An intelligent optimization control method and system for on-load voltage regulation of a transformer

Abstract

The application discloses a kind of intelligent optimization control method and system of transformer on-load voltage regulation.The method comprises: obtaining transformer multidimensional operating state data to establish database;Using genetic algorithm, with voltage quality, network loss and tap changer action frequency as target to construct multi-objective optimization model, generate Pareto optimal solution set;Using fuzzy control, according to voltage deviation, voltage rate of change and load rate of change to determine optimal adjustment timing and amplitude, output fuzzy control instruction;Through intelligent decision algorithm, fuse Pareto optimal solution set and fuzzy control instruction, generate final adjustment instruction;Finally, through closed-loop feedback and online learning algorithm, continuously optimize control parameter.The application combines global optimization genetic algorithm with real-time response fuzzy control, realizes the comprehensive collaborative optimization of voltage quality, economy and equipment life, significantly improves the intelligent level of power grid voltage regulation and the overall benefit of system operation.

Description

Technical Field

[0001] This invention belongs to the field of power system automation technology. Specifically, it relates to a transformer voltage regulation control technology applied in smart grids, and in particular, a transformer on-load voltage regulation intelligent optimization control method and system that combines multi-objective optimization algorithms and intelligent decision-making. Background Technology

[0002] On-load tap changing of transformers is a key means of maintaining voltage stability in distribution networks. It changes the transformer ratio by adjusting the position of the transformer taps, thereby stabilizing the supply voltage at the user end. With the large-scale grid connection of renewable energy and the increasing complexity of electricity loads, unprecedented challenges have been posed to the refined and intelligent regulation of grid voltage.

[0003] In existing technologies, the control strategies for on-load tap changer transformers mainly rely on relatively simple control logic. For example, the traditional PID (proportional-integral-derivative) control method primarily adjusts the output voltage based on the deviation from the rated value. While this method offers a direct response, it can lead to excessively frequent tap changer operation under conditions of frequent load fluctuations. This not only increases grid losses but also accelerates the mechanical wear of the tap changer, shortening the equipment's lifespan.

[0004] To overcome the aforementioned shortcomings, some improvement methods have introduced expert systems or fuzzy logic control. These methods simulate expert experience by setting a series of "if-then" rules, comprehensively considering multiple factors such as voltage deviation and load change rate to make decisions, thus reducing unnecessary tap changes to some extent. However, the rule base of these methods is usually fixed, making it difficult to adapt to the diversity and dynamic changes in power grid operating conditions. More importantly, they are essentially based on real-time decision-making using local information, lacking a global and long-term optimization perspective.

[0005] Therefore, existing technologies generally face a core technical problem: how to simultaneously achieve multi-objective synergistic optimization while meeting grid voltage quality requirements and considering operational economy (such as reducing grid losses) and equipment lifespan (such as controlling tap changer operation frequency). Existing methods often focus on optimizing a single objective, making it difficult to find a globally optimal balance point among these mutually constraining objectives, resulting in low overall system operating efficiency. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide an intelligent optimization control method and system for on-load tap changer of transformers, which aims to overcome the defects of the prior art that only focuses on a single objective, lacks global optimization capability, and is difficult to balance multiple constraints such as voltage quality, economy and equipment life.

[0007] To solve the above-mentioned technical problems, the present invention provides an intelligent optimization control method for on-load tap changing of transformers, comprising the following steps: The transformer load current, voltage fluctuation, tap position, and ambient temperature are acquired. The transformer load current, voltage fluctuation, tap position, and ambient temperature are collected in real time through a sensor network to establish an operating status database. Based on the historical operating data and real-time status information in the operating status database, a genetic algorithm is used to encode the tap adjustment strategy into chromosomes and construct a multi-objective optimization model. With the optimization objectives of minimizing voltage quality deviation, reducing network loss, and controlling tap operation frequency, a fitness evaluation function is established to generate a Pareto optimal solution set. Based on the voltage deviation, voltage change rate, and load change rate in the operating status database, the voltage deviation, voltage change rate, and load change rate are fuzzified, and fuzzy inference rules for tap adjustment are formulated. The optimal adjustment timing and adjustment range are determined through fuzzy inference and defuzzification, and fuzzy control commands are output. The Pareto optimal solution set is used as a candidate solution library by intelligent decision-making algorithm, the fuzzy control command is used as a real-time constraint, and the comprehensive score of each candidate solution is calculated by weighted comprehensive evaluation method. The solution with the highest score is selected to generate the tap changer adjustment command. Using the tap adjustment command, the tap switch controller is driven to perform a gear adjustment action and collect the adjusted voltage and current parameters. The weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule are updated through an online learning algorithm to obtain optimized control parameters.

[0008] As a preferred technical solution, the transformer load current, voltage fluctuation, tap position, and ambient temperature are acquired. A sensor network is used to collect these parameters in real time, establishing the operational status database. This includes: installing intelligent power sensors on the high and low voltage sides of the transformer to collect three-phase voltage, current, and power factor, with a sampling frequency of 1 second / time, based on the transformer load current, voltage fluctuation, tap position, and ambient temperature; installing position sensors on the tap changer to monitor the current tap position in real time, acquiring tap position data; arranging a temperature sensor array inside the transformer tank to monitor the oil temperature distribution, acquiring ambient temperature data; and transmitting the electrical parameter data, tap position data, and ambient temperature data to a monitoring center via a wireless communication module, establishing the operational status database containing historical operational data, real-time status information, and predictive data.

[0009] As another preferred technical solution, based on historical operating data and real-time status information in the operating status database, a genetic algorithm is used to encode the tap changer regulation strategy into chromosomes and construct a multi-objective optimization model, establish the fitness evaluation function, and generate the Pareto optimal solution set. This includes: encoding the tap changer regulation strategy into chromosomes based on the historical operating data and real-time status information, with each gene representing a regulation scheme within a specific time window, and defining an individual coding scheme; based on the individual coding scheme, constructing the fitness evaluation function F=w1×f1+w2×f2+w3×f3, where f1 calculates the root mean square value of the voltage deviation, f2 calculates the system active power loss, and f3 counts the number of tap changer switching actions, and establishing a multi-objective optimization model; based on the multi-objective optimization model, setting the population size to 100, the crossover probability to 0.8, and the mutation probability to 0.05, and generating the Pareto optimal solution set through selection, crossover, and mutation operations.

[0010] Furthermore, based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt in the operating status database, fuzzification processing is performed on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt, and fuzzy inference rules for tap adjustment are formulated. This includes: based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt, dividing the voltage deviation into seven linguistic variables: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large; dividing the voltage change rate and the load change rate into three linguistic variables: negative, zero, and positive; and establishing membership functions for the input variables. Based on the membership functions, a fuzzy controller with the adjustment level and adjustment time as output variables is established, and 49 fuzzy inference rules are formulated. Based on the 49 fuzzy inference rules and the membership functions, defuzzification processing is performed using the centroid method, and the fuzzy control command is output.

