A transformer voltage active defense and autonomous recovery method

Abstract

The present application relates to a kind of table area voltage active defense and autonomous recovery method, comprising: obtaining the node monitoring data in table area, the node monitoring data in table area is input into voltage overrun risk probability prediction model, obtains the voltage overrun risk probability of future time section;Voltage overrun risk probability prediction model is obtained using training set to train deep neural network model;Set based on voltage and cost sensitivity, using the absolute value of sensitivity the adjustable resource in table area is divided into multiple economic echelon;When voltage overrun risk probability exceeds preset active defense threshold, then call the adjustable resource corresponding to economic echelon, reduce voltage overrun risk probability to safety threshold.The present application aims at solving the technical problems existing in the response lag, poor economy, dependent on after-regulation, unable to actively prevent voltage overrun in the existing voltage control strategy.

Description

Technical Field

[0001] This invention relates to the field of active distribution network technology, and in particular to a method for active protection and autonomous recovery of transformer substation voltage. Background Technology

[0002] As the penetration rate of distributed photovoltaic (PV) power in low-voltage distribution networks increases, the challenges it poses to the safe and stable operation of the power grid are becoming increasingly severe. "The voltage over-limit problem caused by the large-scale grid connection of low-voltage distributed PV, characterized by strong randomness and volatility, is becoming increasingly prominent." Especially during periods of ample sunshine and low electricity load (such as midday), large-scale PV output leads to power backflow to the upstream grid, a phenomenon known as "power flow reversal." This causes a significant rise in voltage along the feeder lines, particularly at the end of the lines, resulting in frequent voltage over-limit problems. Voltage over-limit not only seriously affects the power quality for users but can also damage electrical equipment and threaten the safety of grid equipment, becoming a core challenge in ensuring power quality under a high proportion of renewable energy integration.

[0003] To address the voltage over-limit problem, existing technologies mainly employ the following methods: Traditional mechanical voltage regulating equipment, such as on-load tap changers (OLTC) and line-connected capacitor banks (CB) of distribution transformers, has the following main limitations: slow response speed, usually on the order of minutes, making it unable to effectively track the rapid fluctuations in photovoltaic output on the order of seconds or minutes; coarse adjustment levels and limited control precision; and as mechanical equipment, its number of operations is limited, and frequent adjustments will seriously affect its service life and result in high operating costs.

[0004] Passive-Response Distributed Resource Regulation: With the increasing prevalence of distributed energy resources (DERs) and energy storage systems (ESS), utilizing the rapid regulation capabilities of their power electronic inverters has become a new technological direction. However, most current regulation strategies remain in a "passive-response" mode. That is, the system monitors the voltage in real time through voltage monitoring devices, and only triggers the control logic to calculate and issue regulation commands when the monitored value exceeds a preset safety threshold. This "post-event remedy" mode inevitably suffers from regulation lag; there is a time difference between the occurrence of voltage exceeding the limit and the effective implementation of the control command. During this period, users have already experienced substandard voltage, and the problem of instantaneous deterioration of power quality still exists.

[0005] The fundamental flaw in existing technologies lies in the lack of a proactive defense mechanism that integrates risk prediction, multi-resource coordination, and economic cost accounting. Their core idea is "treatment" rather than "prevention," failing to address problems before they arise. This paradigm shift from passive control to proactive risk management is crucial for enhancing grid resilience and the ability to accommodate new energy sources, yet existing technologies have not provided effective solutions. Therefore, a new method is urgently needed that can predict voltage exceedance risks in advance and proactively intervene by coordinating the scheduling of various distributed resources in a cost-optimal manner, thus nipping voltage problems in the bud. Summary of the Invention

[0006] In order to solve the problems existing in the prior art, the purpose of this invention is to provide a method for active defense and autonomous recovery of transformer voltage based on risk probability prediction and multi-level economic coordination, which aims to solve the technical problems of response lag, poor economy, reliance on ex-post adjustment, and inability to actively prevent voltage over-limit in existing voltage control strategies.

