Voltage control system adaptive control method, system and device based on risk quantitative perception and medium

Abstract

The invention relates to the technical field of power system operation and control, and discloses a voltage control system adaptive control method, system, equipment and medium based on risk quantitative perception, and the method comprises the steps: deeply mining historical operation data and meteorological information through an artificial intelligence technology, constructing an online risk quantitative model, and outputting a dynamic risk factor; a dynamic mapping mechanism between a risk factor and an automatic voltage control optimization target is established, a safety weight and an economic weight in the control target are designed as an adaptive function of the risk factor, a risk adaptive strategy is embedded into a model prediction control rolling optimization framework, and a closed-loop control system is formed. According to the method, the voltage out-of-limit risk in an extreme scene can be remarkably reduced, meanwhile, the network loss is effectively reduced in conventional operation, and an innovative solution is provided for safe and economical operation of a high-uncertainty power grid.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, and in particular to an adaptive control method, system, device and medium for voltage control systems based on risk quantification perception. Background Technology

[0002] Domestic and foreign scholars have conducted extensive research on voltage control under uncertain environments, and their technical routes can be roughly divided into three stages.

[0003] The first stage is the traditional AVC method, which mainly relies on sensitivity analysis for reactive power optimization. This type of method calculates the sensitivity of each control device to node voltage to achieve coordinated allocation of reactive power resources. Studies have shown that traditional AVC is effective in dealing with steady-state and small-range fluctuations, but its shortcomings lie in its high dependence on accurate system models and measurements, and its control decisions are based on the current situation, lacking foresight regarding future changes. With the continuous increase in the penetration rate of new energy sources, output prediction errors can lead to significant voltage shifts within minutes, and the "lag" characteristic of traditional methods makes them unsuitable for online real-time control requirements.

[0004] The second stage introduces the AVC strategy using Model Predictive Control (MPC). MPC utilizes the system's predictive model to optimize a series of future control actions within each control cycle. Its forward-looking rolling optimization mechanism naturally aligns with the preventative control requirements of voltage. Research shows that MPC can effectively improve the dynamic performance of the system. However, most existing MPC-AVC research relies on deterministic predictions of renewable energy output points, neglecting the inherent uncertainty of the predictions themselves. When there is a significant deviation between the actual output and the predicted value, the optimization results based on deterministic predictions may not be optimal, and may even lead to erroneous control decisions, failing to fundamentally address the risks posed by uncertainty. Compared to some deep reinforcement learning methods that directly learn control strategies from data, MPC's reliance on model accuracy is its core limitation.

[0005] The third stage involves optimization control methods that consider uncertainty. To address predictive uncertainty, stochastic optimization and robust optimization have been introduced into the field of voltage control. Stochastic optimization characterizes the probability distribution of uncertain variables through methods such as scenario generation, aiming to optimize system performance under expected conditions. However, it has high computational complexity, and the typicality of the scenarios is difficult to guarantee. Robust optimization constructs a set of uncertainties, seeking solutions that still satisfy all constraints in the worst case, ensuring system safety. However, its main drawback lies in its "one-size-fits-all" conservatism; to withstand extremely low-probability events, it may sacrifice the economic efficiency of system operation under most normal operating conditions. It is worth noting that how to dynamically adjust this conservatism to match the actual risk level faced by the system is a current research challenge. The development of data-driven technologies provides a new approach to this, allowing for more precise guidance of the optimization process by mining risk patterns from massive amounts of data. Summary of the Invention

[0006] In view of the aforementioned existing problems, the present invention provides an adaptive control method, system, device and medium for voltage control systems based on risk quantification perception.

[0007] Therefore, the technical problem solved by this invention is that while existing research has made significant progress in dealing with voltage control uncertainty, there are still shortcomings: First, the risk characterization is insufficient, failing to fully explore the deep correlation between factors such as sudden weather changes and new energy prediction errors, making it difficult to accurately perceive low-probability, high-risk events; Second, the control strategy is static, and the conservatism of robust optimization or the expected goal of stochastic optimization remains unchanged throughout the entire control time domain once set, making it impossible to adaptively adjust according to the real-time changing risk level, resulting in the system either being too conservative or operating recklessly.

