JKW control method based on adaptive fuzzy control

Abstract

The invention provides a JKW control method based on adaptive fuzzy control, which belongs to the technical field of reactive power compensation of a power system, and comprises the following steps of: interacting reactive power compensation states of JKW devices in adjacent regions based on an edge physical agent, constructing a regional reactive power topological graph in combination with local power grid topological data, generating a regional coordination instruction according to a supply-demand relationship, and establishing a regional coordination graph according to the regional coordination instruction; the method comprises the following steps: acquiring proportion deviation, power factor deviation and reactive power change gradient of a mutual inductor, constructing a standard fuzzy input vector in combination with a regional cooperation instruction, mapping the standard fuzzy input vector into a fuzzy linguistic variable through a dynamically updated fuzzy rule base, and reasoning and outputting an initial switching strategy; and comparing and analyzing the proportion deviation and the power factor deviation through a hardware comparator, determining a trigger condition and existence of a fault, and dynamically optimizing an initial switching strategy by combining a regional reactive topological graph to realize the control of the JKW. And the operation stability of the power grid is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of reactive power compensation technology in power systems, and in particular to a JKW control method based on adaptive fuzzy control. Background Technology

[0002] In power system operation, reactive power balance and dynamic compensation are crucial for ensuring grid voltage stability and reducing line losses. JKW units, as core equipment for dynamic reactive power regulation, directly impact the economy and reliability of the distribution network through their control accuracy and coordination capabilities. However, existing JKW control technology suffers from the following significant drawbacks:

[0003] Traditional JKW control often employs a local independent regulation mode, generating switching strategies solely based on local grid parameters (such as voltage and power factor), without considering the reactive power compensation status correlation between adjacent JKW units. For example, in an industrial park, multiple JKW units often exhibit alternating overcompensation and undercompensation due to a lack of cross-regional coordination, resulting in voltage fluctuations exceeding ±5% and increased line transmission losses by 15%-20%. Existing fuzzy control methods primarily use preset fixed value rule bases, unable to dynamically update based on real-time operating conditions. For instance, when grid load changes abruptly (such as the start-up of a large motor), the switching strategies generated by the fixed rule base lag behind actual needs, causing power factor correction times to exceed 10 seconds, deviating from the target value of 0.95 by ±0.08. Existing fault detection methods often rely on single sensor parameter threshold judgments (such as detecting faults solely through voltage deviation), lacking multi-dimensional verification; fault code design is simple (such as only including fault type), failing to achieve precise fault location and severity. For example, in one substation, a misjudgment of transformer deviation led to erroneous switching of JKW units, with troubleshooting taking over 2 hours. All of the above factors lead to a decrease in the adaptability, coordination, and reliability of JKW control.

[0004] Therefore, this invention proposes a JKW control method based on adaptive fuzzy control. Summary of the Invention

[0005] This invention provides a JKW control method based on adaptive fuzzy control to solve the aforementioned technical problems.

[0006] This invention provides a JKW control method based on adaptive fuzzy control, comprising:

[0007] Step 1: Based on the reactive power compensation status of JKW devices in adjacent areas through edge physical agent interaction, construct a regional reactive power topology map by combining local power grid topology data, and generate regional coordination instructions according to supply and demand relationship, wherein the supply and demand relationship is determined by the ratio of total reactive power deficit to total compensation capacity in the region.

[0008] Step 2: Collect the proportional deviation, power factor deviation, and reactive power change gradient of the current transformer, and construct a standard fuzzy input vector in combination with the regional collaborative instructions. Map the standard fuzzy input vector into fuzzy language variables through a dynamically updated fuzzy rule base, and infer and output the initial switching strategy. The initial switching strategy is related to the capacitor switching amount, thyristor trigger phase, and collaborative enable flag.

[0009] Step 3: Compare and analyze the proportional deviation and power factor deviation using a hardware comparator to determine the triggering conditions and the existence of faults;

[0010] Step 4: Based on the triggering conditions and the existence of faults, and combined with the regional reactive power topology map, dynamically optimize the initial switching strategy to achieve control of JKW.

[0011] Preferably, step 1 includes:

[0012] Real-time monitoring of the reactive power change rate of the local power grid, and matching it with the change rate-cycle reference table to obtain the interaction cycle;

[0013] Based on the interaction cycle, the reactive power compensation status of the JKW device in the adjacent area and the compensation capacity and power factor of the local JKW device are collected to obtain interaction data, and the interaction data is encrypted to construct a cross-domain reactive power status dataset.

[0014] Based on the cross-domain reactive power status dataset and combined with the node connection relationship, line impedance parameters and equipment priority of the local power grid, a regional reactive power topology map is dynamically constructed. The node characteristics of the regional reactive power topology map are related to reactive power deficit, equipment health and voltage deviation, and the edge characteristics of the regional reactive power topology map are related to the calculated reactive power transmission loss and transmission delay between nodes.

[0015] With the optimization goals of reactive power balance, minimum transmission loss, and balanced equipment lifespan, and in conjunction with the supply and demand relationship of the regional reactive power topology map, regional coordination instructions are generated. These regional coordination instructions are related to the regional status symbol, the compensation amount of the JKW device, and the instruction priority.

[0016] Preferably, encrypting the interactive data includes:

[0017] Obtain the node property association and physical distance between local nodes and any neighboring nodes in the reactive power topology map of the region, and determine the potential association granularity;

[0018] Task category matching is performed on the first task set of the local node and the second task set of any neighboring node, and the potential transmission granularity is obtained by combining the network transmission paths of the local node and any neighboring node.

[0019] Based on the potential association granularity and the potential transmission granularity, a fuzzy relationship between the local node and any of the neighboring nodes is constructed to obtain the corresponding data granularity;

[0020] The interactive data is divided into real-time sensitive slices, device feature slices, and topology-related slices according to the data granularity, and an initial security array is generated based on the slice attributes of different slices.

[0021] Obtain the working log of each node in the reactive power topology map of the region, and construct the security dimension array of the corresponding node;

[0022] A first-stage encryption program matching the initial security array is loaded from the secure storage medium, and a second-stage encryption program matching the security dimension array is loaded from the secure storage medium simultaneously.

[0023] The first-stage encryption procedure is executed to obtain the storage offset parameter of each second-stage encryption procedure. The storage offset parameter is the offset of the starting address of the corresponding second-stage encryption procedure relative to the starting address of the first-stage encryption procedure. It is constrained by the minimum erasable unit size of the real-time sensitive processing unit, device feature processing unit, and topology association unit compatible with the encryption procedure, the communication time slot period, and the transmission delay tolerance and phase difference tolerance associated with the distance deviation in the regional reactive topology.

[0024] Based on the interaction direction between neighboring nodes in adjacent regions and local nodes in the local power grid, and combined with the corresponding offset, a new dimension array is obtained by performing a security upgrade analysis on the corresponding security dimension array in the same dimension.

[0025] A secure sequence is obtained based on the initial secure array and all new dimension arrays, and the interactive data is encrypted.