[0011] Specifically, the 49 fuzzy inference rules include: when the voltage deviation is large and the voltage change rate is positive, immediately adjust down by 2 levels; when the voltage deviation is large and the voltage change rate is negative, immediately adjust up by 2 levels; when the voltage deviation is zero and the load change rate is zero, keep the current level unchanged; and based on different combinations of the voltage deviation, the voltage change rate, and the load change rate, establish the 49 fuzzy inference rules covering all operating conditions.

[0012] As another preferred technical solution, an intelligent decision-making algorithm is used to use the Pareto optimal solution set as a candidate solution library and the fuzzy control command as a real-time constraint. A weighted comprehensive evaluation method is then employed to calculate the comprehensive score of each candidate solution. This includes: designing a decision fusion device based on the Pareto optimal solution set and the fuzzy control command, using the Pareto optimal solution set as the candidate solution library and the fuzzy control command as a real-time constraint; establishing a dynamic decision weight allocation mechanism based on the decision fusion device, prioritizing the Pareto optimal solution set under normal operating conditions and the fuzzy control command under emergency conditions; and calculating the comprehensive score of each candidate solution using a weighted comprehensive evaluation method based on the dynamic decision weight allocation mechanism, selecting the solution with the highest score to generate the tap changer adjustment command.

[0013] Furthermore, driving the tap changer controller to perform gear adjustment actions and collecting the adjusted voltage and current parameters includes: based on the tap adjustment command, driving the tap changer controller to perform gear adjustment actions; immediately collecting the adjusted voltage and current parameters after each gear adjustment action; calculating the deviation between the actual adjustment effect and the expected target to obtain adjustment effect evaluation data; based on the adjustment effect evaluation data, establishing a performance evaluation index system including voltage qualification rate, adjustment accuracy, and response time; based on the performance evaluation index system, automatically starting a parameter retuning program when the adjustment effect is unsatisfactory after multiple consecutive adjustments to obtain optimized control parameters.

[0014] As another preferred technical solution, the optimized control parameters are obtained by updating the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm. This includes: updating the weight coefficients of w1, w2, and w3 in the fitness evaluation function using a gradient descent algorithm based on the adjustment effect evaluation data and the performance evaluation index system; updating the membership function parameters of each linguistic variable in the fuzzy inference rule using an adaptive adjustment algorithm based on the deviation between the adjustment effect and the expected target; and establishing an anomaly detection module based on the weight coefficients and the membership function parameters, which automatically switches to a safe mode and issues an alarm signal when a device malfunction or abnormal operating condition is detected, thereby obtaining the optimized control parameters.

[0015] To address the aforementioned technical problems, the present invention also provides an intelligent optimization control system for on-load tap changing of transformers, comprising: The data acquisition module is used to acquire transformer load current, voltage fluctuation, tap position and ambient temperature. It collects the transformer load current, voltage fluctuation, tap position and ambient temperature in real time through a sensor network and establishes an operating status database. The genetic algorithm optimization module is used to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model based on the historical running data and real-time status information in the running status database, establish a fitness evaluation function, and generate a Pareto optimal solution set. The fuzzy control module is used to perform fuzzification processing on the voltage deviation, voltage change rate and load change rate based on the voltage deviation, voltage change rate and load change rate in the operating status database, formulate fuzzy inference rules for tap adjustment, and output fuzzy control commands. The intelligent decision-making module is used to generate tap changer adjustment instructions based on the Pareto optimal solution set and the fuzzy control instructions, using an intelligent decision-making algorithm to use the Pareto optimal solution set as a candidate solution library, the fuzzy control instructions as real-time constraints, and a weighted comprehensive evaluation method to calculate the comprehensive score of each candidate solution. The closed-loop feedback module is used to drive the tap changer controller to perform a gear adjustment action based on the tap adjustment command and to collect the adjusted voltage and current parameters. It updates the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain optimized control parameters.

[0016] The beneficial effects of this invention are as follows: 1. Strong comprehensive optimization capability: By constructing a multi-objective optimization model with voltage quality, network loss and tap changer operation frequency as objectives, this invention can effectively balance the safety, economy and equipment reliability of power grid operation, and maximize the overall benefits of the system.

[0017] 2. Balancing global and real-time decision-making: The system creatively combines a genetic algorithm that provides a global optimization solution set with a fast-responding and highly adaptable fuzzy control. Through an intelligent decision fusion device, the control strategy has both long-term optimization goals and the ability to cope with real-time changes in operating conditions, thereby improving the robustness and accuracy of the control.

[0018] 3. Possesses adaptive and self-learning capabilities: By establishing a closed-loop feedback and online learning mechanism, the system can continuously optimize its control model parameters (such as the weights of the genetic algorithm and the membership function of the fuzzy rule) based on the actual adjustment effect, thereby achieving adaptive evolution of the control strategy and maintaining optimal control performance in the long term.

[0019] 4. Enhance the intelligence level of the power grid: This invention provides a complete intelligent control closed loop of "data acquisition - global optimization - real-time decision-making - feedback learning", which provides an efficient and intelligent solution for transformer on-load tap change and even broader power grid control problems.

[0020] To address the aforementioned technical problems, the present invention also provides an intelligent optimization control system for on-load tap changing of transformers, comprising: The data acquisition module is used to acquire transformer load current, voltage fluctuation, tap position and ambient temperature. It collects the transformer load current, voltage fluctuation, tap position and ambient temperature in real time through a sensor network and establishes an operating status database. The genetic algorithm optimization module is used to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model based on the historical running data and real-time status information in the running status database, establish a fitness evaluation function, and generate a Pareto optimal solution set. The fuzzy control module is used to perform fuzzification processing on the voltage deviation, voltage change rate and load change rate based on the voltage deviation, voltage change rate and load change rate in the operating status database, formulate fuzzy inference rules for tap adjustment, and output fuzzy control commands. The intelligent decision-making module is used to generate tap changer adjustment instructions based on the Pareto optimal solution set and the fuzzy control instructions, using an intelligent decision-making algorithm to use the Pareto optimal solution set as a candidate solution library, the fuzzy control instructions as real-time constraints, and a weighted comprehensive evaluation method to calculate the comprehensive score of each candidate solution. The closed-loop feedback module is used to drive the tap changer controller to perform a gear adjustment action based on the tap adjustment command and to collect the adjusted voltage and current parameters. It updates the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain optimized control parameters.

[0021] The beneficial effects of this invention are as follows: 1. Strong comprehensive optimization capability: By constructing a multi-objective optimization model with voltage quality, network loss and tap changer operation frequency as objectives, this invention can effectively balance the safety, economy and equipment reliability of power grid operation, and maximize the overall benefits of the system.