[0008] To address the aforementioned technical problems, this invention provides the following technical solution: an adaptive control method for a voltage control system based on risk quantification perception, comprising: calculating dynamic risk factors using a pre-trained artificial intelligence risk inference model based on real-time acquired renewable energy power prediction curves, meteorological forecast data, and power grid measurement information through a risk perception layer; dynamically adjusting the safety and economic weights in the model predictive control optimization objective function by using the dynamic risk factors as input through an adaptive control layer; and solving and executing the optimal reactive power and voltage control strategy based on the adjusted optimization objective function and power system physical constraints within the MPC rolling optimization framework.

[0009] As a preferred embodiment of the adaptive control method for voltage control system based on risk quantification perception described in this invention, the calculation of dynamic risk factors includes: constructing a multi-source historical database, performing feature engineering, extracting risk features related to new energy output fluctuations and extreme events, training a gradient boosting decision tree model as a risk inference model, performing online inference on real-time feature vectors, and outputting normalized dynamic risk factors.

[0010] As a preferred embodiment of the adaptive control method for voltage control system based on risk quantification perception described in this invention, the feature engineering includes: extracting dynamic features of predicted power from historical new energy power prediction data, extracting meteorological change features from meteorological data, and extracting morphological features from the predicted power curve to construct a feature vector for model training.

[0011] As a preferred embodiment of the adaptive control method for voltage control system based on risk quantification perception described in this invention, the adaptive control layer includes: constructing a risk-adaptive multi-objective optimization function, establishing a nonlinear mapping relationship between risk factors and weights, and dynamically balancing safety and economic objectives according to the real-time risk level.

[0012] As a preferred embodiment of the adaptive control method for a voltage control system based on risk quantification perception described in this invention, the construction of the risk-adaptive multi-objective optimization function includes the objective function J being constructed as a weighted sum of safety and economic terms: in, Representing the system's safety margin, it is quantified as the sum of squares of the deviations between the critical node voltages and their safety limits. , These are the safety weighting coefficient and the economic weighting coefficient, respectively. As a dynamic risk factor, the goal is to minimize the risk of voltage exceeding limits: in, For the weight coefficient corresponding to node i, , This is the weight coefficient for the over-limit penalty corresponding to the dominant node j. , This represents the actual voltage amplitude at the node. This refers to the set of all nodes in the system whose voltage needs to be monitored. As the dominant node set, These are the safe lower and upper limits for node voltage. Representing the system's operational economy, the active power network loss of the system is used as the indicator. This means the goal is to minimize losses: in, For branch road collection, , Let be the voltage phase angles at nodes i and j at the two ends of the branch, respectively. Let be the electrical conductance between nodes i and j.

[0013] As a preferred embodiment of the adaptive control method for voltage control system based on risk quantification perception described in this invention, the physical constraints of the power system include: power flow equations as equality constraints, node voltage safety limits, generator reactive power output range, dynamic reactive power compensation device output limits, limits on the number of parallel capacitor switching groups, and control equipment adjustment rate and frequency as inequality constraints.

[0014] As a preferred embodiment of the adaptive control method for voltage control system based on risk quantification perception described in this invention, the method is implemented in an automatic voltage control (AVC) system, wherein the risk perception layer and the adaptive control layer are integrated in the power grid dispatch center and connected to the SCADA / WAMS system through a real-time data interface to realize closed-loop risk perception and adaptive voltage control.

[0015] This invention provides an adaptive control system for a voltage control system based on risk quantification perception.

[0016] As a preferred embodiment of the voltage control system adaptive control system based on risk quantification perception described in this invention, it includes: a data acquisition and processing module, a risk perception module, an adaptive control decision module, and an execution and feedback control module. The system comprises a data acquisition and processing module, a risk perception module, an adaptive control decision-making module, and an execution and feedback control module. The data acquisition and processing module is responsible for collecting power grid operation data, new energy forecast data and meteorological information in real time, and performing data cleaning and feature extraction. The risk perception module, through its built-in artificial intelligence risk reasoning model, performs online analysis of input features and quantifies and outputs dynamic voltage risk factors in the near future. The adaptive control decision module dynamically adjusts the weights of safety and economy in the optimization objective based on real-time risk factors, and solves the optimal reactive voltage control strategy in a rolling manner while satisfying system constraints. The execution and feedback control module is responsible for converting control commands into actual operations of the reactive power regulation equipment, and continuously tracking the control effect through a closed-loop feedback mechanism to achieve dynamic adaptive adjustment of the system.

[0017] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of an adaptive control method for a voltage control system based on risk quantification perception.

[0018] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of an adaptive control method for a voltage control system based on risk quantification perception.