[0007] To achieve the above objectives, the present invention provides the following solution: A method for active protection and autonomous recovery of transformer substation voltage, comprising: The node monitoring data within the transformer area is acquired, and the node monitoring data within the transformer area is input into the voltage over-limit risk probability prediction model to obtain the voltage over-limit risk probability at future time sections; the voltage over-limit risk probability prediction model is obtained by training a deep neural network model using a training set. Set a sensitivity based on voltage and cost, and use the absolute value of the sensitivity to divide the adjustable resources in the transformer area into multiple economic tiers; When the probability of voltage exceeding the limit exceeds a preset active defense threshold, the adjustable resources corresponding to the economic tier are used to reduce the probability of voltage exceeding the limit to a safe threshold.

[0008] Optionally, the node monitoring data includes: weather data, photovoltaic output, load curves, and node voltage.

[0009] Optionally, training the deep neural network model using the training set includes: The training set is obtained; the training set includes: historical node monitoring data collected from similar transformer areas and local node monitoring data; The deep neural network model is trained using the historical node monitoring data. During the continuous training process, the deep neural network model gradually learns the deep time-series characteristics and nonlinear relationships between voltage dynamic changes and multidimensional input variables. The parameters of the trained deep neural network model are fine-tuned based on the local node monitoring data.

[0010] Optionally, setting the sensitivity includes: ; in, Sensitivity based on voltage and cost, To adjust the voltage change of target node i caused by the action of resource j. C j For economic costs.

[0011] Optionally, the economic costs include: energy storage charging and discharging costs, photovoltaic inverter reactive power regulation costs, and photovoltaic active power reduction costs.

[0012] Optionally, the economic echelon includes: The first tier is used for reactive power regulation using photovoltaic inverters; The second tier is used for charging and discharging using energy storage systems; The third tier is used for photovoltaic active power reduction.

[0013] Optionally, reducing the probability of voltage exceeding the limit to a safe threshold includes: When the voltage over-limit risk probability exceeds the first-level active defense threshold, the first echelon is invoked to regulate the reactive power of the photovoltaic inverter and reduce the voltage over-limit risk probability to a safe threshold. If the first echelon fails to reduce the voltage over-limit risk probability to a safe threshold, the second echelon is invoked to charge and discharge the energy storage system. If the second echelon fails to reduce the voltage over-limit risk probability to a safe threshold, the third echelon is invoked to reduce the active power of the photovoltaic system until the voltage over-limit risk probability is reduced to a safe threshold.

[0014] The beneficial effects of this invention are as follows: This invention utilizes a deep neural network model to predict the probability of voltage exceedance risks at future time points, using the probability of risk rather than real-time voltage exceedance as the triggering condition. This allows the system to intervene before voltage exceedances occur, transforming reactive post-event management into proactive pre-event defense, effectively ensuring power supply quality, achieving proactive voltage exceedance prevention, and solving the problem of delayed response in traditional control systems.

[0015] This invention constructs an evaluation index based on "voltage-cost" sensitivity, dividing adjustable resources into three economic tiers: low-cost (PV reactive power), medium-cost (energy storage), and high-cost (active power curtailment). By prioritizing the use of the low-cost tier and only gradually calling upon the high-cost tier when necessary, energy storage losses and PV curtailment are minimized, balancing grid security and operational economy. This achieves multi-level coordinated control and optimizes economics through tiered scheduling while ensuring voltage safety.

[0016] This invention adopts a deep transfer learning architecture of "cloud pre-training + edge fine-tuning". It uses massive historical data to pre-train the model to learn voltage timing features, and then uses local data to fine-tune the parameters. This not only effectively solves the problem of difficult model training and low accuracy caused by the small number of samples in a single transformer area, but also improves the model's adaptability to the electrical characteristics of different transformer areas, as well as its generalization ability and prediction accuracy in transformer areas with scarce data. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. 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.