[0026] Preferably, step 2 includes:

[0027] Constructing a four-dimensional fuzzy input vector The dimensions are then standardized and mapped to the interval between -1 and 1 to obtain the standard fuzzy input vector X0, where ΔK is the proportional deviation of the mutual inductor and Δcosφ is the power factor deviation. The reactive power gradient, ΔQ c For regional coordination commands;

[0028] The standard fuzzy input vector is mapped to fuzzy language variables based on a dynamically updated fuzzy rule base, and the initial switching strategy is inferred and output.

[0029] Among them, capacitor switching amount Where m1 represents the number of rules in the dynamically updated fuzzy rule base; w iLet represent the weight of the i-th rule; fi(X0) represents the output function of the i-th rule;

[0030] The thyristor trigger phase θ0 is the optimal trigger angle dynamically calculated based on the voltage zero-crossing point;

[0031] The collaboration enable flag is activated when the set standard conditions are met.

[0032] Preferably, step 3 includes:

[0033] The proportional deviation and power factor deviation of the current transformer are connected to a dual-channel high-speed hardware comparator, and a threshold comparison mechanism is configured for each channel to determine the decision result. The dual-channel high-speed hardware comparator has a built-in temperature compensation circuit, and the threshold dynamic adjustment logic is linked to the local power grid frequency.

[0034] A fault code is generated based on the judgment result, and the fault code is used to locate the fault.

[0035] Preferably, generating a fault code based on the judgment result and using the fault code for fault location includes:

[0036] The judgment results, the cumulative number of anomalies in the timing logic judgments, and the offset of real-time monitored equipment operating parameters are aligned to construct a fault feature vector;

[0037] The fault feature vector is matched with the pre-constructed 16-bit fault code, and the fault category is locked according to the TypeCode and SeverityCode. The fault propagation path is calculated by combining the LocationCode, TimeStamp and the shortest path algorithm of the topology graph to determine the most likely fault source.

[0038] A fault cause-effect graph is constructed. When a compound fault is detected, the main fault source is determined by Bayesian network inference. The node weights of the fault cause-effect graph are taken from the equipment health of the regional reactive power topology graph, and the fault propagation probability of the edges is obtained by statistical analysis of historical fault data from the past m0 months.

[0039] Preferably, step 4 includes:

[0040] Convert the trigger signal output by the hardware comparator into a trigger strength coefficient Kt;

[0041] The fault is encoded using 3-bit binary code to generate the fault impact weight Wf.

[0042] Extract the node correlation degree R from the reactive power topology map of the region. i1j1 And the reactive power transmission loss coefficient γ=0.01+0.05L·(1+δ1·(T(t)-T0)+δ2·μ(t)), where Ri1j1 δ1 represents the node correlation degree between local node i1 and neighboring node j1; L represents the line length; T(t) represents the real-time ambient temperature; T0 represents the reference temperature; δ1 represents the temperature influence coefficient of 0.002 / ℃; μ(t) represents the real-time line load rate; δ2 represents the load rate influence coefficient.

[0043] The initial switching strategy is dynamically optimized:

[0044] Under normal operating conditions, the following optimizations are performed:

[0045]

[0046] in, The optimized capacitor switching amount; m2 represents the total number of neighboring nodes; ΔQ j1 The reactive power deficit of the j1-th neighboring node; is the health decay coefficient of the j1th neighboring node; For time period correction factors;

[0047]

[0048] in, The optimized thyristor trigger phase;

[0049] When the operating condition is faulty, the following optimizations are performed:

[0050] If an abnormality is detected in the current transformer, the switching amount will be recalculated based on the data from the redundant current transformers.

[0051] If a system-level fault is triggered, the collaborative enable flag is forcibly activated: B0=1, and the preset percentage of switching amount is allocated to the neighboring nodes according to the regional reactive power topology map, where B0 is the result of the collaborative enable flag;

[0052] If the reactive power topology map of the region fails due to communication interruption, a preset neighborhood compensation protocol is activated, and the topology recovery status is monitored in real time. Once the topology is restored, the system switches to dynamic optimization mode.

[0053] Preferred options also include:

[0054] Collect the optimized power factor and neighborhood compensation response, and calculate the optimization effect evaluation coefficient E;

[0055]

[0056] Wherein, cosφ new For the optimized power factor, ΔQ sy The optimized compensation response; ΔQ raThe rated reactive power is Nc; the switching frequency is τ1, τ2, and τ3 are weighting coefficients, and τ1+τ2+τ3=1;

[0057] When the optimization effect evaluation coefficient E is greater than the preset evaluation coefficient, the fuzzy rule base is dynamically updated.

[0058] Compared with the prior art, the beneficial effects of this application are as follows:

[0059] By using edge physical agents to achieve cross-regional collaboration, dynamic fuzzy rule bases to improve policy adaptability, hardware comparators to enhance fault detection, and dynamic optimization to ensure control accuracy, this approach effectively solves the problems of insufficient regional collaboration, policy lag, and slow fault response in traditional JKW control. It shortens reactive power compensation response time, stabilizes power factor, reduces line transmission loss, and significantly improves the stability of power grid operation.

[0060] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0061] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0062] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0063] Figure 1 This is a flowchart of a JKW control method based on adaptive fuzzy control in an embodiment of the present invention. Detailed Implementation

[0064] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0065] This invention provides a JKW control method based on adaptive fuzzy control, such as... Figure 1 As shown, it includes:

[0066] Step 1: Based on the reactive power compensation status of JKW devices in adjacent areas through edge physical agent interaction, construct a regional reactive power topology map by combining local power grid topology data, and generate regional coordination instructions according to supply and demand relationship. The supply and demand relationship is determined by the ratio of the vertical reactive power deficit to the total compensation capacity in the region.

[0067] Step 2: Collect the proportional deviation, power factor deviation, and reactive power change gradient of the current transformer, and construct a standard fuzzy input vector in combination with the regional collaborative instructions. Map the standard fuzzy input vector into fuzzy language variables through a dynamically updated fuzzy rule base, and infer and output the initial switching strategy. The initial switching strategy is related to the capacitor switching amount, thyristor trigger phase, and collaborative enable flag.

[0068] Step 3: Compare and analyze the proportional deviation and power factor deviation using a hardware comparator to determine the triggering conditions and the existence of faults;

[0069] Step 4: Based on the triggering conditions and the existence of faults, and combined with the regional reactive power topology map, dynamically optimize the initial switching strategy to achieve control of JKW.

[0070] In this embodiment, the edge physical agent is a physical agent module deployed on the edge computing node, which is responsible for the interaction, preprocessing and instruction forwarding of reactive power compensation status data between the local and neighboring JKW devices, and has low-latency data processing capabilities.

[0071] In this embodiment, the JKW device is an intelligent reactive power compensation device. By controlling the switching of capacitor banks and the triggering of thyristors, it realizes dynamic adjustment of the reactive power of the power grid, ensuring voltage stability and power factor compliance.