[0019] The beneficial effects of this invention are as follows: This invention proposes a new control paradigm that integrates closed-loop feedback between predicted uncertainty risk and AVC control objectives. It breaks down the barriers between the "prediction" and "control" links in traditional methods. By dynamically adjusting and optimizing the objective through online quantified risk factors, the control system can proactively adapt to changes in the uncertain environment, becoming more conservative when risk is high and pursuing economy when risk is low, thus achieving an intelligent closed-loop risk management system.

[0020] An AI-based online risk quantification method was designed to accurately detect low-probability, high-risk events. This method does not rely on any specific probability distribution assumptions, but instead directly learns the complex nonlinear mapping relationship between "predictive features and actual risks" from massive historical data. In particular, it extracts key features from historical weather mutations and extreme events, thereby significantly improving the early warning capability for potential large fluctuations in power output in the future.

[0021] A dynamic adaptive optimization mechanism for the safety and economic objectives of AVC was constructed, improving the level of control intelligence. By designing the safety and economic weights in the optimization objectives as functions of risk factors, the AVC system is endowed with the ability to make autonomous decisions under different risk levels, completing the evolution of AVC from deterministic control and static stochastic / robust control to dynamic risk optimization control. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This invention provides a two-layer collaborative control framework for an adaptive control method of a voltage control system based on risk quantification perception, as an embodiment of the present invention.

[0024] Figure 2 This invention provides a dynamic mapping mechanism for risk factors and optimization weights in an adaptive control method for a voltage control system based on risk quantification perception, as an embodiment of the present invention.

[0025] Figure 3 This paper presents a simulation comparison experiment design for an adaptive control method of a voltage control system based on risk quantification perception, as provided in an embodiment of the present invention.

[0026] Figure 4 A comparison of the dynamic risk factor R(t) under different scenarios for an adaptive control method of a voltage control system based on risk quantification perception provided in an embodiment of the present invention.

[0027] Figure 5 This diagram illustrates the comparison of active power and dynamic risk factors under different scenarios for an adaptive control method of a voltage control system based on risk quantification perception, provided as an embodiment of the present invention. Detailed Implementation

[0028] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0029] Example 1, the first embodiment of the present invention, provides an adaptive control method for a voltage control system based on risk quantification perception, comprising: S1: Construct an online risk quantification model based on artificial intelligence to achieve accurate perception of future voltage risks.

[0030] S2: Design a dynamic mapping mechanism between risk and control objectives to achieve an adaptive trade-off between safety and economy in the AVC strategy.

[0031] S3: Embed the AVC strategy into the MPC framework to form a closed-loop risk-aware preventive control.

[0032] It should be noted that this embodiment can effectively learn and identify prediction patterns similar to historical high-risk events from massive amounts of data without relying on specific uncertainty distribution assumptions, significantly improving the system's risk perception and forward-looking prediction capabilities for low-probability high-risk events such as sudden weather changes.

[0033] The constructed closed-loop feedback mechanism between risk and control objectives successfully transforms abstract risk quantification indicators into specific control tendencies, enabling the AVC control strategy to dynamically and adaptively seek a balance between safety and economy. This completes the evolution from deterministic control to stochastic risk optimization control and improves the level of intelligence in control.

[0034] Simulation results on the improved IEEE 39-bus system clearly demonstrate the comprehensive advantages of the proposed strategy: when dealing with extreme event impacts, it can significantly improve grid resilience through pre-emptive risk avoidance; and when the system experiences normal fluctuations, it can effectively improve operational economy by dynamically reducing safety weights.

[0035] Example 2, refer to Figure 1 and Figure 2 As an embodiment of the present invention, based on the above embodiment, an adaptive control method for a voltage control system based on risk quantification perception is provided.

[0036] Furthermore, in this embodiment, step S1 constructs an online risk quantification model based on artificial intelligence to achieve accurate perception of future voltage risks. Specific steps include S101-S103: S101: To achieve accurate risk perception and adaptive adjustment of control strategies, the following design is implemented: Figure 1 The diagram illustrates a two-layer collaborative control framework: a risk perception layer and an adaptive control layer. This framework operates under the rolling optimization mechanism of Model Predictive Control (MPC). It includes a risk perception layer and an adaptive control layer. S102: Risk Perception Layer, serving as the "outpost" and "brain" of the entire control system. Its core function is to calculate, online, the voltage instability risk faced by the system within a future control time domain (e.g., 15 minutes to 1 hour) based on real-time acquired renewable energy power prediction curves, weather forecast data, and power grid measurement information, using a pre-trained artificial intelligence risk reasoning model. This risk is quantified into a normalized dynamic risk factor. And transmit it to the adaptive control layer in real time.

