatch is identified, a spatiotemporal shift mitigation strategy is initiated: Based on the power output forecast data and load forecast data, a source-load power net difference curve is plotted on a future daily time scale, and the shape of the net difference curve is defined as the predicted time-series imbalance trend; a third reactive power optimization model is constructed with the goal of offsetting the time-series imbalance trend and flattening the daily voltage curve; control commands are generated through the third reactive power optimization model to plan and control the adjustable transformer taps and capacitor reactor groups, performing voltage reduction operations before the predicted overvoltage period and voltage boosting and compensation operations before the predicted undervoltage period.

6. A method for voltage management in distribution transformer areas with a high proportion of new energy access according to claim 5, characterized in that, The specific process of sending the reactive power optimization control command to the voltage regulation device within the distribution area to control the voltage regulation device to perform voltage regulation and suppress the expected risk of voltage exceeding the limit in the future preset period is as follows: When the target of governance is the efficiency-sudden change-dominant mode, a master-slave collaborative mechanism is adopted: the inverter with the fastest response is set as the master execution unit, responsible for quickly tracking and smoothing fluctuations; at the same time, the static reactive power compensation device is set as the slave execution unit to provide background reactive power support to compensate for the capacity limitation of the inverter. During the execution of the master-slave collaborative mechanism, the master execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its own remaining regulation capacity and governance effect, so that the output of the slave execution unit is coordinated with the action of the master execution unit to jointly complete the fluctuation smoothing task. When the target of the governance is the high impedance sensitive mode, a voltage sensitivity feedforward mechanism is adopted: before the control command is issued, the equivalent Thevenin impedance parameter updated in real time is used as the feedforward quantity to dynamically correct the amplitude of the control command, so as to ensure that the same control command can produce the expected voltage regulation effect when the grid impedance changes, and avoid under-regulation or over-regulation. When the target of governance is the source-load time-series mismatch mode, a plan-feedback verification mechanism is adopted: the real-time collected source-load short-term forecast data is compared with the forecast reference curve on which the original control command was generated, and the absolute deviation or root mean square error of the source-load short-term forecast data and the forecast reference curve at key time points is calculated. The source-load short-term forecast data includes the new energy output forecast curve and the load forecast curve within a preset time period. If the deviation or error exceeds the reconstruction threshold, it is determined that the actual time-series imbalance trend has changed significantly, and the rapid recalculation of the third reactive power optimization model is triggered. Based on the new timing imbalance trend obtained from the recalculation, a new set of control instructions is generated to re-plan and control the adjustable transformer tap changer and capacitor reactor group.

7. A method for voltage management in distribution transformer areas with a high proportion of new energy access according to claim 6, characterized in that, The specific process by which the main execution unit dynamically calculates and generates the output reference command of the slave execution unit based on its remaining adjustment capacity and governance effect is as follows: A1. The main execution unit monitors its output current in real time and calculates the difference between the current reactive power output and the rated capacity of the main execution unit as the real-time residual regulation capacity; at the same time, it monitors the voltage fluctuation rate of local key nodes. If the voltage fluctuation rate exceeds the smoothing target threshold, it is determined that there is a fluctuation power gap that has not been completely smoothed. A2. Based on the fluctuating power gap and the real-time remaining adjustment capability, the initial output reference value of the slave execution unit is calculated through a first-order hysteresis filter function; the time constant of the filter function is set to be greater than the response period of the master execution unit but less than the response period of the slave execution unit to ensure the stability of the output reference command. A3. Limit the initial output reference value, wherein the upper limit of the initial output reference value does not exceed the rated capacity of the slave execution unit, and the lower limit of the initial output reference value is not lower than the minimum reactive power allowed by the system; and issue the final output reference command after limiting to the slave execution unit.

8. A method for voltage management in distribution substations with a high proportion of renewable energy access according to claim 6, characterized in that, The specific process of dynamically correcting the amplitude of the control command by using the real-time updated equivalent Thevenin impedance parameter as a feedforward is as follows: B1. Based on the equivalent Thevenin impedance parameters, establish a linear relationship model between the change in the control command and the change in the voltage of the key node, and define the linear relationship coefficient of the linear relationship model as the real-time command-voltage sensitivity. B2. Based on the control command, the theoretical voltage regulation amount required to eliminate the expected voltage deviation is determined; B3. Divide the theoretical voltage regulation by the real-time command-voltage sensitivity to calculate the reactive power control command amplitude after feedforward correction. The calculation formula is: Corrected command amplitude = Theoretical voltage regulation amount / Real-time command - Voltage sensitivity; B4. The control command using the modified command amplitude is sent to the static reactive power compensation device.

9. A method for voltage management in distribution transformer areas with a high proportion of new energy access according to claim 6, characterized in that, The key timing points include power extreme points or zero-crossing points.

10. A method for voltage management in distribution transformer areas with a high proportion of new energy access according to claim 6, characterized in that, In addition to controlling the voltage regulating device to perform voltage regulation, it also includes: C1. Within a preset time window after the control command is issued, monitor the actual voltage change of key nodes in real time and calculate the correlation coefficient between the actual voltage change curve and the expected voltage change curve; if the correlation coefficient is consistently lower than the reliability threshold, it is determined that the treatment effect has not met expectations. C2. Establish a dual safety protection mechanism with the voltage acceptable range as the hard boundary and the voltage change rate as the soft boundary; inside the voltage acceptable range, a stricter safety action boundary is preset; once the actual voltage is detected to exceed this safety action boundary, it is determined to be approaching the hard boundary, and an emergency correction command is immediately generated and executed to restore the voltage to the safe target range within the voltage acceptable range; once the voltage change rate is detected to exceed the soft boundary, a reverse damping command is immediately superimposed on the current control command, and the magnitude of the reverse damping command is proportional to the degree of voltage change rate exceeding the limit. C3. Whenever the dual-boundary security protection mechanism is triggered, the type of the event, the control command before the trigger, and the parameters of the protection command are recorded as a security event. Based on the statistical pattern of security events over a period of time, with the optimization goal of reducing false activation and refusal to activate the security protection mechanism, the threshold of the soft boundary, the position of the security action boundary, and the proportional coefficient of the reverse damping command are automatically adjusted to match the security protection strategy with the dynamic characteristics of the transformer area.

Images