[0072] In this embodiment, the reactive power compensation status is the current set of operating parameters of the JKW device, including real-time compensation capacity, switching status (on / off), power factor, and equipment health (such as capacitor aging degree).

[0073] Local power grid topology data is the basic data describing the structure of the local power grid, including node connection relationships (such as the connection methods of transformers, loads, and switches), line parameters (conductor type, length, impedance), equipment attributes (rated capacity, priority), etc.

[0074] The regional reactive power topology map is a visualization map constructed based on local power grid topology data and the reactive power compensation status of neighboring areas. Nodes are labeled with features such as reactive power deficit, equipment health, and voltage deviation, while edges are labeled with features such as reactive power transmission loss and transmission delay between nodes. For example, node A (transformer) is labeled with "reactive power deficit -80kvar (overcompensated), health 0.95", and node B (load) is labeled with "reactive power deficit 120kvar (undercompensated), health 0.88"; the AB edge is labeled with "transmission loss 3%, transmission delay 40ms".

[0075] In this embodiment, a supply-demand ratio >1 indicates supply is less than demand, and <1 indicates supply is greater than demand.

[0076] In this embodiment, the total reactive power deficit in the region is the sum of the differences between the reactive power demand of all load nodes in the region and the current compensation amount (positive value is undercompensation, negative value is overcompensation).

[0077] Total compensation capacity is the sum of the rated compensation capacities of all JKW devices in the region, reflecting the region's maximum reactive power compensation capability.

[0078] Regional coordination instructions are cross-regional coordination instructions generated based on supply and demand relationships. They include regional status symbols (such as "under-compensation" or "over-compensation"), compensation adjustment values ​​for target JKW devices, and instruction priorities (levels 1-3, with level 3 being the highest). For example, for a supply and demand relationship where "supply is less than demand", the regional coordination instruction is generated as follows: "Status symbol: under-compensation; Target devices: JKW-1#, JKW-2#; Compensation adjustment: increase by 100kvar for each; Priority: level 2".

[0079] In this embodiment, the proportional deviation of the current / voltage transformer is the proportional error between the measured value and the actual value of the current / voltage transformer, reflecting the measurement accuracy of the transformer (calculation formula: (measured value - actual value) / actual value × 100%). The power factor deviation is the difference between the actual power factor of the power grid and the target power factor (usually 0.98), reflecting the degree of deviation of reactive power compensation (positive value is leading, negative value is lagging).

[0080] The reactive power gradient is the second derivative of reactive power over time, reflecting the rate of change of reactive power (positive values ​​indicate acceleration and negative values ​​indicate deceleration).

[0081] The standard fuzzy input vector is a four-dimensional vector formed by normalizing input parameters such as proportional deviation, power factor deviation, reactive power change gradient, and regional coordination commands (mapping them to the interval -1 to 1). It is used for fuzzy rule base inference. For example, proportional deviation -4% (normalized to -0.3), power factor deviation -0.03 (normalized to -0.2), and reactive power change gradient 1 kvar / s 2 (Normalized to 0.4), regional coordination instruction adjustment coefficient 0.2 (normalized to 0.2), forming a standard fuzzy input vector X0 = {-0.3, -0.2, 0.4, 0.2}.

[0082] The dynamically updated fuzzy rule base is a collection of multiple fuzzy control rules. For example, the rule base contains 80 rules, such as "if the proportional deviation is negative and the power factor deviation is negative, then the capacitor switching amount is positive and small". The rule weights are adjusted every hour according to the compensation effect evaluation coefficient (e.g., if a rule has a good optimization effect, its weight increases from 0.12 to 0.15).

[0083] Fuzzy linguistic variables are variables described using fuzzy terms (such as "positive large", "negative medium", and "zero") to transform precise input parameters into a form that can be processed by fuzzy logic. For example, seven fuzzy linguistic variables are defined for power factor deviation: negative large (NB), negative medium (NM), negative small (NS), zero (ZO), positive small (PS), positive medium (PM), and positive large (PB), where "-0.03" corresponds to "negative small (NS)".

[0084] The initial switching strategy is a preliminary control scheme generated based on fuzzy rule base reasoning. It includes three core parameters: capacitor switching amount, thyristor trigger phase, and cooperative enable flag. For example, the initial switching strategy is "capacitor switching amount + 120kvar (3 groups of 40kvar capacitors are connected), thyristor trigger phase 65°, cooperative enable flag 0 (cooperative is not activated)".

[0085] The hardware comparator is an electronic comparator with high-speed response capability (this solution uses dual channels) used to compare the proportional deviation and power factor deviation with preset thresholds and output a decision result (normal / abnormal). It has a built-in temperature compensation circuit to offset the influence of ambient temperature.

[0086] In this embodiment, the triggering condition is the condition that triggers subsequent strategy optimization when the hardware comparator determines that the input parameters are abnormal. It is usually "proportional deviation exceeds the threshold" or "power factor deviation exceeds the threshold". For example, the triggering condition is set as "absolute value of proportional deviation > 5% for 100ms" or "absolute value of power factor deviation > 0.05 for 200ms". Meeting either condition will trigger the switching strategy optimization.

[0087] A fault exists when the hardware comparator determines that the fault type (such as transformer fault, capacitor aging fault, etc.) is determined by combining the equipment operating parameters. For example, if channel 1 detects a proportional deviation of -6% (exceeding the -5% threshold) for 150ms, and simultaneously monitors a voltage fluctuation of more than 2% on the secondary side of the transformer, it is determined that there is a "transformer proportional deviation fault".

[0088] In this embodiment, dynamic optimization is a process of adjusting the initial switching strategy in real time based on triggering conditions, fault status, and regional reactive power topology map, so that the strategy is more in line with the actual working conditions (such as reducing the switching amount when there is a fault, and increasing the switching amount when there is undercompensation in the neighborhood). For example, the initial switching amount is 40kvar. If the triggering conditions are met and there is no fault, and the neighboring nodes in the topology map are significantly undercompensated, the dynamic optimization is 50kvar. If there is a transformer fault, the redundant transformer data is used to recalculate, and the optimization is 35kvar.

[0089] The beneficial effects of the above technical solution are: cross-regional collaboration is achieved through edge physical agents, the adaptability of the strategy is improved by dynamic fuzzy rule base, fault detection is enhanced by hardware comparators, and control accuracy is guaranteed by dynamic optimization. It effectively solves the problems of insufficient regional collaboration, strategy lag, and slow fault response in traditional JKW control, shortens the reactive power compensation response time, stabilizes the power factor, reduces line transmission loss, and significantly improves the stability of power grid operation.

[0090] This invention provides a JKW control method based on adaptive fuzzy control, step 1 of which includes:

[0091] Real-time monitoring of the reactive power change rate of the local power grid, and matching it with the change rate-cycle reference table to obtain the interaction cycle;

[0092] Based on the interaction cycle, the reactive power compensation status of the JKW device in the adjacent area and the compensation capacity and power factor of the local JKW device are collected to obtain interaction data, and the interaction data is encrypted to construct a cross-domain reactive power status dataset.