[0037] S103: Adaptive control layer, the "decision-making" and "execution" center of the control strategy. It receives risk factors from the perception layer. This is used as a key input to dynamically adjust the weights representing system safety and economy in the MPC optimization objective function. Within one MPC rolling optimization cycle, this risk-adaptive optimization problem is solved to obtain the optimal reactive power and voltage control strategy at the current moment (such as generator excitation regulation, capacitor switching, and SVG output adjustment), and then the strategy is issued for execution.

[0038] The framework uses the rolling execution mechanism of MPC to continuously perform a closed-loop cycle of "risk perception - strategy adaptation - optimization solution - control execution", which enables the voltage control system to continuously follow the time-series changes in risk and make dynamic optimal decisions that take into account both safety and economy.

[0039] Furthermore, in this embodiment, step S2 designs a dynamic mapping mechanism between risk and control objectives to achieve an adaptive trade-off between safety and economy in the AVC strategy. Specific steps include S201-S204: S201: Construct a multi-source heterogeneous time-series database, mainly including: Power grid operation data, including historical key node voltages, line power flow, generator output, and reactive power equipment status collected by the SCADA / WAMS system; New energy data, including historical power forecast data and actual power output data of new energy power plants during the same period; Meteorological data, including historical meteorological data (such as wind speed, wind direction, total solar radiation, cloud cover, etc.) and weather forecast data corresponding to the geographical location of the new energy power station.

[0040] S202: The goal of feature engineering is to extract indicators from raw data that can effectively predict risk. Unlike conventional forecasting tasks, this study focuses on capturing “high-risk” features that may lead to large forecast deviations and drastic voltage fluctuations. Research shows that relying solely on the absolute value of the predicted power is insufficient to assess risk. Therefore, the key feature set includes: (1) Predicted power dynamic characteristics Predicted Ramp Rate: This characterizes the drastic degree of future power changes.

[0041] Predicted volatility: The standard deviation of the predicted series over a time window. It characterizes the stability of power.

[0042] Historical characteristics of prediction errors: recent prediction errors (e.g., the past hour). The statistical properties (mean, variance) reflect the performance of the current prediction model.

[0043] (2) Characteristics of meteorological changes Wind shear: The rapid rate of change of wind speed is an important precursor to sudden changes in wind power.

[0044] Irradiance drop: Characterizes a rapid decline in solar power output caused by factors such as cloud cover.

[0045] Extreme weather warnings: Encode weather warnings such as typhoons and blizzards as input features.

[0046] (3) Predicted curve shape characteristics: For the predicted power curves of the next 15 minutes or 1 hour, principal component analysis (PCA) or convolutional neural network (CNN) is used to extract their shape characteristics in order to identify curve patterns similar to historical high-risk events.

[0047] In an optional embodiment, the extraction of dynamic features of predicted power can also be achieved through moving window statistics. Specifically, using predicted power data within a fixed time window, a moving average and a moving range are calculated as features. The moving average reflects the recent trend of predicted power, and the moving range (the difference between the maximum and minimum values ​​within the window) characterizes the magnitude of power fluctuations.

[0048] In another optional embodiment, the extraction of dynamic features of predicted power can also be achieved through time series decomposition, specifically by decomposing the predicted power sequence into trend components, periodic components and residual components, and extracting the amplitude and rate of change of each component as features.

[0049] S203: Gradient Boosting Decision Tree (GBDT) is adopted as the core risk inference model. GBDT was chosen because of its strong ability to process tabular data, fast training speed, good interpretability, and ability to effectively handle nonlinear relationships, which aligns well with the needs of the power grid dispatch center. Compared to deep neural networks, GBDT is less prone to overfitting and less sensitive to hyperparameters in this task.

[0050] In an optional embodiment, the core risk inference model can also be a random forest model, specifically, which makes predictions by constructing multiple decision trees and aggregating their results. This model has strong capabilities for processing tabular data, fast training speed, and good resistance to overfitting and interpretability.

[0051] In another alternative embodiment, the core risk reasoning model can also be a multilayer perceptron model, specifically, handling nonlinear relationships through hidden layers and activation functions. This model can learn a complex mapping from input features to risk labels, and the loss function is optimized using gradient descent during training. A "risk label" is defined for each sample in the historical database. In this embodiment, the risk label is defined as the severity of voltage exceedance in post-event analysis or the adjustment cost incurred to maintain voltage stability. For example, the label for a sample can be defined as the probability of a level 3 or higher voltage exceedance event occurring in the system within the next hour, or the reactive power reserve consumption rate.