[0022] 2. Balancing global and real-time decision-making: The system creatively combines a genetic algorithm that provides a global optimization solution set with a fast-responding and highly adaptable fuzzy control. Through an intelligent decision fusion device, the control strategy has both long-term optimization goals and the ability to cope with real-time changes in operating conditions, thereby improving the robustness and accuracy of the control.

[0023] 3. Possesses adaptive and self-learning capabilities: By establishing a closed-loop feedback and online learning mechanism, the system can continuously optimize its control model parameters (such as the weights of the genetic algorithm and the membership function of the fuzzy rule) based on the actual adjustment effect, thereby achieving adaptive evolution of the control strategy and maintaining optimal control performance in the long term.

[0024] 4. Enhance the intelligence level of the power grid: This invention provides a complete intelligent control closed loop of "data acquisition - global optimization - real-time decision-making - feedback learning", which provides an efficient and intelligent solution for transformer on-load tap change and even broader power grid control problems. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of an intelligent optimization control method for on-load tap changing of a transformer, provided as an embodiment of the present invention.

[0027] Figure 2 This is a structural block diagram of an intelligent optimization control system for on-load tap changing of a transformer, provided in an embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram illustrating the data acquisition and operation status database establishment in an embodiment of the present invention.

[0029] Figure 4 This is a schematic diagram of the multi-objective optimization process of the genetic algorithm in an embodiment of the present invention.

[0030] Figure 5 This is a schematic diagram of the fuzzy control logic in an embodiment of the present invention.

[0031] Figure 6 This is a schematic diagram of the intelligent decision-making and closed-loop feedback mechanism in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0033] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0034] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0035] 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, not all, of the embodiments of the present invention. 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.

[0036] Example 1 This embodiment provides an intelligent optimization control method for on-load tap changing of transformers. For example... Figure 1 As shown, this method is the core process of the present invention, and subsequent embodiments are refinements or preferred solutions to specific steps in this embodiment. The method specifically includes the following steps: S1: Acquire transformer load current, voltage fluctuation, tap position and ambient temperature, and collect the transformer load current, voltage fluctuation, tap position and ambient temperature in real time through a sensor network to establish an operating status database. The sensor network includes intelligent power sensors installed on the high and low voltage sides of the transformer, position sensors installed on the tap switch device and temperature sensor array arranged in the transformer tank. S2: Based on the historical operating data and real-time status information in the operating status database, a genetic algorithm is used to encode the tap changer adjustment strategy into chromosomes and construct a multi-objective optimization model. With the optimization objectives of minimizing voltage quality deviation, reducing network loss, and controlling tap changer operation frequency, a fitness evaluation function F=w1×f1+w2×f2+w3×f3 is established to generate the Pareto optimal solution set, where f1 calculates the root mean square value of voltage deviation, f2 calculates the system active power loss, f3 counts the number of tap changer switching operations, and w1-w3 are user-defined weights; S3: Based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt in the operating status database, perform fuzzification processing on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt, and formulate fuzzy inference rules for tap adjustment. Determine the optimal adjustment timing and adjustment range through fuzzy inference and defuzzification processing, and output fuzzy control commands. S4: Using an intelligent decision-making algorithm, the Pareto optimal solution set is used as a candidate solution library, the fuzzy control command is used as a real-time constraint, and a weighted comprehensive evaluation method is used to calculate the comprehensive score of each candidate solution. The solution with the highest score is selected to generate the tap changer adjustment command. S5: Using the tap adjustment command, drive the tap switch controller to perform a gear adjustment action and collect the adjusted voltage and current parameters. Update the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain optimized control parameters.

[0037] Specifically, the detailed explanation of the above steps is as follows: S1: Data Acquisition and Database Construction. This step is fundamental to the entire intelligent control system. By deploying a comprehensive sensor network at key locations on the transformer, multi-dimensional and high-frequency data acquisition of the transformer's operating status is achieved. The established operating status database not only contains real-time status information but also integrates historical operating data and data-driven predictive information, providing reliable and comprehensive data support for subsequent optimization algorithms and decision-making. The specific structure of the sensor network and data processing are further detailed in Example 2.

[0038] S2: Global Optimization via Genetic Algorithm. This step aims to find an optimal control strategy that balances voltage quality, operational economy, and equipment lifespan from a global and long-term perspective. The genetic algorithm, by simulating biological evolution, transforms the complex control problem into chromosome encoding and fitness evaluation. By setting a multi-objective fitness function, the algorithm can search a large solution space, ultimately generating a set of Pareto optimal solutions. Each solution in this set represents an optimal strategy with different preferences (e.g., some prioritize voltage quality, others energy saving), providing diverse candidate solutions for the final decision. The specific parameters and execution process of the genetic algorithm are further elaborated in Example 3.

[0039] S3: Real-time Response of Fuzzy Control. This step aims to address real-time fluctuations and uncertainties in power grid operation. Fuzzy control excels at handling complex, nonlinear, and time-varying systems, transforming operator experience and knowledge into fuzzy rules. By fuzzifying key indicators such as voltage deviation, voltage change rate, and load change rate, the system can quickly determine the current urgency and adjustment needs, thereby identifying the optimal timing and magnitude of adjustment. This compensates for the shortcomings of genetic algorithms, such as long computation time and difficulty in handling sudden situations. The specific design of the fuzzy controller is described in more detail in Examples 4 and 5.

[0040] S4: Intelligent Decision Fusion. This step is one of the key innovations of this invention, organically combining global optimization with real-time response. This step designs a decision fusion device that uses the Pareto optimal solution set generated by a genetic algorithm as a "candidate solution library" and the instructions output by fuzzy control as "real-time constraints" or "immediate demands." Through a weighted comprehensive evaluation method, each solution in the candidate library is scored, and the solution with the highest score is selected as the final tap changer adjustment instruction. This fusion mechanism ensures that control decisions possess both strategic foresight (from the genetic algorithm) and tactical flexibility (from fuzzy control). The specific mechanism of decision fusion is further described in Example 6.

[0041] S5: Closed-Loop Feedback and Online Learning. This step constructs a continuously optimizing closed-loop system, enabling the control method to be adaptive and self-learning. After executing the adjustment command, the system immediately collects the adjustment effect data and compares it with the expected target. Based on this deviation, the online learning algorithm automatically updates the fitness function weights of the genetic algorithm and the rule parameters of the fuzzy control. This means that the system can learn from each control experience, continuously improve its own model, and thus maintain efficient and accurate control performance under varying operating conditions. The specific implementation of this closed-loop feedback mechanism is further detailed in Examples 7 and 8.