[0018] Figure 1 This is a flowchart of a method for active protection and autonomous recovery of transformer voltage based on risk probability prediction and multi-level economic coordination, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the prediction process for the voltage over-limit risk probability in an embodiment of the present invention. Figure 3 This is a decision-making flowchart for multi-level economical collaborative defense in an embodiment of the present invention; Figure 4 This is a comparison curve of the key node voltage before and after adopting the present invention in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] like Figure 1As shown, this embodiment discloses a method for proactive voltage defense and autonomous recovery of transformer substations based on risk probability prediction and multi-level economic coordination. The method includes: acquiring node monitoring data within the transformer substation area; inputting the node monitoring data into a voltage exceedance risk probability prediction model to obtain the voltage exceedance risk probability at future time points; training a deep neural network model using a training set to obtain the voltage exceedance risk probability prediction model; setting a sensitivity based on voltage and cost; dividing adjustable resources within the transformer substation area into multiple economic tiers using the absolute value of the sensitivity; and when the voltage exceedance risk probability exceeds a preset proactive defense threshold, calling upon the adjustable resources corresponding to the economic tier to reduce the voltage exceedance risk probability to a safe threshold.

[0022] Furthermore, node monitoring data includes: weather data, photovoltaic output, load curves, and node voltage.

[0023] Furthermore, training a deep neural network model using a training set includes: acquiring a training set; the training set includes: historical node monitoring data collected from similar transformer areas and local node monitoring data; training a deep neural network model using historical node monitoring data, and during the continuous training process, the deep neural network model gradually learns the deep time-series characteristics and nonlinear relationships between voltage dynamic changes and multidimensional input variables; and fine-tuning the model parameters of the trained deep neural network model based on local node monitoring data.

[0024] Specifically, voltage limit exceedance risk probability prediction based on deep transfer learning, such as... Figure 2 As shown: The control logic of this invention is not triggered by the actual voltage measurement value, but driven by the predicted probability of future risks, realizing a fundamental shift from "passive response" to "active defense".

[0025] Prediction Objective: This invention does not predict a single future voltage curve, but rather directly calculates and outputs the "probability" that the voltage value of a key monitoring node (such as the end of a line or a node with a high historical incidence of voltage exceedances) within a short future time segment (e.g., 15 minutes) will exceed a safe threshold (e.g., 1.07 pu as specified in the national standard). This "risk probability" is the sole basis for initiating all subsequent proactive defense measures.

[0026] Model Construction and Application: This prediction function is based on the abnormal operating condition diagnosis method of deep transfer learning. First, on the cloud server, a deep neural network model (such as a composite model containing convolutional layers and long short-term memory network layers) is pre-trained using historical operating big data (including weather data, photovoltaic output, load curves, node voltage, etc.) collected from a large number of similar transformer areas. This enables the model to learn the deep temporal characteristics and nonlinear relationships between voltage dynamic changes and multidimensional input variables.

[0027] Edge fine-tuning and real-time prediction: The pre-trained generalized model in the cloud is distributed to edge computing terminals (such as converged terminals) deployed in the distribution area. The edge terminals use a small amount of locally accumulated historical data to fine-tune the model, making the model parameters more suitable for the unique electrical topology and load characteristics of the distribution area. In actual operation, the edge terminals use ultra-short-term photovoltaic output prediction data, load prediction data, and current real-time measurement data as model inputs, and the model can then output the probability of voltage over-limit risk in future time segments in real time and on a rolling basis.

[0028] Furthermore, economic costs include: energy storage charging and discharging costs, photovoltaic inverter reactive power regulation costs, and photovoltaic active power reduction costs.

[0029] Specifically, dynamic resource assessment and classification based on voltage-cost sensitivity: To achieve cost-optimal collaborative control, this invention introduces a unified metric for evaluating heterogeneous distributed resources.