[0093] Based on the cross-domain reactive power status dataset and combined with the node connection relationship, line impedance parameters and equipment priority of the local power grid, a regional reactive power topology map is dynamically constructed. The node characteristics of the regional reactive power topology map are related to reactive power deficit, equipment health and voltage deviation, and the edge characteristics of the regional reactive power topology map are related to the calculated reactive power transmission loss and transmission delay between nodes.

[0094] With the optimization goals of reactive power balance, minimum transmission loss, and balanced equipment lifespan, and in conjunction with the supply and demand relationship of the regional reactive power topology map, regional coordination instructions are generated. These regional coordination instructions are related to the regional status symbol, the compensation amount of the JKW device, and the instruction priority.

[0095] In this embodiment, the reactive power change rate of the local power grid is the change in reactive power per unit time, reflecting the fluctuation speed of reactive load, and the unit is kvar / s. For example, the reactive power at time t1 is 200 kvar, and at time t2 (t2-t1=10s) it is 300 kvar. The reactive power change rate is (300-200) / 10=10 kvar / s. Specifically, the edge controller calculates the reactive power change rate of the local power grid every 100ms and uses a sliding window algorithm (window size 5 sampling points) to smooth the fluctuation.

[0096] In this embodiment, the rate of change-cycle lookup table is a preset correspondence table between reactive power change rate and data interaction cycle, used to dynamically adjust the frequency of data interaction between adjacent areas according to the reactive power fluctuation speed (the faster the fluctuation, the shorter the cycle). For example, the lookup table is set as follows:

[0097] Rate of change ≤ 5 kvar / s → Interaction period 30 s;

[0098] 5 kvar / s < rate of change ≤ 20 kvar / s → interaction period 10 s;

[0099] Rate of change > 20 kvar / s → Interaction period 2 s.

[0100] In this embodiment, the interaction period is the time interval for transmitting reactive power compensation status data between the local JKW device and the JKW device in the adjacent area, which is dynamically determined by the rate of change-period lookup table.

[0101] The adjacent area JKW device is a JKW device (intelligent reactive power compensation device) installed in an area that is geographically or electricalally adjacent to the local power grid, and it interacts with the local device through a communication link.

[0102] The compensation capacity of the local JKW device is the total capacity of the capacitors that can be put into operation at the local JKW device (including the already put into operation and the standby part), reflecting its reactive power compensation capability.

[0103] Interactive data is the set of data transmitted between the local JKW device and the JKW device in the adjacent area during the interactive cycle. It includes key information such as the compensation capacity, power factor, and reactive power change rate of both parties.

[0104] Encryption is the process of encrypting interactive data. It divides data granularity based on the correlation between local and neighboring nodes and the characteristics of the transmission path, and uses dynamic keys to ensure data security.

[0105] Cross-domain reactive power status dataset is a structured dataset formed by integrating interaction data from local and neighboring regions. It contains reactive power compensation status, equipment parameters, etc. of each region and is used to construct regional reactive power topology maps. For example, the cross-domain reactive power status dataset contains 20 parameters such as compensation capacity, power factor, reactive power deficit, and equipment health of region A (local), region B, and region C, indexed by region ID and timestamp.

[0106] The node connection relationship of a local power grid describes the topological relationship of the physical connection between various electrical nodes (such as transformers, loads, switches, and JKW devices) in the local power grid. It is usually represented by a "node-edge" model. For example, the node connection relationship of a local power grid is "10kV bus (node ​​1) → transformer (node ​​2) → JKW device (node ​​3) → load 1 (node ​​4), load 2 (node ​​5)". Node 1 and node 2 are connected by a line, and node 3 and node 2 are connected by a switch.

[0107] Line impedance parameters are parameters that describe the electrical characteristics of a line, including resistance (R) and reactance (X), which affect the transmission loss of reactive power, and are measured in Ω.

[0108] Equipment priority is a priority level set according to the importance of the equipment in the power grid (such as level 1 to 5, with level 1 being the highest). Equipment with higher priority will be given priority in ensuring its operating status during reactive power distribution.

[0109] Dynamic construction is the process of updating the regional reactive power topology map in real time based on the cross-domain reactive power state dataset and real-time grid parameters (such as switch status and load changes) to ensure that the map reflects the current state of the grid. For example, when a switch on a line in area B is disconnected, the edge features corresponding to that line are automatically deleted and the node connection relationships and transmission losses are updated during dynamic construction.

[0110] The node characteristics of a regional reactive power topology map are the attribute information of the nodes in the regional reactive power topology map, reflecting the reactive power status and equipment condition of the nodes. They mainly include reactive power deficit, equipment health, and voltage deviation. For example, the characteristics of a certain node are "reactive power deficit +120kvar, equipment health 0.85, voltage deviation +2% (higher than rated voltage)".

[0111] Reactive power deficit is the difference between the actual reactive power demand of a node and the current compensation amount (positive value is undercompensation, negative value is overcompensation), reflecting the degree of imbalance between reactive power supply and demand.

[0112] Equipment health is an indicator reflecting the operating status of the JKW unit (0-1, 1 being in good condition), calculated based on the equipment's operating years, number of switching operations, and historical fault data.

[0113] Voltage deviation is the percentage difference between the actual voltage and the rated voltage of a node, reflecting voltage stability (the national standard allows a deviation of ±7%). For example, if the rated voltage of a node is 0.4kV and the actual measured voltage is 0.42kV, the voltage deviation is (0.42-0.4) / 0.4×100%=+5%.

[0114] Edge features are the attribute information of edges in the regional reactive power topology map, reflecting the transmission characteristics of the line, mainly including reactive power transmission loss and transmission delay between nodes.

[0115] Inter-node reactive power transmission loss is the proportion of reactive power lost during transmission between two nodes, and it is related to line impedance, transmission distance, and the magnitude of reactive power.

[0116] Transmission delay is the time delay in the transmission of reactive power-related data between two nodes, and it is related to the line length and communication rate.

[0117] Reactive power balance is an indicator that measures the uniformity of reactive power deficit distribution among nodes in a region. It is calculated as the ratio of the standard deviation of reactive power deficit to the average value of each node (the smaller the ratio, the higher the balance).

[0118] Minimizing transmission loss is one of the goals of regional reactive power optimization. By adjusting the allocation of reactive power compensation, the total reactive power transmission loss in the region can be reduced to the minimum.

[0119] Equipment life balancing is one of the goals of regional reactive power optimization. By controlling the switching frequency and compensation amount of each JKW device, we can prevent a certain device from aging prematurely due to overuse, and make the equipment life balancing.

[0120] The optimization objective is a set of core objectives for regional collaborative control, including reactive power balance, minimum transmission loss, and equipment lifespan balance, which are integrated into a single optimization index through a weighted algorithm.

[0121] The instruction format for regional coordination commands is "region ID + compensation adjustment value + symbol + priority". They are sent via encrypted communication, and the recipient needs to provide feedback on the execution result.