[0042] Example 2 This embodiment describes the specific implementation of step S1 in Embodiment 1 (acquiring transformer load current, voltage fluctuation, tap position, and ambient temperature; real-time data collection of the transformer load current, voltage fluctuation, tap position, and ambient temperature via a sensor network to establish the operating status database). Please refer to [link to previous document]. Figure 3 The process specifically includes: S1.1: Deploy high-precision intelligent power sensors on the high and low voltage sides of the transformer to synchronously collect key electrical parameters such as three-phase voltage, current, active power, reactive power, and power factor.

[0043] High-voltage side data primarily reflects the voltage level and fluctuation characteristics of the power grid, while low-voltage side data is directly related to the power supply quality at the user end. The sensor's accuracy is preferably 0.2S to ensure the accuracy of the measurement data. A sampling frequency of 1 second is set. This frequency effectively captures transient changes in the power grid caused by large load starts and stops, supporting the accurate calculation of voltage change rate (dU / dt) and load change rate (dP / dt), while avoiding data redundancy and processing pressure caused by excessively high sampling frequencies. The collected voltage data is the core basis for evaluating the voltage quality target (f1), while current and power data are the foundation for calculating the system's active power loss (f2).

[0044] S1.2: Install an absolute encoder or similar high-precision position sensor on the drive mechanism of the on-load tap changer of the transformer to achieve millisecond-level real-time monitoring of the current tap position and obtain accurate tap position data. This data has a dual key function: First, as direct state feedback of the control system, it verifies whether the adjustment command is executed correctly, forming an indispensable part of the closed-loop control, and can promptly detect mechanical faults such as tap jamming; Second, it is a direct data source for accurately calculating the target tap operation frequency (f3), providing a basis for evaluating and optimizing the switch life.

[0045] S1.3: To comprehensively understand the thermal state of the transformer, a temperature sensor array is arranged vertically inside the transformer tank to monitor the oil temperature at the top, middle, and bottom layers, thereby understanding the oil temperature distribution and promptly detecting anomalies such as localized overheating. Simultaneously, an ambient temperature sensor is installed near the transformer. The purpose of acquiring this multi-point temperature data is twofold: firstly, as an important indicator for assessing the transformer's health status, preventing insulation aging and shortened lifespan due to overheating; secondly, as a key correction parameter for network loss calculation, since winding resistance changes with temperature, real-time temperature data can be used to correct the copper loss calculation model, thereby improving the accuracy of the economic target (f2) assessment.

[0046] S1.4: The electrical parameter data, tap position data, and temperature data collected by the various sensors are encapsulated and encrypted using a wireless communication module (e.g., supporting 4G / 5G networks) integrated in the field terminal unit (RTU), and transmitted securely and with low latency to the back-end monitoring center via standard protocols such as IEC 61850 or Modbus. At the monitoring center, a dedicated operational status database is constructed, preferably using a time-series database architecture. This database undergoes rigorous preprocessing of the received data, including timestamp alignment, outlier removal, and missing value imputation, before being stored in a structured format. The database not only contains long-term historical operational data for model training and trend analysis, and second-level real-time status information for real-time control decisions, but also integrates future load forecast data generated by the short-term load forecasting module, providing forward-looking input for predictive optimization of genetic algorithms. Ultimately, this forms a comprehensive, high-quality data foundation, providing unified and reliable data services for all upper-level intelligent algorithms.

[0047] Example 3 This embodiment describes the specific implementation of step S2 in Embodiment 1 (based on historical running data and real-time status information in the running status database, using a genetic algorithm to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model, establishing the fitness evaluation function, and generating the Pareto optimal solution set). Please refer to [link to previous document]. Figure 4 The process specifically includes: S2.1: Based on the historical operation data and real-time status information, the splitter regulation strategy is encoded into chromosomes, with each gene locus representing a regulation scheme within a specific time window, and an individual coding scheme is defined.

[0048] Specifically, to achieve global optimization for a future scheduling cycle (e.g., 24 hours), this cycle is divided into N time windows of equal length (e.g., N=96, each window is 15 minutes). A complete regulation strategy, i.e., an individual, is encoded as a chromosome of length N. Each gene position i (i=1, 2, ..., N) on the chromosome corresponds to the regulation action in the i-th time window. Its gene value is selected from a predefined action set, such as {-1, 0, +1}, representing the execution of one "downgrade one level", "no action", and "upgrade one level" operation within that time window, respectively. This integer encoding method intuitively maps complex temporal control strategies into gene sequences that are easy for genetic algorithms to process, providing convenience for subsequent genetic operations such as crossover and mutation. S2.2: Based on the individual encoding scheme, construct the fitness evaluation function and establish a multi-objective optimization model. This invention aims to simultaneously optimize voltage quality, operational economy, and equipment lifespan. Therefore, an optimization model is constructed that includes the following three mutually constraining objective functions: 1) Voltage quality objective (f1): The objective is to minimize the root mean square value of the voltage deviation over the entire scheduling cycle, and its mathematical expression is:

[0049] Where Vi(gi) is the predicted bus voltage value calculated by power flow calculation after the adjustment action gi is performed in the i-th time window, and Vref is the preset reference voltage (e.g., 1.0 pu). The smaller this objective function value, the more stable the voltage control and the higher the quality. 2) Operational economic objective (f2): The objective is to minimize the system active power loss throughout the entire scheduling cycle, and its mathematical expression is:

[0050] in, The adjustment action is performed within the i-th time window. The system's active power loss can be accurately obtained through power flow calculations combined with line resistance parameters corrected for real-time temperature, where Δt is the length of the time window. The smaller the objective function value, the more economical the system operation. 3) Equipment lifespan objective (f3): The objective is to minimize the total number of tap changer operations within the entire scheduling cycle, and its mathematical expression is:

[0051] in, Let be the value of the i-th gene locus, and sgn be the sign function. This objective function directly counts the number of non-zero adjustment actions. The smaller the value, the less mechanical wear on the tap changer, which helps extend the service life of the equipment.

[0052] S2.3: Based on the multi-objective optimization model, the Pareto optimal solution set is generated using the non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy.

[0053] The process is as follows: First, an initial population P0 containing M individuals (e.g., M=100) is randomly generated. Then, an iterative evolutionary process begins: In generation t, genetic operators such as tournament selection, simulated binary crossover (SBX), and polynomial mutation (PM) are applied to population Pt to generate a progeny population Qt of the same size. The parent and progeny generations are merged into a temporary population Rt = Pt ∪ Qt of size 2M. Next, Rt is non-dominated and sorted, stratifying it into multiple non-dominated fronts F1, F2, ... . Simultaneously, the crowding degree of individuals in each front is calculated to maintain solution diversity. Finally, the optimal M individuals are selected from Rt based on non-dominated level (higher levels are preferred) and crowding degree (within the same level, higher crowding degree is preferred) to form the next generation population P(t+1). This process is repeated until the preset maximum number of generations is reached. When the algorithm terminates, all solutions in the first non-dominated front F1 constitute the desired Pareto optimal solution set. Each solution in this solution set is a non-dominated daily adjustment strategy that achieves different trade-offs among the three objectives, providing a series of high-quality candidate solutions for the subsequent intelligent decision-making module.