[0052] S204: The extracted historical features are used as the model input (X), and the corresponding historical risk event labels are used as the model output (Y). By minimizing the loss function (such as cross-entropy loss), the GBDT model is trained to learn the non-linear mapping relationship from the input features X to the risk labels Y. .

[0053] Once the model is trained, it can be deployed on the AVC system for online risk inference. At the beginning of each MPC control cycle, the system acquires the latest renewable energy power prediction curves and meteorological information in real time, and extracts online feature vectors using the same method as in the training phase. The feature vector is then input into the pre-trained GBDT model.

[0054] The model outputs a continuous value between 0 and 1, defined as the normalized dynamic risk factor. The derivation of this factor is as follows: in, For the current moment Extracted feature vectors The trained GBDT model. The closer the value is to 1, the more similar the model judges the current operating condition to historical high-risk event patterns, and the greater the risk of voltage instability the system faces in the next control cycle. Conversely, if... A value close to 0 indicates that the system is in a stable operating state.

[0055] Furthermore, in this embodiment, step S3 embeds the AVC strategy into the MPC framework to form a closed-loop risk-aware preventive control, specifically including steps S301-S30: S301: Obtaining Dynamic Risk Factors Then, a risk-adaptive multi-objective optimization function is constructed to guide the control strategy. The control objectives of AVC typically include both safety and economic aspects. Therefore, the objective function J is constructed as a weighted sum of safety and economic terms: in, Representing the system's safety margin, it is quantified as the sum of squares of the deviations between the critical node voltages and their safety limits. , These are the safety weighting coefficient and the economic weighting coefficient, respectively. As a dynamic risk factor, the goal is to minimize the risk of voltage exceeding limits: in, For the weight coefficient corresponding to node i, , This is the weight coefficient for the over-limit penalty corresponding to the dominant node j. , This represents the actual voltage amplitude at the node. This refers to the set of all nodes in the system whose voltage needs to be monitored. As the dominant node set, These are the safe lower and upper limits for node voltage. Representing the system's operational economy, the active power network loss of the system is used as the indicator. This means the goal is to minimize losses: in, For branch road collection, , Let be the voltage phase angles at nodes i and j at the two ends of the branch, respectively. Let be the electrical conductance between nodes i and j.

[0056] Safety weight and economic weight It is no longer a fixed constant, but rather a matter of risk factors. An adaptive function is used to establish a direct link between risk and control conservatism.

[0057] S302: The Sigmoid function is used to construct the above nonlinear mapping relationship, providing a smooth transition: in, and These are the upper and lower limits of the safety weight, for example, 0.9 and 0.1; This is the risk threshold, usually set to 0.5, representing the neutral point of risk perception; It is the slope parameter, which controls the degree of drastic change in weights.

[0058] When the risk perception layer judges that the future risk is very high ( ), It will approach its maximum value. At this point, the optimization objective function will mainly consist of the safety term. Under normal circumstances, AVC will adopt a more conservative strategy (such as reserving more reactive power reserves and maintaining the voltage in the middle of the safe range) to ensure safety. Conversely, when the system is running smoothly ( ), It will approach its minimum value Economic items weight As the voltage increases accordingly, AVC will adopt a more economical strategy (such as allowing the voltage to fluctuate within a wider range within a safe range to reduce regulation costs and network losses). This dynamic balancing effectively overcomes the shortcomings of traditional robust optimization methods, which are too conservative under all operating conditions.

[0059] In an optional embodiment, the nonlinear mapping relationship can also be constructed using a lookup table method. Specifically, the interval [0,1] of the risk factor R(t) is divided into several continuous sub-intervals, and within each sub-interval, the safety weight α changes linearly with R(t). By setting different slopes for different intervals, a smooth transition effect is achieved.

[0060] In another alternative embodiment, the nonlinear mapping relationship can also be constructed using piecewise linear functions. Specifically, a discrete mapping table is defined, specifying the optimal safety weight values ​​corresponding to several typical risk factor values. In online applications, the current safety weight α is obtained based on the real-time risk factor R(t) by looking up the table and using linear interpolation.