[0030] Definition of "Voltage-Cost" Sensitivity: The "Voltage-Cost" sensitivity proposed in this invention ( This sensitivity serves as a core indicator for resource assessment. It not only includes traditional voltage-power sensitivity (i.e., how much voltage at the target node can be changed per unit of power adjustment, representing "technical effectiveness"), but also innovatively incorporates the "economic cost" per unit of adjustment, defined as: ; in, Sensitivity based on voltage and cost, To regulate resources j Target node caused by the action i The amount of voltage change. C j For economic costs.

[0031] Quantifying economic costs: The adjustment costs of different resources vary. This invention provides a refined quantification of costs, including at least the following: Energy storage charging and discharging costs: mainly the cost of cycle life loss, which is related to charging and discharging power and depth.

[0032] Reactive power regulation costs of photovoltaic inverters: mainly the costs of equipment wear and tear and reduced lifespan caused by long-term reactive power regulation.

[0033] The cost of curtailing solar power is the opportunity cost of curtailing solar power, which is equal to the amount of electricity cut off multiplied by the grid-connected electricity price. It is the most expensive regulation method.

[0034] Furthermore, the economic tiers include: the first tier, used for reactive power regulation using photovoltaic inverters; the second tier, used for charging and discharging using energy storage systems; and the third tier, used for active power reduction from photovoltaic systems.

[0035] Reducing the voltage over-limit risk probability to a safe threshold includes: when the voltage over-limit risk probability exceeds the first-level active defense threshold, the first-tier team is activated to regulate the reactive power of the photovoltaic inverter to reduce the voltage over-limit risk probability to a safe threshold; if the first-tier team fails to reduce the voltage over-limit risk probability to a safe threshold, the second-tier team is activated to charge and discharge the energy storage system; if the second-tier team fails to reduce the voltage over-limit risk probability to a safe threshold, the third-tier team is activated to reduce the active power of the photovoltaic system, until the voltage over-limit risk probability is reduced to a safe threshold.

[0036] Specifically, the economic tiers are dynamically divided: based on the real-time calculated absolute values ​​of the "voltage-cost" sensitivity of each resource, the edge terminal dynamically and in real-time divides all adjustable resources within the distribution area into three economic tiers. This division is not fixed, but changes dynamically with electricity prices, equipment status, and grid operating conditions.

[0037] First tier (zero / negative cost resources): This typically refers to reactive power regulation from photovoltaic inverters. Their regulation costs are extremely low, and they have the highest priority for deployment.

[0038] The second tier (medium-cost resources): typically consists of energy storage systems for charging and discharging. Their cost is mainly due to lifespan degradation, significantly lower than the cost of curtailment.

[0039] The third tier (high-cost resources): active power curtailment (curtailment of solar power). This is the most economically costly measure and should only be used as a last resort in emergency situations.

[0040] like Figure 3 As shown, the risk-driven, multi-level, economically coordinated defense control logic is as follows: This is the core control strategy of the present invention, which tightly couples risk probability with economic tiers, achieving a balance between safety and economy.

[0041] Triggering mechanism: When the "risk probability" predicted in step one exceeds the preset level one active defense threshold (e.g., 70%), the system initiates the active defense process.

[0042] Level 1 Active Defense (Cost-Optimal Defense): The system strictly follows the economic priority order, first and only utilizing resources in the first tier (photovoltaic reactive power). Through optimization algorithms, the system accurately calculates the minimum reactive power adjustment command required to reduce the predicted risk probability to below a safe threshold (e.g., 30%) using only the first tier resources, and sends it to the relevant photovoltaic inverters.

[0043] Level 2 Active Defense (Suboptimal Cost Defense): If, after all Level 1 resources have been utilized (e.g., the inverter's reactive power output reaches its capacity limit), the predicted risk probability still exceeds the safety threshold, the system will automatically and seamlessly activate Level 2 defense. At this point, the Level 1 defense commands are locked and remain in effect. The system uses the remaining risk probability as the control objective to optimize the scheduling of second-tier energy storage resources. Its optimization goal is to minimize the overall cost of energy storage while ensuring the elimination of risk.