[0054] Example 4 This embodiment describes the specific implementation of step S3 in Embodiment 1 (based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt in the operating status database, performing fuzzification processing on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt, and formulating fuzzy inference rules for tap changer adjustment). Please refer to... Figure 5 The process specifically includes: S3.1: Establish the fuzzification interface for input variables. This step involves converting the precise, real-time physical quantities obtained from the operational status database into fuzzy linguistic variables that the fuzzy controller can understand. Specifically, this involves defining and fuzzifying three key input variables: 1) Voltage deviation ΔU: Define its universe of discourse, for example, [-0.1 pu, +0.1 pu]. Within this universe of discourse, it is divided into seven fuzzy subsets: {Negative Large (NB), Negative Medium (NM), Negative Small (NS), Zero (ZE), Positive Small (PS), Positive Medium (PM), Positive Large (PB)}. Each fuzzy subset is described by a membership function, preferably a triangular or trapezoidal membership function that is computationally simple and intuitive for engineering. For example, when a precise voltage deviation value ΔU = +0.03 pu is input, it may belong to both the "Zero (ZE)" and "Positive Small (PS)" fuzzy sets, with membership degrees μ_ZE(0.03) = 0.4 and μ_PS(0.03) = 0.6, respectively. This partial membership characteristic is the core of fuzzy logic in handling uncertainty.

[0055] 2) Voltage change rate dU / dt and load change rate dP / dt: Similarly, the universe of discourse for these two dynamic indicators is defined and divided into three fuzzy subsets: {negative (N), zero (Z), positive (P)}. This reflects the changing trends of voltage and load (decreasing, stable, increasing). Triangular or trapezoidal membership functions are also used for quantization. Through this fuzzification process, the system converts precise real-time data (e.g., ΔU=+0.03, dU / dt=-0.001, dP / dt=+0.05) into a set of fuzzy inputs with membership degrees (e.g., {ΔU is 0.6 PS and 0.4 ZE}, {dU / dt is 0.8 N and 0.2 Z}, {dP / dt is 1.0 P}), laying the foundation for subsequent fuzzy inference.

[0056] S3.2: Construct the fuzzy inference rule base and inference engine. This step is the "brain" of the fuzzy controller, containing expert knowledge and control strategies.

[0057] 1) Define the output variable: The output of the fuzzy controller is also a fuzzy linguistic variable. In this embodiment, the output variable is "Tap Change (TC)", and its universe of discourse can be defined as {-2, -1, 0, +1, +2}, which correspond to the five fuzzy subsets {down by 2 levels (DN2), down by 1 level (DN1), no action (NO), up by 1 level (UP1), up by 2 levels (UP2)} respectively.

[0058] 2) Formulate fuzzy inference rules: Based on the fuzzy sets of inputs and outputs, establish a series of fuzzy rules in the form of "IF-THEN". The rule base is designed to comprehensively cover all possible power grid operating conditions. For example: - Rule R1: IF ΔU is PB AND dU / dt is P THEN TC is DN2. (Voltage is severely high and still rising rapidly, requiring immediate and significant reduction) - Rule R2: IF ΔU is PS AND dP / dt is P THEN TC is NO. (Voltage is slightly high, but the load is increasing, which is expected to lower the voltage, so no action is taken for now) - Rule R3: IF ΔU is ZE AND dU / dt is NS THEN TC is UP1. (Voltage is normal but has a downward trend, and a slight upward adjustment should be prepared) This embodiment constructs a total of 49 core rules, mainly based on the combination of voltage deviation ΔU and voltage change rate dU / dt (7×3=21 rules), and the combination of voltage deviation ΔU and load change rate dP / dt (7×3=21 rules), and supplements key linkage rules to form a complete rule matrix.

[0059] 3) Establishing a fuzzy inference engine: A Mamdani-type fuzzy inference engine is adopted. For a given fuzzy input, the inference engine will traverse the entire rule base, calculate the trigger strength of each rule (usually by taking the "AND" operation on the membership degree of the premise, i.e., the minimum value), and perform "peak-shaving" processing on the conclusion (output fuzzy set) of the rule according to the trigger strength to obtain the fuzzy output of each rule.

[0060] S3.3: Perform defuzzification to generate precise control instructions. This step transforms the result of fuzzy inference back into a clear, executable numerical value.

[0061] After multiple rules are triggered simultaneously, the system obtains a final fuzzy output set composed of the superposition of the fuzzy outputs of all triggered rules. To obtain a precise adjustment value, defuzzification is required. This invention preferably employs the Center of Gravity (CoG) method, which is widely used due to its smoothness and robustness. The CoG method calculates the abscissa of the geometric center (centroid) of the region covered by the final output fuzzy set. Mathematically, it calculates a weighted average of the output variable's domain, weighted by membership degree. For example, if the final superimposed fuzzy output forms an irregular shape between "No Action (NO)" and "Upgrade by 1 level (UP1)," the CoG method may calculate a precise value, such as +0.7. This precise value is the final fuzzy control command. Upon receiving the "+0.7" command, the controller can determine to execute the "Upgrade by 1 level" operation based on preset execution logic (e.g., when the absolute value of the command is greater than 0.5, execute the nearest integer level adjustment), thus achieving a closed loop from fuzzy decision-making to precise control.

[0062] Example 5

[0063] This embodiment further illustrates the "formulation of 49 fuzzy inference rules" in Embodiment 4, aiming to more clearly demonstrate the construction method of fuzzy rules. These rules are preset in the knowledge base of the fuzzy controller, and some examples are as follows: Rule 1: When the voltage deviation is positive and the voltage change rate is positive, immediately adjust down by 2 levels. This rule corresponds to emergency situations where the voltage is severely high and still rising rapidly, requiring immediate and significant adjustment to suppress voltage exceeding limits.

[0064] Rule 2: When the voltage deviation is negative and the voltage change rate is negative, immediately adjust upwards by 2 levels. This rule corresponds to emergency situations where the voltage is severely low and still rapidly decreasing, requiring immediate and significant reverse adjustment.

[0065] Rule 3: When the voltage deviation is zero and the load change rate is zero, maintain the current tap position. This rule corresponds to the system operating in an ideal state, with qualified voltage and stable load. In this case, unnecessary tap changes should be avoided to extend equipment life.