[0122] Regional status symbols are used in regional coordination instructions to identify the regional supply and demand status. For example, "+" indicates that supply exceeds demand (reduced compensation is required), and "-" indicates that supply is less than demand (increased compensation is required).

[0123] The compensation amount area coordination instruction of the JKW device specifies the reactive capacity that the JKW device needs to be put into or removed (positive value is put into, negative value is removed).

[0124] The execution priority of instructions in the instruction priority area is coordinated (levels 1 to 3, with level 1 being the highest). Instructions with higher priority are executed before local policies. For example, level 1 instructions (such as emergency compensation during a fault) must be executed within 1 second, level 2 instructions (such as peak load adjustment) must be executed within 5 seconds, and level 3 instructions (such as daily optimization) must be executed within 30 seconds.

[0125] The beneficial effects of the above technical solution are: by dynamically adjusting the interaction cycle through real-time monitoring of reactive power change rate, and by constructing a regional reactive power topology map in combination with encrypted cross-domain data, collaborative instructions are generated through multi-objective optimization, which solves the problems of insufficient regional collaboration, data insecurity, and uneven reactive power distribution in traditional JKW control, thereby improving the regional reactive power compensation response speed and reducing transmission loss.

[0126] This invention provides a JKW control method based on adaptive fuzzy control, which encrypts the interactive data, including:

[0127] Obtain the node property association and physical distance between local nodes and any neighboring nodes in the reactive power topology map of the region, and determine the potential association granularity;

[0128] Task category matching is performed on the first task set of the local node and the second task set of any neighboring node, and the potential transmission granularity is obtained by combining the network transmission paths of the local node and any neighboring node.

[0129] Based on the potential association granularity and the potential transmission granularity, a fuzzy relationship between the local node and any of the neighboring nodes is constructed to obtain the corresponding data granularity;

[0130] The interactive data is divided into real-time sensitive slices, device feature slices, and topology-related slices according to the data granularity, and an initial security array is generated based on the slice attributes of different slices.

[0131] Obtain the working log of each node in the reactive power topology map of the region, and construct the security dimension array of the corresponding node;

[0132] A first-stage encryption program matching the initial security array is loaded from the secure storage medium, and a second-stage encryption program matching the security dimension array is loaded from the secure storage medium simultaneously.

[0133] The first-stage encryption procedure is executed to obtain the storage offset parameter of each second-stage encryption procedure. The storage offset parameter is the offset of the starting address of the corresponding second-stage encryption procedure relative to the starting address of the first-stage encryption procedure. It is constrained by the minimum erasable unit size of the real-time sensitive processing unit, device feature processing unit, and topology association unit compatible with the encryption procedure, the communication time slot period, and the transmission delay tolerance and phase difference tolerance associated with the distance deviation in the regional reactive topology.

[0134] Based on the interaction direction between neighboring nodes in adjacent regions and local nodes in the local power grid, and combined with the corresponding offset, a new dimension array is obtained by performing a security upgrade analysis on the corresponding security dimension array in the same dimension.

[0135] A secure sequence is obtained based on the initial secure array and all new dimension arrays, and the interactive data is encrypted.

[0136] In this embodiment, the node property association is weighted according to load type (industrial / civilian / new energy) and voltage level (10kV / 0.4kV) (e.g., the association weight between industrial nodes is 0.6, and between industrial and civil nodes it is 0.3). The association is calculated by the rule engine of the edge server by reading the node attribute database (stored in MySQL).

[0137] Physical distance is calculated by obtaining latitude and longitude from the power grid GIS system (or Baidu Maps Enterprise API) and then using the Haversine formula to calculate the distance in kilometers.

[0138] Potential association granularity: Formula Grel = 0.6 × type matching degree + 0.4 × (1 - distance / 10) (10km is the maximum association radius). For example, if the distance between the local node (industrial, 10kV) and the neighboring node (industrial, 10kV) is 5km, the type matching degree is 1 → Grel = 0.6 × 1 + 0.4 × 0.5 = 0.8 (high association).

[0139] In this embodiment, the first / second task set: the instructions to be executed by the local / neighborhood JKW (such as "compensate 15kvar" and "suppress 5th harmonic"), are synchronized to the edge node every 2 minutes in JSON format via the MQTT protocol.

[0140] Task category matching: Use the Jaccard coefficient J = |rw1∩rw2| / |rw1∪rw2|, where rw1 is the first task set and rw2 is the second task set. The Python script parses the JSON task fields (such as "task_type":"reactive_compensation") and calculates the intersection / union. For example, the local task set T1 (reactive power compensation, harmonic suppression) and the neighboring T2 (reactive power compensation, voltage regulation) → J = 0.5.

[0141] The network transmission path is obtained by using the traceroute tool (Linux command) or network diagnostic module to obtain the route hop count and latency (ms), which is executed periodically by the built-in diagnostic process of the edge node.

[0142] Potential transmission granularity: Formula Gtrans=0.7×J+0.3×(1-delay / 100) (100ms is the maximum delay threshold), for example, J=0.5, transmission delay 50ms→Gtrans=0.7×0.5+0.3×0.5=0.5 (medium transmission feasibility).

[0143] Fuzzy relations are defined by 5 predefined rules (e.g., IF high correlation (Grel>0.7) AND fast transmission (Gtrans)).

[0144] >0.6) THEN data granularity is fine (16 sub-chips), stored in the JKW controller Flash memory, and executed through the Mamdani fuzzy inference engine (ported from MATLAB FuzzyLogic algorithm), with a response time ≤50ms.

[0145] Data granularity: Strong cohesion (Grel>0.8 and Gtrans>0.7) → 16 sub-slices (millisecond-level encryption); Weak cohesion → 4 sub-slices (minute-level encryption). For example, Grel=0.8 and Gtrans=0.6 → inference yields medium cohesion, which is divided into 8 sub-slices.

[0146] The interactive data consists of reactive power values, harmonic content, and equipment status (such as "capacitor life 2500h") exchanged between nodes, transmitted via an SSL encrypted channel in JSON format.

[0147] Data segment division: Real-time sensitive segment: contains current reactive power deficit (e.g., -30kvar) and harmonic mutation (THD from 3% to 8%), requiring millisecond-level encryption; Equipment characteristic segment: contains capacity (200kvar) and lifespan (2500h), requiring minute-level encryption; Topology related segment: contains line impedance (0.5Ω) and connection relationship, requiring hour-level encryption.

[0148] The initial security array is generated according to chip attributes (e.g., real-time sensitive chip → SA0 = [security level: 8, key length: 256, algorithm: AES-256]), dynamically generated by the national cryptographic SM4 encryption chip through API and stored in the on-chip RAM.

[0149] The work log is recorded by JKW as the number of switching operations (1200 times) and fault codes (0x0001). A CSV file is generated every 10 minutes and uploaded to the edge server via FTP protocol.

[0150] Security dimension array: Extract the number of faults, alarms, and key update times, and perform min-max normalization using the Pandas library (Python) (e.g., fault count 2 / monthly average 4 → 0.5).