[0061] S303: The optimization solution of MPC must be performed under the condition that the physical operating limits of the power system are met. The constraints constitute the feasible region of the optimization problem, mainly including: equality constraints and power flow equation constraints, ensuring the balance of injected power and outflow power at each node.

[0062] Node voltage safety constraints: The voltage at all load nodes must be within permissible limits. Reactive power output of generators is constrained; the reactive power output of a generator cannot exceed its limit. Output constraints of dynamic reactive power compensation devices such as SVG: The constraint on the number of parallel capacitor switching groups: in, For the active power output of the generator at node i, Let i be the active power of the load at node i. Let i be the set of all adjacent nodes that are directly connected to node i via a line. The conductance of the branch connecting nodes i and j. The susceptance of the branch connecting nodes i and j. For the reactive power output of the generator at node i, Let i be the reactive power generated by the parallel reactive power compensation device at node i. Let i be the reactive power of the load at node i. Let be the lower and upper limits of the voltage safety for node i. Let be the lower and upper limits of the reactive power output of generator i. These are the lower and upper limits of the reactive power output of the SVG (Static Var Compensator). The actual reactive power output of the SVG at time t. Let i be the number of capacitor banks i that have been put into operation at time t. Let i be the total number of capacitor banks.

[0063] The control equipment's adjustment rate and number of adjustments are constrained, taking into account the mechanical life of equipment such as transformer taps and capacitors, limiting its adjustment rate and total number of adjustments within a certain time period.

[0064] These constraints collectively define a complex, nonlinear, mixed-integer optimization problem that needs to be solved efficiently within each MPC control step.

[0065] In an optional embodiment, constraint handling can also be achieved by hardening critical constraints and softening secondary constraints. Specifically, critical constraints such as node voltage safety are still treated as hard constraints that must be met. For non-critical constraints such as generator reactive power output range, they can be introduced into the objective function as soft constraints, allowing slight exceedances but imposing a high penalty term.

[0066] In another alternative embodiment, the constraint conditions can also be handled through constraint aggregation. Specifically, multiple inequality constraints of the same type are aggregated to summarize the voltage over-limit situations of all nodes into a system-wide voltage over-limit index constraint.

[0067] Example 3 is the third embodiment of the present invention, which differs from the previous two embodiments in that: This embodiment also provides an adaptive control system for a voltage control system based on risk quantification perception, including: a data acquisition and processing module, a risk perception module, an adaptive control decision module, and an execution and feedback control module; The data acquisition and processing module is responsible for collecting real-time power grid operation data, new energy forecast data, and meteorological information, and for performing data cleaning and feature extraction. The risk perception module uses a built-in artificial intelligence risk reasoning model to analyze input features online and quantify and output dynamic voltage risk factors in the near future. The adaptive control decision module dynamically adjusts the weights of safety and economy in the optimization objective based on real-time risk factors, and solves the optimal reactive voltage control strategy in a rolling manner while satisfying system constraints. The execution and feedback control module is responsible for converting control commands into actual operations of the reactive power regulation equipment, and continuously tracking the control effect through a closed-loop feedback mechanism to achieve dynamic adaptive adjustment of the system.

[0068] This embodiment also provides an electronic device, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the adaptive control method of the voltage control system based on risk quantification perception proposed in the above embodiment.

[0069] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the adaptive control method for a voltage control system based on risk quantification perception as proposed in the above embodiments.

[0070] The storage medium proposed in this embodiment and the adaptive control method for a voltage control system based on risk quantification perception proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0071] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0072] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0073] Example 4, refer to Figures 3-5 As an embodiment of the present invention, based on the above embodiment, an adaptive control method for a voltage control system based on risk quantification perception is provided.

[0074] To comprehensively evaluate the performance of the method in this embodiment, two typical scenarios were designed and compared with two mainstream methods.

[0075] Scenario 1 (Normal Fluctuation Scenario): Simulating normal weather conditions with clear skies and wind. In this scenario, the output of new energy sources experiences small-scale random fluctuations, but its prediction error generally conforms to a normal distribution, indicating a low-risk operating state.

[0076] Scenario 2 (Extreme Event Scenario): This scenario simulates a low-probability, high-risk event where solar power output forecasts deviate significantly due to severe afternoon convective weather in summer (such as the passage of thunderstorm clouds). Specifically, photovoltaic output drops sharply from 90% of the predicted value to 20% within 15 minutes, and then slowly recovers, posing a severe test to the system's voltage support capability.