[0066] In addition, other rules based on different combinations of the voltage deviation, the voltage change rate, and the load change rate are also included, for example: If the voltage deviation is "small positive" and the voltage change rate is "zero" while the load change rate is "positive", then "adjust down by 1 level later". This indicates that the voltage is slightly high but stable, while the load is trending upward (which may cause the voltage to drop), so the adjustment can be delayed or reduced for observation.

[0067] If the voltage deviation is "zero" and the voltage change rate is "negative", then "prepare to adjust upward by 1 level". This indicates that the voltage is normal but has a downward trend, and the system should enter a warning state, taking immediate action once a deviation occurs.

[0068] By establishing 49 fuzzy inference rules covering all major operating conditions, the fuzzy controller can make reasonable and rapid responses in various scenarios.

[0069] Example 6 This embodiment describes the specific implementation of step S4 in Embodiment 1 (using the Pareto optimal solution set as a candidate solution library through an intelligent decision-making algorithm, using the fuzzy control command as a real-time constraint, and calculating the comprehensive score of each candidate solution using a weighted comprehensive evaluation method). Please refer to [link to previous document]. Figure 6 The process specifically includes: S4.1: Based on the Pareto optimal solution set and the fuzzy control instructions, a decision fusion processor is designed, using the Pareto optimal solution set as a candidate solution library and the fuzzy control instructions as real-time constraints. The decision fusion processor, as the core algorithm module connecting global optimization and real-time response, is specifically implemented as follows: First, the fusion processor receives two key inputs. The first input is the Pareto optimal solution set {S1, S2, ..., Sk} generated by the genetic algorithm, where each solution Sj is a complete sequence of tap changer strategies covering a future scheduling cycle (e.g., a vector consisting of 96 time window changers {-1, 0, +1}), representing a globally optimal trade-off. The second input is a precise and continuous control command C_fuzzy (e.g., a real value between -2 and +2, such as +0.7) generated by the fuzzy controller for the current grid state (ΔU, dU / dt, dP / dt) at the i-th time window. This command represents the system's most pressing regulation requirement and ideal regulation magnitude at the current moment. The task of the decision fusion processor is to select the optimal regulation action for the current i-th time window from the candidate solution library. Instead of directly executing C_fuzzy, it uses C_fuzzy as a "real-time demand benchmark" to evaluate the suitability of the action g_j,i suggested by each solution Sj in the Pareto solution set in the current time window i, thereby achieving intelligent integration of global optimization goals and real-time adjustment needs.

[0070] S4.2: Based on the aforementioned decision fusion unit, a dynamic allocation mechanism for decision weights is established. Under normal operating conditions, the Pareto optimal solution set takes precedence, while under emergency conditions, the fuzzy control commands take priority. To achieve this dynamic allocation, this invention introduces a "real-time response weight" α, whose value dynamically changes within the interval [0, 1]. The value of α is determined by a state evaluation function, which calculates based on the severity of voltage deviation ΔU and voltage change rate dU / dt. Specifically, an urgency index E = k1*|ΔU| + k2*|dU / dt| is defined, where k1 and k2 are weighting coefficients. Then, the urgency index E is mapped to the weight α using a Sigmoid function: α = 1 / (1 + exp(-c*(E - E_threshold))). Here, c is the steepness coefficient of the function, and E_threshold is a preset urgency threshold. When the power grid is operating smoothly (E is much smaller than E_threshold), the value of α is close to 0, and the decision will primarily be based on the overall global performance of the scheme. When the power grid condition becomes abnormal (E approaches or exceeds E_threshold), the value of α will increase rapidly and approach 1. At this point, the decision will overwhelmingly prioritize the degree of conformity between the scheme and the fuzzy control command, ensuring that the system can make the fastest and most direct response to deal with potential risks.

[0071] S4.3: Based on the aforementioned dynamic allocation mechanism of decision weights, the comprehensive score of each candidate scheme is calculated using a weighted comprehensive evaluation method, and the scheme with the highest score is selected to generate the tap changer adjustment command. Specifically, for each candidate scheme Sj in the Pareto optimal solution set, its comprehensive score Score(Sj) is composed of two weighted parts: global performance score and real-time compliance score.

[0072] 1) Global Performance Score (Score_global): This score evaluates the long-term comprehensive benefits of scheme Sj. First, the three objective function values ​​f1(Sj), f2(Sj), and f3(Sj) corresponding to scheme Sj are normalized. Then, a weighted sum is calculated based on preset operating preferences (reflected by weights wg1, wg2, and wg3). Score_global(Sj) = wg1*norm(f1) + wg2*norm(f2) + wg3*norm(f3).

[0073] 2) Real-time compliance score (Score_realtime): This score evaluates the consistency between the adjustment action g_j,i suggested by the scheme Sj in the current time window and the fuzzy control instruction C_fuzzy. It is calculated using a Gaussian similarity function. Score_realtime(Sj) = exp( - (gj,i - C_fuzzy)2 / (2σ2) ) Where σ is the variance parameter, used to adjust the tolerance for deviation. This score is 1 when g_j,i is completely consistent with C_fuzzy; the larger the deviation, the closer the score is to 0. Finally, the comprehensive score of each candidate solution Sj is: Score(Sj) = (1 - α) * Score_global(Sj) + α * Score_realtime(Sj). The decision fusion unit traverses all candidate solutions, calculates their respective comprehensive scores, and selects the solution with the highest score, S_best. Finally, the adjustment action g_best,i corresponding to the current time window in S_best is output as the final tap changer adjustment command, driving the controller to execute. This mechanism ensures that the final decision inherits the global optimization characteristics of the Pareto solution while fully incorporating the real-time adaptability of fuzzy control.

[0074] Example 7 This embodiment describes the specific implementation of the execution and preliminary feedback stage (driving the tap changer controller to perform the gear adjustment action and collecting the adjusted voltage and current parameters) in step S5 of embodiment 1. Please refer to [link to embodiment 1]. Figure 6 The process specifically includes: S5.1: Based on the tap changer adjustment command, the tap changer controller is driven to perform a tap position adjustment action. Immediately after each tap position adjustment action, the adjusted voltage and current parameters are collected, and the deviation between the actual adjustment effect and the expected target is calculated to obtain adjustment effect evaluation data. This step is the execution end of the control loop. The adjustment command is sent to the field controller via reliable communication methods such as industrial Ethernet, driving the tap changer to complete the tap position switching. Once the adjustment action is completed, the system immediately collects the power grid status after the action through sensors and compares it with the expected status at the time the command was issued, quantifying the accuracy of this control and forming evaluation data.

[0075] S5.2: Based on the aforementioned regulation effect evaluation data, establish a performance evaluation index system including voltage qualification rate, regulation accuracy, and response time. To systematically evaluate the long-term performance of the control method, a set of key performance indicators (KPIs) needs to be established. Voltage qualification rate measures voltage quality, regulation accuracy assesses the accuracy of control, and response time measures the system's speed. These indicators will be continuously recorded and analyzed.