[0151] Example: SD = [Number of faults: 0.5, Number of alarms: 0.6, Key update: 0.8].

[0152] The secure storage medium is an encrypted SD card (such as the KS8695 encryption module), which uses AES-256 encryption to store multiple versions of encryption programs (adapted to different security levels).

[0153] Phased Programming: Phase 1: Lightweight SM4 program (≤10KB), adapted for real-time sensitive chips, with a fixed load address of 0x1000; Phase 2: AES + Chinese cryptographic hybrid program (≤50KB), adapted for devices / topology chips, with a dynamically calculated load address.

[0154] Load matching: Load via the SPI interface and perform CRC32 check (to ensure program integrity) through the security array-program mapping table (e.g., SA0→Program 1).

[0155] The storage offset parameter (Offset) is the address difference between the second-stage program and the first-stage program, and must satisfy:

[0156] Storage erasure unit: the least common multiple (e.g., 32KB) of the real-time processing unit (4KB), the device feature unit (8KB), and the topology association unit (16KB);

[0157] Communication time slot period: aligned with IEC61850 sampling period (e.g., 20ms, 32KB corresponds to 0.625μs per KB);

[0158] Distance-related tolerance: Transmission delay tolerance (0.1ms per km, ±0.05ms): End-to-end delay measured via OWAMP protocol;

[0159] Phase difference tolerance (0.01° per km, ±0.005°): Phase difference is measured by phasor synchronization via PMU. Specifically, this involves reading the memory controller register (to obtain the erase page size), parsing the IEC61850 message (to obtain the sampling rate), and calculating the offset (e.g., 32KB) using the least common multiple (LCM) algorithm.

[0160] In this embodiment, the method for obtaining the new dimension array is as follows:

[0161] Determine the basic structure of the original security dimension array: The original security dimension array (denoted as SD) contains the core security features of the node operation, usually 3-5 dimensions, for example: SD = [normalized value of fault count, normalized value of alarm count, key update frequency]. Specifically, the values ​​of each dimension are normalized by min-max (range [0,1]) to reflect the relative severity of the security status.

[0162] Security enhancement features are added based on the interaction direction: The interaction direction is divided into "upload" (local → neighboring domain) and "download" (neighboring domain → local), and security enhancement features need to be added accordingly: Upload scenario: Data integrity needs to be ensured, and CRC32 check bits (hexadecimal value) need to be added; Download scenario: The legality of the data source needs to be ensured, and ECDSA digital signature digest (first 8 hexadecimal digits) needs to be added.

[0163] Adjust the dimension value in conjunction with the storage offset parameter:

[0164] The storage offset parameter reflects hardware storage constraints and needs to affect the new dimension array in the following ways:

[0165] If Offset ≥ 32KB (high hardware compatibility), increase the weight of security enhancement features (such as the proportion of checksum / signature digest in the array);

[0166] If Offset < 32KB (strict hardware constraints), the original dimension value is truncated (keeping 2 decimal places) to ensure it matches the storage precision;

[0167] The final structure of the new dimension array (SD′) is: SD′ = [original dimension 1, original dimension 2, original dimension 3, security enhancement feature].

[0168] For example, original dimensions are preserved: the three basic dimensions of SD are directly preserved: [0.5, 0.6, 0.8];

[0169] Add security enhancement features: Add CRC32 check bit to the upload scenario, assuming the CRC32 value of the interactive data (such as reactive power deficit -20kvar) is 0x1A3B5C7D;

[0170] Adjusting based on Offset: Since Offset = 32KB ≥ 32KB, there is no need to truncate the original value; the checksum is fully preserved.

[0171] The final new dimension array SD′ = [0.5, 0.6, 0.8, 0x1A3B5C7D].

[0172] The security sequence is generated by merging the initial security array (70% weight) with the new dimension array (30% weight), such as SS = [Level: 8, Key: 0xABCD..., Verification: 0x5678], through a weighted fusion algorithm.

[0173] Encryption execution involves calling a hardware encryption module (such as an AES coprocessor) to perform parallel encryption on the chips: real-time sensitive chips → SM4 algorithm (millisecond level); device / topology chips → AES + national cryptographic algorithm (microsecond level). Specifically, the JKW controller triggers the encryption chip through the SPI interface, with a delay of ≤1ms / chip.

[0174] The beneficial effects of the above technical solution are as follows: by dynamically dividing data granularity, matching node association and transmission characteristics, and combining interaction direction to upgrade security strategies, on-demand encryption of interactive data is achieved (high-strength protection for sensitive data and lightweight processing for non-sensitive data). This not only improves the security of cross-regional data transmission, but also reduces the computing load of edge devices. It solves the problems of excessive encryption consuming resources and insufficient encryption posing risks in traditional unified encryption, and adapts to the dynamic and security requirements of regional reactive power collaborative control.

[0175] This invention provides a JKW control method based on adaptive fuzzy control, step 2 of which includes:

[0176] Constructing a four-dimensional fuzzy input vector The dimensions are then standardized and mapped to the interval between -1 and 1 to obtain the standard fuzzy input vector X0, where ΔK is the proportional deviation of the mutual inductor and Δcosφ is the power factor deviation. The reactive power gradient, ΔQ c For regional coordination commands;

[0177] The standard fuzzy input vector is mapped to fuzzy language variables based on a dynamically updated fuzzy rule base, and the initial switching strategy is inferred and output.

[0178] Among them, capacitor switching amount Where m1 represents the number of rules in the dynamically updated fuzzy rule base; w iLet represent the weight of the i-th rule; fi(X0) represents the output function of the i-th rule;

[0179] The thyristor trigger phase θ0 is the optimal trigger angle dynamically calculated based on the voltage zero-crossing point;

[0180] The collaboration enable flag is activated when the set standard conditions are met.

[0181] In this embodiment, ΔK=(K 测 -K 实 ) / K 实 ;

[0182] Δcosφ=cosφ 实 -cosφ 目标 ;

[0183] Q = 100 + 10t + 0.5t 2 ;

[0184] fi(X0)=min{Ai(x1),Ai(x2),Ai(x3),Ai(x4)}, where X0={x1,x2,x3,x4}, Ai represents the membership relationship between the input variable and the fuzzy rule base, taking values ​​from 0 to 1, where x1,x2,x3,x4 are the values ​​of proportional deviation, power factor deviation, reactive power change gradient, and regional collaborative command mapping, respectively.

[0185] The thyristor trigger phase, for example, the fundamental voltage phase is 120°, and the optimal trigger angle is 60° (60° trigger after zero crossing, at which point the voltage change rate is moderate and the inrush current is ≤1.5 times the rated current). Specifically, the phase-locked loop (PLL) tracks the voltage zero crossing, calculates the trigger angle every cycle (20ms), and outputs the trigger pulse through the FPGA.