[0077] Comparison of options: Option 1 (Traditional AVC): This method uses reactive power optimization based on current measurement information. Its goal is to minimize the voltage deviation and network loss at the current moment, but it does not have predictive capabilities.

[0078] Option 2 (Deterministic MPC-AVC): This option employs the standard MPC framework, where the optimization objective is fixed, and the safety and economic weights are set as constants. Furthermore, it directly uses the point prediction values ​​of new energy power for optimization, ignoring the uncertainty of the prediction.

[0079] Option 3 (Method of this embodiment): The dynamic strategy of risk perception driven by fusion data proposed in this embodiment is adopted.

[0080] Figure 4 This embodiment demonstrates the online generation of dynamic risk factors using the method described in this example, under both normal volatility scenarios and extreme event scenarios. The curves showing the changes in active power and the corresponding fluctuations in active power.

[0081] from Figure 4 It can be seen that in scenario one (normal fluctuations), the risk factor The consistently stable fluctuations below 0.2 indicate that the AI ​​model accurately determined the system to be in a safe state. However, in scenario two (extreme event), as the photovoltaic power output began to deviate significantly from the predicted value, the risk factor... The value rapidly increased, reaching a peak of 0.88 at the moment of power drop, which is far higher than in normal scenarios. This result strongly demonstrates that the AI-based risk quantification method proposed in this embodiment can successfully extract high-risk characteristics from the shape and trend of the prediction curve, providing an effective forward warning capability for upcoming extreme disturbances and a reliable decision-making basis for subsequent adaptive control.

[0082] Security comparison in extreme event scenarios Figure 5The voltage change curve of the critical bus 37, which is most severely affected, is shown under the impact of Scenario 2 (extreme event). Table 1 quantifies and compares the various safety performance indicators of the three schemes under this scenario.

[0083] Comparative Experimental Data 2: Critical Bus Voltage under Extreme Events [Insert here] Figure 5 The results show that Scheme 3 exhibits the smallest voltage drop and the fastest recovery, outperforming Schemes 1 and 2. Figure 5 Comparison of critical bus voltages under extreme events Table 1 Comparison of safety indicators under extreme event scenarios

[0084] analyze Figure 5 As shown in Table 1, during extreme shocks: Option 1 (traditional AVC) lacks foresight and only begins to passively adjust after a significant voltage drop, resulting in the voltage dropping as low as 0.89 pu, which is far beyond the safe lower limit (0.95 pu) and has the longest recovery time.

[0085] While Option 2 (Deterministic MPC) has predictive capabilities, it fails to prepare sufficient reactive power reserves in advance because it relies on erroneous deterministic predictions. When the actual power drops sharply, its control strategy lacks sufficient margin, causing the voltage to still drop to 0.92 pu, resulting in a prolonged period of over-limit operation.

[0086] Option 3 (the method in this embodiment) is based on the early detection of high risks ( (Raise), its control target automatically adjusts the safety weight. By increasing the voltage margin to above 0.8, the MPC proactively increased the system's reactive power reserve before the impact occurred, maintaining the voltage at critical nodes at a relatively high level close to 1.0 pu. Therefore, when the impact occurred, thanks to the ample margin, the system voltage only slightly dropped to 0.96 pu, remaining within the safe range and demonstrating optimal safety and the strongest system resilience.

[0087] Economic Comparison under Typical Fluctuation Scenarios Table 2 compares the economic performance indicators of the three schemes under scenario one (normal fluctuations) after running continuously for 24 hours.

[0088] Table 2 Comparison of economic indicators under normal fluctuation scenarios

[0089] Analysis of Table 2 shows that when the system is running smoothly with low risk: Option 2 (deterministic MPC) uses fixed safety weights, and its control strategy is relatively conservative. Maintaining a certain safety margin leads to additional network losses and adjustment costs.

[0090] Option 3 (the method in this embodiment) is due to risk factors. If it remains at a low level for an extended period, its optimization objective will automatically allocate more weight to economic efficiency. The increased voltage allows the control strategy to focus on finding the operating point with the minimum network loss, enabling more flexible voltage fluctuations within a safe range and reducing unnecessary reactive power adjustments. Compared to Scheme 2, the method in this embodiment reduces the total daily network loss by approximately 5.1% and significantly lowers equipment adjustment costs, demonstrating that it can effectively improve the system's operational economy while ensuring safety.