[0076] S5.3: Based on the aforementioned performance evaluation index system, when multiple consecutive adjustments fail to achieve the desired effect, a parameter retuning procedure is automatically initiated to obtain optimized control parameters. This is a self-diagnostic and self-repair mechanism. If the system detects a continuous deterioration in the KPI indicators (e.g., the voltage deviation increases after five consecutive adjustments), it determines that the current control model may no longer be adapted to changes in the power grid state, thereby automatically triggering the parameter retuning procedure to comprehensively recalibrate the core parameters of the upper-level algorithm.

[0077] Example 8

[0078] This embodiment describes the specific implementation of the online learning step in step S5 of Embodiment 1 (updating the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain the optimized control parameters). Please refer to [link to previous document]. Figure 6 The process specifically includes: S5.4: Based on the aforementioned adjustment effect evaluation data and the aforementioned performance evaluation index system, the weight coefficients of w1, w2, and w3 in the fitness evaluation function are updated using the gradient descent algorithm. This process aims to optimize the target preference of the genetic algorithm. If historical data shows that although the voltage qualification rate is high, the network loss and action frequency are too high, the online learning algorithm will automatically fine-tune the weights of w1, w2, and w3 using the gradient descent method according to the preset optimization objective (such as "minimizing network loss and action frequency while ensuring a 99.9% voltage qualification rate"). For example, it may appropriately reduce the weight of w1 and increase the weights of w2 and w3.

[0079] S5.5: Based on the deviation between the adjustment effect and the expected target, an adaptive adjustment algorithm is used to update the membership function parameters of each linguistic variable in the fuzzy inference rule. This process aims to optimize the performance of the fuzzy controller. For example, if the system is found to be slow to respond to the case of "small voltage deviation", the adaptive algorithm will fine-tune the membership function of the linguistic variable "small" to make its boundaries more sensitive, thereby allowing the controller to identify and respond to such cases earlier.

[0080] S5.6: Based on the weighting coefficients and membership function parameters, an anomaly detection module is established. When a device malfunction or abnormal operating condition is detected, it automatically switches to a safe mode and issues an alarm signal, obtaining the optimized control parameters. This is an additional module to ensure system safety. It continuously monitors power grid data using the learned normal operating parameter model. Once a significant deviation of the data pattern from the normal model is detected (which may indicate sensor failure, tap jamming, or a major disturbance in the external power grid), the system immediately switches to a preset, most conservative safe operating mode (e.g., stopping all automatic adjustments) and issues an alarm signal to maintenance personnel.

[0081] Example 9 This embodiment provides an intelligent optimization control system for on-load tap changing of transformers, which serves as the physical carrier for implementing the aforementioned methods. For example... Figure 2 As shown, the system includes: The data acquisition module 10 is used to acquire transformer load current, voltage fluctuation, tap position, and ambient temperature. It collects these data in real time via a sensor network to establish an operational status database. This module corresponds to the function described in Embodiment 2 and serves as the system's perception layer, providing data input for the entire system.

[0082] The genetic algorithm optimization module 20 is used to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model based on historical running data and real-time status information in the running status database, establish a fitness evaluation function, and generate a Pareto optimal solution set. This module is the system's global optimization engine, corresponding to the function described in Embodiment 3, and is responsible for generating long-term, globally optimal control strategy candidates.

[0083] The fuzzy control module 30 is used to perform fuzzification processing on the voltage deviation, voltage change rate, and load change rate based on the operating status database, formulate fuzzy inference rules for tap adjustment, and output fuzzy control commands. This module is the system's real-time response engine, corresponding to the functions described in embodiments 4 and 5, and is responsible for quickly generating immediate adjustment suggestions based on the current operating conditions.

[0084] The intelligent decision-making module 40 is used to generate tap changer adjustment instructions based on the Pareto optimal solution set and the fuzzy control instructions. This is achieved by using an intelligent decision-making algorithm to calculate the comprehensive score of each candidate solution using the Pareto optimal solution set as a candidate solution library and the fuzzy control instructions as real-time constraints. This module is the core of the system's decision-making process, corresponding to the function described in Embodiment 6, responsible for integrating global optimization results and real-time response suggestions to make the final control decision.

[0085] The closed-loop feedback module 50 is used to drive the tap changer controller to perform a tap adjustment action based on the tap adjustment command, and to collect the adjusted voltage and current parameters. It then updates the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain optimized control parameters. This module is the system's execution and learning layer, corresponding to the functions described in embodiments 7 and 8. It is responsible for executing commands, evaluating effects, and continuously optimizing system performance, thus forming a complete intelligent control closed loop.

[0086] The system modules described above can be integrated into a central controller or deployed in a distributed manner on edge computing devices and cloud servers.

[0087] This disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of the intelligent optimization control method for on-load tap changing of a transformer as described in the above-described method embodiments. The storage medium can be either volatile or non-volatile computer-readable storage.

[0088] Furthermore, this disclosure also provides a computer program product storing a computer program. When the computer program is run by a processor, it executes the steps of an intelligent optimization control method for on-load tap change of a transformer provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0089] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0090] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and apparatuses described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this disclosure, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0091] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0092] In addition, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0093] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0094] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A smart optimization control method for on-load tap changing of a transformer, characterized in that, Includes the following steps: The transformer load current, voltage fluctuation, tap position, and ambient temperature are acquired. The transformer load current, voltage fluctuation, tap position, and ambient temperature are collected in real time through a sensor network to establish an operating status database. The sensor network includes intelligent power sensors installed on the high and low voltage sides of the transformer, position sensors installed on the tap switch device, and a temperature sensor array arranged in the transformer tank. Based on the historical operating data and real-time status information in the operating status database, a genetic algorithm is used to encode the tap changer adjustment strategy into chromosomes and construct a multi-objective optimization model. With the optimization objectives of minimizing voltage quality deviation, reducing network loss, and controlling tap changer operation frequency, a fitness evaluation function F=w1×f1+w2×f2+w3×f3 is established to generate the Pareto optimal solution set, where f1 calculates the root mean square value of voltage deviation, f2 calculates the system active power loss, f3 counts the number of tap changer switching operations, and w1-s3 are user-defined weights. Based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt in the operating status database, the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt are fuzzified, and fuzzy inference rules for tap adjustment are formulated. The optimal adjustment timing and adjustment range are determined through fuzzy inference and defuzzification, and fuzzy control commands are output. The Pareto optimal solution set is used as a candidate solution library by intelligent decision-making algorithm, the fuzzy control command is used as a real-time constraint, and the comprehensive score of each candidate solution is calculated by weighted comprehensive evaluation method. The solution with the highest score is selected to generate the tap changer adjustment command. Using the tap adjustment command, the tap switch controller is driven to perform a gear adjustment action and collect the adjusted voltage and current parameters. The weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule are updated through an online learning algorithm to obtain optimized control parameters.