[0186] The collaboration enable flag is set to 1 when the reactive power deficit in the region is greater than 80% of the total compensation capacity. For example, if the total regional deficit is 500 kvar and the total capacity is 600 kvar, 500 / 600 ≈ 0.83 > 0.8, then the flag = 1 (activating neighborhood collaboration).

[0187] The beneficial effects of the above technical solution are: by dynamically constructing a four-dimensional input vector (integrating transformer deviation, power factor, reactive power change gradient and regional coordination command), combined with a self-updating rule base and refined fuzzy reasoning, precise control of capacitor switching and triggering phase is achieved, supporting the intelligent upgrade of reactive power compensation in the distribution network.

[0188] This invention provides a JKW control method based on adaptive fuzzy control, step 3 of which includes:

[0189] The proportional deviation and power factor deviation of the current transformer are connected to a dual-channel high-speed hardware comparator, and a threshold comparison mechanism is configured for each channel to determine the decision result. The dual-channel high-speed hardware comparator has a built-in temperature compensation circuit, and the threshold dynamic adjustment logic is linked to the local power grid frequency.

[0190] A fault code is generated based on the judgment result, and the fault code is used to locate the fault.

[0191] Preferably, generating a fault code based on the judgment result and using the fault code for fault location includes:

[0192] The judgment results, the cumulative number of anomalies in the timing logic judgments, and the offset of real-time monitored equipment operating parameters are aligned to construct a fault feature vector;

[0193] The fault feature vector is matched with the pre-defined 16-bit fault code, and the fault category is determined based on the TypeCode and SeverityCode. The fault propagation path is calculated by combining the LocationCode, TimeStamp, and the shortest path algorithm of the topology graph to determine the most likely fault source.

[0194] A fault cause-effect graph is constructed. When a compound fault is detected, the main fault source is determined by Bayesian network inference. The node weights of the fault cause-effect graph are taken from the equipment health of the regional reactive power topology graph, and the fault propagation probability of the edges is obtained by statistical analysis of historical fault data from the past m0 months.

[0195] In this embodiment, a 16-bit fault code is constructed:

[0196] Err = TypeCode (4 bits) + SeverityCode (3 bits) + LocationCode (5 bits) + TimeStamp (4 bits), where TypeCode is the type code, SeverityCode is the severity code, LocationCode is the location code, and TimeStamp is the time code.

[0197] In this embodiment, the dual-channel high-speed hardware comparator is a high-speed analog circuit with two independent comparison channels. It compares ΔK, Δcosφ with the threshold and outputs the judgment result. For example, channel 1 compares ΔK (threshold ± 5%), and channel 2 compares Δcosφ (threshold ± 0.05%). When ΔK = -6%, channel 1 outputs an abnormal value; when Δcosφ = -0.04, channel 2 outputs a normal value.

[0198] In this embodiment, the threshold comparison mechanism presets a dynamic threshold for each channel (such as ΔK threshold ±5%, Δcosφ threshold ±0.05). The comparator outputs a normal / abnormal / critical decision based on the relationship between the input value and the threshold. For example, if ΔK = -4.9% (close to the -5% threshold), it is judged as critically abnormal; if ΔK = -5.1%, it is judged as abnormal.

[0199] The cumulative number of abnormal decisions in the timing logic is the number of abnormal decisions made by the same channel within a unit of time (e.g., 10s), reflecting the persistence of the fault (the more times, the more likely the fault is real). For example, if the ΔK channel makes 3 consecutive abnormal decisions within 10s, the cumulative number is 3.

[0200] The deviation of equipment operating parameters is the deviation between the real-time parameters (such as temperature and voltage) of the equipment and the rated values. For example, the temperature deviation of the current transformer is Tactual - Trated, and the voltage deviation is Uactual - Urated.

[0201] The fault feature vector is a multi-dimensional vector that integrates the decision results, the cumulative number of anomalies, and the parameter offsets. It is used to characterize fault features (such as the cumulative number of anomalies). For example, the vector dimension is 4 (2 decision results + 2 parameter offsets), and the numerical standardization range is [-1, 1], which facilitates subsequent matching.

[0202] In this embodiment, the 16-bit fault code (Err) has the following structure:

[0203] TypeCode (4 bits): Fault type (e.g., 0001 = ΔK fault, 0010 = Δcosφ fault, 0011 = compound fault).

[0204] SeverityCode (3 bits): Severity level (000 = Minor, 001 = Moderate, 010 = Severe, 100 = Urgent).

[0205] LocationCode (5 bits): Fault location (e.g., 00001 = 1# transformer, 00010 = 2# JKW).

[0206] TimeStamp (4-bit): Time segmentation (every 15 minutes, 0000 = 0–15 minutes, 0001 = 15–30 minutes, ...).

[0207] For example, at 8:05 (belonging to segment 0000), the ΔK critical fault of transformer #1 occurred, and the fault code was: Err=TypeCode0001+SeverityCode010+LocationCode00001+TimeStamp0000=0001010000010000 (binary).

[0208] The shortest path algorithm for the topology graph is based on the regional reactive power topology graph (nodes are devices, edges are connections). It uses Dijkstra's algorithm to calculate the shortest path from the faulty node to other nodes to locate the source of the fault propagation. For example, if the fault code indicates that transformer #1 is abnormal, in the topology, transformer #1 is connected to node A (distance 1), node B (distance 2), and node C (distance 3). The shortest path is node A, so node A is investigated first.

[0209] The fault cause-effect graph consists of nodes representing devices (weight = device health, e.g., the health of transformer #1 is 0.9) and edges representing fault propagation relationships (e.g., transformer failure → JKW malfunction, with the probability calculated from historical data over the past m0 months, where m0 = 6 months).

[0210] Bayesian network inference is used to determine the primary fault source when there are multiple faults (such as simultaneous anomalies in ΔK and Δcosφ). For example, if the posterior probability of a current transformer fault is 80% and the fault probability of JKW is 20%, then the primary fault source is the current transformer.

[0211] The beneficial effects of the above technical solution are: it enables rapid fault judgment through dual-channel high-speed comparison and dynamic threshold, accurately locates the fault location with 16-bit fault code and topological shortest path, and analyzes the root cause of compound faults with Bayesian network and causal graph, thus solving the problems of "high false judgment rate, ambiguous location and difficulty in analyzing compound faults" in traditional fault detection.

[0212] This invention provides a JKW control method based on adaptive fuzzy control, step 4 of which includes:

[0213] Convert the trigger signal output by the hardware comparator into a trigger strength coefficient Kt;

[0214] The fault is encoded using 3-bit binary code to generate the fault impact weight Wf.

[0215] Extract the node correlation degree R from the reactive power topology map of the region. i1j1 And the reactive power transmission loss coefficient γ=0.01+0.05L·(1+δ1·(T(t)-T0)+δ2·μ(t)), where R i1j1 δ1 represents the node correlation degree between local node i1 and neighboring node j1; L represents the line length; T(t) represents the real-time ambient temperature; T0 represents the reference temperature; δ1 represents the temperature influence coefficient of 0.002 / ℃; μ(t) represents the real-time line load rate; δ2 represents the load rate influence coefficient.