[0091] In summary, the risk-adaptive AVC strategy proposed in this embodiment demonstrates superiority across various scenarios. When facing extreme threats, it can proactively assess risks and adjust the system to a "high-alert" state, prioritizing security. During stable system operation, it can "lower its guard," enabling refined adjustments based on economic efficiency. This flexible and adaptable intelligent characteristic is not found in traditional AVC and deterministic MPC methods, achieving a dynamic optimal balance between security and economy in dynamic environments.

Claims

1. An adaptive control method for voltage control systems based on risk quantification perception, characterized in that: include, Based on real-time acquired new energy power prediction curves, meteorological forecast data and power grid measurement information, the risk perception layer calculates dynamic risk factors using a pre-trained artificial intelligence risk reasoning model. By using dynamic risk factors as input through an adaptive control layer, the safety weight and economic weight in the model predictive control optimization objective function are dynamically adjusted. In the MPC rolling optimization framework, the optimal reactive voltage control strategy is solved and executed based on the adjusted optimization objective function and power system physical constraints.

2. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 1, characterized in that: The calculation of dynamic risk factors includes constructing a multi-source historical database, performing feature engineering, extracting risk features related to fluctuations in new energy output and extreme events, training a gradient boosting decision tree model as a risk inference model, performing online inference on real-time feature vectors, and outputting normalized dynamic risk factors.

3. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 2, characterized in that: The feature engineering includes extracting dynamic features of predicted power from historical new energy power prediction data, extracting meteorological change features from meteorological data, and extracting morphological features from predicted power curves to construct feature vectors for model training.

4. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 3, characterized in that: The adaptive control layer includes constructing a risk-adaptive multi-objective optimization function, establishing a nonlinear mapping relationship between risk factors and weights, and dynamically balancing safety and economic objectives based on real-time risk levels.

5. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 4, characterized in that: The construction of the risk-adaptive multi-objective optimization function includes: The objective function J is constructed as a weighted sum of safety and economic terms: in, Representing the system's safety margin, it is quantified as the sum of squares of the deviations between the critical node voltages and their safety limits. , These are the safety weighting coefficient and the economic weighting coefficient, respectively. As a dynamic risk factor, the objective is to minimize the risk of voltage exceeding limits: in, For the weight coefficient corresponding to node i, , This is the weight coefficient for the over-limit penalty corresponding to the dominant node j. , This represents the actual voltage amplitude at the node. This refers to the set of all nodes in the system whose voltage needs to be monitored. As the dominant node set, These are the safe lower and upper limits for node voltage. Representing the system's operational economy, the active power network loss of the system is used as the indicator. This means the goal is to minimize losses: in, For branch road collection, , Let be the voltage phase angles at nodes i and j at the two ends of the branch, respectively. Let be the electrical conductance between nodes i and j.

6. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 5, characterized in that: The physical constraints of the power system include power flow equations as equality constraints, node voltage safety limits, generator reactive power output range, dynamic reactive power compensation device output limits, limits on the number of parallel capacitor banks to be switched, and control equipment adjustment rates and frequencies as inequality constraints.

7. The adaptive control method for a voltage control system based on risk quantification perception as described in claim 6, characterized in that: The method is implemented in an automatic voltage control (AVC) system, where the risk perception layer and the adaptive control layer are integrated in the power grid dispatch center and connected to the SCADA / WAMS system through a real-time data interface to achieve closed-loop risk perception and adaptive voltage control.

8. An adaptive control system for a voltage control system based on risk quantification perception, employing the adaptive control method for a voltage control system based on risk quantification perception as described in any one of claims 1 to 7, characterized in that, include: The system comprises a data acquisition and processing module, a risk perception module, an adaptive control decision-making module, and an execution and feedback control module. The data acquisition and processing module is responsible for collecting power grid operation data, new energy forecast data and meteorological information in real time, and performing data cleaning and feature extraction. The risk perception module, through its built-in artificial intelligence risk reasoning model, performs online analysis of input features and quantifies and outputs dynamic voltage risk factors in the near future. The adaptive control decision module dynamically adjusts the weights of safety and economy in the optimization objective based on real-time risk factors, and solves the optimal reactive voltage control strategy in a rolling manner while satisfying system constraints. The execution and feedback control module is responsible for converting control commands into actual operations of the reactive power regulation equipment, and continuously tracking the control effect through a closed-loop feedback mechanism to achieve dynamic adaptive adjustment of the system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the adaptive control method for a voltage control system based on risk quantification perception as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive control method for a voltage control system based on risk quantification perception as described in any one of claims 1 to 7.

Images