2. The method according to claim 1, characterized in that, The system acquires transformer load current, voltage fluctuation, tap position, and ambient temperature. It uses a sensor network to collect these data in real time, establishing an operational status database, including: Based on the transformer load current, voltage fluctuation, tap position and ambient temperature, intelligent power sensors are installed on the high and low voltage sides of the transformer to collect three-phase voltage, current and power factor and the sampling frequency is set to 1 second / time to obtain electrical parameter data. A position sensor is installed on the tap changer device to monitor the current tap position in real time and obtain tap position data. A temperature sensor array is arranged inside the transformer oil tank to monitor the oil temperature distribution and acquire ambient temperature data. Based on the electrical parameter data, the tap position data, and the ambient temperature data, the electrical parameter data, the tap position data, and the ambient temperature data are transmitted to the monitoring center via a wireless communication module to establish an operating status database containing historical operating data, real-time status information, and predictive data.

3. The method according to claim 1 or 2, characterized in that, Based on historical operational data and real-time status information in the operational status database, a genetic algorithm is used to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model. The fitness evaluation function is established, and the Pareto optimal solution set is generated, including: Based on the historical operating data and real-time status information, the splitter regulation strategy is encoded into chromosomes, with each gene locus representing a regulation scheme within a specific time window, and an individual coding scheme is defined. Based on the individual coding scheme, the fitness evaluation function F=w1×f1+w2×f2+w3×f3 is constructed, where f1 calculates the root mean square value of voltage deviation, f2 calculates the system active power loss, and f3 counts the number of tap changer actions, thus establishing a multi-objective optimization model. Based on the multi-objective optimization model, the population size is set to 100, the crossover probability is 0.8, and the mutation probability is 0.

05. The Pareto optimal solution set is generated through selection, crossover, and mutation operations.

4. The method according to claim 3, characterized in that, Based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt in the operating status database, the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt are fuzzified, and fuzzy inference rules for tap changer adjustment are formulated, including: Based on the voltage deviation ΔU, voltage change rate dU / dt, and load change rate dP / dt, the voltage deviation is divided into seven linguistic variables: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large. The voltage change rate and the load change rate are each divided into three linguistic variables: negative, zero, and positive. A membership function for the input variables is established. Based on the membership function, a fuzzy controller with the number of adjustment levels and adjustment time as output variables is established, and 49 fuzzy inference rules are formulated. Based on the 49 fuzzy inference rules and the membership function, the centroid method is used for defuzzification processing, and the fuzzy control command is output.

5. The method according to claim 4, characterized in that, The 49 rules for fuzzy inference include: When the voltage deviation is positive and the voltage change rate is positive, immediately adjust the voltage down by 2 levels; When the voltage deviation is negative and the voltage change rate is negative, immediately adjust the voltage up by 2 levels; When the voltage deviation is zero and the load change rate is zero, the current gear level remains unchanged; And based on different combinations of the voltage deviation, the voltage change rate, and the load change rate, 49 fuzzy inference rules covering all operating conditions are established.

6. The method according to claim 1, characterized in that, Based on the Pareto optimal solution set and the fuzzy control command, an intelligent decision-making algorithm is used to calculate the comprehensive score of each candidate solution by using the Pareto optimal solution set as a candidate solution library and the fuzzy control command as a real-time constraint. This includes: Based on the Pareto optimal solution set and the fuzzy control instructions, a decision fusion device is designed, using the Pareto optimal solution set as a candidate solution library and the fuzzy control instructions as real-time constraints. Based on the decision fusion device, a dynamic decision weight allocation mechanism is established, which prioritizes the Pareto optimal solution set under normal operating conditions and the fuzzy control command under emergency conditions. Based on the dynamic allocation mechanism of decision weights, the comprehensive score of each candidate scheme is calculated by a weighted comprehensive evaluation method, and the scheme with the highest score is selected to generate the tap changer adjustment command.

7. The method according to claim 6, characterized in that, The tap changer controller is driven to perform tap adjustment and collect the adjusted voltage and current parameters, including: Based on the tap adjustment command, the tap switch controller is driven to perform a gear adjustment action. After each gear adjustment action is completed, the adjusted voltage and current parameters are collected immediately, the deviation between the actual adjustment effect and the expected target is calculated, and adjustment effect evaluation data is obtained. Based on the aforementioned regulation effect evaluation data, a performance evaluation index system including voltage qualification rate, regulation accuracy, and response time is established. Based on the aforementioned performance evaluation index system, when repeated adjustments fail to achieve the desired effect, the parameter retuning program is automatically initiated to obtain optimized control parameters.

8. The method according to claim 1, characterized in that, The optimized control parameters are obtained by updating the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm, including: Based on the adjustment effect evaluation data and the performance evaluation index system, the weight coefficients of w1, w2, and w3 in the fitness evaluation function are updated using the gradient descent algorithm. Based on the deviation between the adjustment effect and the expected goal, an adaptive adjustment algorithm is used to update the membership function parameters of each linguistic variable in the fuzzy inference rule; Based on the weighting coefficients and the membership function parameters, an anomaly detection module is established. When a device malfunction or abnormal operating condition is detected, the module automatically switches to a safe mode and issues an alarm signal, thereby obtaining the optimized control parameters.

9. An intelligent optimization control system for on-load tap changing of a transformer, characterized in that, include: The data acquisition module is used to acquire transformer load current, voltage fluctuation, tap position and ambient temperature. It collects the transformer load current, voltage fluctuation, tap position and ambient temperature in real time through a sensor network and establishes an operating status database. The genetic algorithm optimization module is used to encode the splitter regulation strategy into chromosomes and construct a multi-objective optimization model based on the historical running data and real-time status information in the running status database, establish a fitness evaluation function, and generate a Pareto optimal solution set. The fuzzy control module is used to perform fuzzification processing on the voltage deviation, voltage change rate and load change rate based on the voltage deviation, voltage change rate and load change rate in the operating status database, formulate fuzzy inference rules for tap adjustment, and output fuzzy control commands. The intelligent decision-making module is used to generate tap changer adjustment instructions based on the Pareto optimal solution set and the fuzzy control instructions, using an intelligent decision-making algorithm to use the Pareto optimal solution set as a candidate solution library, the fuzzy control instructions as real-time constraints, and a weighted comprehensive evaluation method to calculate the comprehensive score of each candidate solution. The closed-loop feedback module is used to drive the tap changer controller to perform a gear adjustment action based on the tap adjustment command and to collect the adjusted voltage and current parameters. It updates the weight coefficients of the fitness evaluation function and the membership function of the fuzzy inference rule through an online learning algorithm to obtain optimized control parameters.

Images