[0044] Level 3 Emergency Defense (Safety-First Defense): The system will only activate Level 3 defense when the predicted risk probability is extremely high (e.g., exceeding 95%), and the risk cannot be completely eliminated even after all Level 1 and Level 2 resources have been mobilized. This is the last resort to ensure grid security. The system will calculate and execute the "minimum" photovoltaic active power reduction that "just" reduces the risk probability below the safety threshold, thereby minimizing economic losses while ensuring safety.

[0045] This embodiment uses an IEEE 33-node distribution network example to illustrate the execution process of the method of the present invention in detail.

[0046] Scenario setting: Power grid model: IEEE 33-node distribution network model.

[0047] Time: 12:00 noon on a summer day, when photovoltaic output reaches its peak, while industrial and residential electricity loads are at their daytime low.

[0048] Key monitoring node: Based on simulation analysis, node 18, which has the most serious voltage over-limit problem, is identified as the key monitoring node in this embodiment.

[0049] Step 1: Risk Prediction Edge terminals deployed in the distribution area acquire ultra-short-term weather forecasts and photovoltaic (PV) output prediction data for the next 15 minutes (12:00-12:15), showing that PV output will reach its daily peak; simultaneously, they acquire load prediction data, showing that the load will remain at a low level. The edge terminals input this data into a locally fine-tuned deep transfer learning prediction model. After calculation, the model outputs: "The probability that the effective voltage value of node 18 will exceed 1.07 pu within the next 15 minutes is 85%."

[0050] Step Two: Resource Assessment The system immediately performs real-time voltage-cost sensitivity calculations and assessments for all adjustable resources within the distribution area. Based on the configuration description of the IEEE 33-node case study (node ​​12 is configured with a 200kW / 400kWh energy storage system, and other nodes such as 16 and 17 are configured with photovoltaics), the assessment results are dynamically divided into: First tier: Reactive power regulation of photovoltaic inverters at nodes 16 and 17. Calculations show that they have high sensitivity to voltage regulation at node 18, and the regulation cost is close to zero.

[0051] The second tier consists of energy storage systems at node 12. Their charging and discharging capabilities moderately regulate the voltage at node 18, with the cost being the battery life loss resulting from each charge / discharge cycle.

[0052] The third tier: Reduce the active power output of photovoltaics at node 18 itself and nearby nodes. This measure has the most direct effect on regulating the voltage at node 18, but the opportunity cost caused by curtailment is the highest.

[0053] Step 3: Execute Level 1 Active Defense: Since the predicted risk probability of 85% is higher than the preset level 1 defense threshold of 70%, the system immediately activates level 1 active defense.

[0054] The system calls upon the first-tier resources and optimizes the calculations to find that: only by ordering the photovoltaic inverters of nodes 16 and 17 to jointly generate a total of 40 kVar of inductive reactive power, the predicted over-limit risk probability of node 18 can be reduced to 20%.

[0055] The predicted probability of 20% is below the safety threshold of 30%, thus meeting the control objective.

[0056] The edge terminal then sends the corresponding reactive power adjustment commands to the inverters at nodes 16 and 17.

[0057] Step Four: Scene Upgrade and Secondary Active Defense Ten minutes later (12:10), the weather suddenly changed. The predicted cloud cover suddenly dissipated, and the actual sunlight intensity far exceeded expectations, causing the actual output of photovoltaic power in the entire region to soar.

[0058] In the new rolling forecast cycle, the system re-predicts risks by combining the latest real-time data and finds that although the primary defense measures are in effect, the probability of voltage over-limit risk at node 18 has risen sharply to 90%.

[0059] The system detected that the reactive power output of the inverters at nodes 16 and 17 has reached their capacity limit, and the primary resources have been exhausted.

[0060] The system immediately and seamlessly activated Level 2 active defense. Based on voltage-cost sensitivity optimization calculations, the optimal strategy was determined: command the energy storage system at node 12 to charge at 50 kW for 15 minutes. Analysis suggests that this command not only effectively absorbs excess power and reduces the risk probability back to below 30%, but also minimizes charging costs due to the current timeframe being within the peak-valley electricity price range.

[0061] The edge terminal sends the charging command to the energy storage PCS at node 12.

[0062] Step 5: Result Verification Through background monitoring and data display, the application of this invention has achieved significant results. (See Table 1 and...) Figure 4 As shown, during the peak photovoltaic power generation period from 12:00 to 13:00: Voltage control effect: The actual maximum voltage of node 18 was successfully suppressed at 1.065 pu, and never reached the over-limit threshold of 1.07 pu, effectively avoiding the occurrence of voltage quality events.

[0063] Economic benefits: Only level one and level two defense resources were used throughout the process, and the most expensive level three defense was not triggered, meaning that no curtailment of solar power occurred, thus achieving full utilization of new energy and maximizing operational economics.

[0064] Table 1 This embodiment fully demonstrates that the method of the present invention can effectively predict and proactively defend against voltage over-limit problems caused by high proportion of photovoltaic grid connection, achieving the dual goals of grid security and economical operation.

[0065] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for active protection and autonomous recovery of transformer substation voltage, characterized in that, include: The node monitoring data within the transformer area is acquired, and the node monitoring data within the transformer area is input into the voltage over-limit risk probability prediction model to obtain the voltage over-limit risk probability at future time sections; the voltage over-limit risk probability prediction model is obtained by training a deep neural network model using a training set. Set a sensitivity based on voltage and cost, and use the absolute value of the sensitivity to divide the adjustable resources in the transformer area into multiple economic tiers; When the probability of voltage exceeding the limit exceeds a preset active defense threshold, the adjustable resources corresponding to the economic tier are used to reduce the probability of voltage exceeding the limit to a safe threshold.

2. The method for active voltage defense and autonomous recovery of transformer substations according to claim 1, characterized in that, The node monitoring data includes: weather data, photovoltaic output, load curves, and node voltage.

3. The active voltage defense and autonomous recovery method for transformer substations according to claim 1, characterized in that, Training the deep neural network model using the training set includes: The training set is obtained; the training set includes: historical node monitoring data collected from similar transformer areas and local node monitoring data; The deep neural network model is trained using the historical node monitoring data. During the continuous training process, the deep neural network model gradually learns the deep time-series characteristics and nonlinear relationships between voltage dynamic changes and multidimensional input variables. The parameters of the trained deep neural network model are fine-tuned based on the local node monitoring data.

4. The method for active voltage defense and autonomous recovery of transformer substations according to claim 1, characterized in that, Setting the sensitivity includes: ； in, Sensitivity based on voltage and cost, To regulate resources j Target node caused by the action i The amount of voltage change. C j For economic costs.

5. The active voltage defense and autonomous recovery method for transformer substations according to claim 4, characterized in that, The economic costs include: energy storage charging and discharging costs, photovoltaic inverter reactive power regulation costs, and photovoltaic active power reduction costs.

6. The method for active protection and autonomous recovery of transformer substation voltage according to claim 1, characterized in that, The economic echelon includes: The first tier is used for reactive power regulation using photovoltaic inverters; The second tier is used for charging and discharging using energy storage systems; The third tier is used for photovoltaic active power reduction.

7. The active voltage defense and autonomous recovery method for transformer substations according to claim 6, characterized in that, Reducing the probability of voltage exceeding the limit to a safe threshold includes: When the voltage over-limit risk probability exceeds the first-level active defense threshold, the first echelon is invoked to perform reactive power regulation of the photovoltaic inverter to reduce the voltage over-limit risk probability to a safe threshold. If the first echelon fails to reduce the voltage over-limit risk probability to a safe threshold, the second echelon is invoked to charge and discharge the energy storage system. If the second echelon fails to reduce the voltage over-limit risk probability to a safe threshold, the third echelon is invoked to reduce the photovoltaic active power until the voltage over-limit risk probability is reduced to a safe threshold.

Images