[0216] The initial switching strategy is dynamically optimized:

[0217] Under normal operating conditions, the following optimizations are performed:

[0218]

[0219] in, The optimized capacitor switching amount; m2 represents the total number of neighboring nodes; ΔQ j1 The reactive power deficit of the j1-th neighboring node; is the health decay coefficient of the j1th neighboring node; For time period correction factors;

[0220]

[0221] in, The optimized thyristor trigger phase;

[0222] When the operating condition is faulty, the following optimizations are performed:

[0223] If an abnormality is detected in the current transformer, the switching amount will be recalculated based on the data from the redundant current transformers.

[0224] If a system-level fault is triggered, the collaborative enable flag is forcibly activated: B0=1, and the preset percentage of switching amount is allocated to the neighboring nodes according to the regional reactive power topology map, where B0 is the result of the collaborative enable flag;

[0225] If the reactive power topology map of the region fails due to communication interruption, a preset neighborhood compensation protocol is activated, and the topology recovery status is monitored in real time. Once the topology is restored, the system switches to dynamic optimization mode.

[0226] The trigger signal of the hardware comparator is the level signal (high / low / critical) output by the dual-channel high-speed comparator, which characterizes whether the grid parameters (proportional deviation, power factor deviation) exceed the limit and the degree of exceeding the limit. For example, when the power factor deviation Δcosφ=-0.06 (exceeding the threshold -0.05), the comparator outputs a high level (5V), which represents "severe exceeding the limit".

[0227] The trigger strength coefficient Kt's hi quantizes the trigger signal into a coefficient in the [0,1] interval, reflecting the urgency of the fault / abnormality (the larger the value, the more urgent the need for adjustment). For example:

[0228] Critical limit exceeded (level jump) → Kt = 0.5;

[0229] Continuous high level (serious fault) → Kt = 0.9.

[0230] The 3-bit binary code (fault) encodes the fault type (such as transformer abnormality, power factor exceeding the limit) into 3 bits (000~111), corresponding to the fault level (minor → severe). For example: transformer proportional deviation exceeds 5% (non-continuous) → code 010 (general fault).

[0231] Power factor deviation exceeds 0.1 (lasting 10s) → Code 101 (serious fault).

[0232] The fault impact weight Wf is the quantization weight (0~1) of the 3-bit code. The larger the code value, the higher the weight (e.g., 000→0.1, 111→0.9). For example, the code 101(5)→Wf=0.5+0.05×5=0.75 (linear mapping).

[0233] Node correlation degree R i1j1 It is the electrical coupling strength (0-1) between the local node and the neighboring nodes, which is determined by power flow calculation or topological distance (the closer the distance, the higher the correlation).

[0234] In this embodiment, μ(t) takes the value of 0 to 1, and T0 is taken as 25°C based on the real-time current / rated current.

[0235]

[0236] The redundant current transformer data is the proportional deviation data of the switch to the backup current transformer when the main current transformer fails, and the switching amount is recalculated.

[0237] The coordination enable flag B0 is a coordination flag that is forcibly activated when there is a system-level fault (such as a bus voltage collapse) (B0=1 means that neighborhood coordination is enabled).

[0238] The default neighborhood compensation protocol is the default compensation strategy (such as allocating according to the historical average ratio) that is enabled when the topology map fails due to communication interruption. For example, when communication is interrupted, the distribution amount is allocated according to neighborhood 1:30%, neighborhood 2:40%, and neighborhood 3:30%, and the topology recovery status is checked every 10 seconds (heartbeat packet).

[0239] The beneficial effects of the above technical solution are: by trigger intensity quantization, fault weight adaptation, and topology-loss dynamic coupling, the working condition adaptive optimization of the reactive power compensation strategy is achieved: under normal working conditions, the compensation accuracy is improved by integrating the neighborhood state and time period characteristics; under fault working conditions, redundancy switching and collaborative control work together to reduce the risk of fault propagation.

[0240] This invention provides a JKW control method based on adaptive fuzzy control, and further includes:

[0241] Collect the optimized power factor and neighborhood compensation response, and calculate the optimization effect evaluation coefficient E;

[0242]

[0243] Wherein, cosφ new For the optimized power factor, ΔQ sy The optimized compensation response; ΔQ raThe rated reactive power is Nc; the switching frequency is τ1, τ2, and τ3 are weighting coefficients, and τ1+τ2+τ3=1;

[0244] When the optimization effect evaluation coefficient E is greater than the preset evaluation coefficient, the fuzzy rule base is dynamically updated.

[0245] The neighborhood compensation response is the actual amount of reactive power compensation adjusted by the neighboring node after receiving the coordination instruction (positive value is input, negative value is cut-off), reflecting the matching degree between the instruction and the execution. For example, if the instruction requires the neighboring node to input 60kvar, the actual input is 55kvar (ΔQsy=55kvar); or if the instruction cuts off 20kvar, the actual cut-off is 18kvar (ΔQsy=-18kvar).

[0246] The switching count Nc is the cumulative number of capacitor switching operations within a statistical period (e.g., 10 minutes), reflecting the risk of equipment lifespan degradation (the more times, the faster the lifespan declines).

[0247] Under normal operating conditions: τ1 = 0.6 (power factor priority), τ2 = 0.3 (compensation response), τ3 = 0.1 (number of switching cycles);

[0248] Fault conditions: τ1 = 0.3, τ2 = 0.5 (priority coordinated response), τ3 = 0.2;

[0249] Load mutation: τ1 = 0.7 (fast correction power factor), τ2 = 0.2, τ3 = 0.1.

[0250] In this embodiment, |cosφ new -0.98|: Power factor deviation (the smaller the better, reflecting the accuracy of compensation);

[0251] The ratio of compensation response to rated demand (the closer to 1 the better, reflecting the synergistic effect);

[0252] Nc: Number of cutting cycles (the fewer the better, reflecting equipment wear and tear).

[0253] The dynamic update logic refers to using reinforcement learning (Q-learning) or genetic algorithms to optimize the weights and number of fuzzy rules when E exceeds the preset value, so that the subsequent compensation strategy is more in line with the working conditions. For example, the weight of "power factor deviation negative large → switching amount positive large" in the original rule base is adjusted from 0.12 to 0.15 to enhance the compensation strength of this scenario.

[0254] The beneficial effects of the above technical solution are: by using the evaluation coefficient E of dynamic weight adaptation and multi-index fusion, the reactive power compensation effect can be accurately quantified, which significantly enhances the dynamic adaptability and long-term stability of reactive power compensation in the distribution network.

[0255] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A JKW control method based on adaptive fuzzy control, characterized in that, include: Step 1: Based on the reactive power compensation status of JKW devices in adjacent areas through edge physical agent interaction, construct a regional reactive power topology map by combining local power grid topology data, and generate regional coordination instructions according to supply and demand relationship, wherein the supply and demand relationship is determined by the ratio of total reactive power deficit to total compensation capacity in the region. Step 2: Collect the proportional deviation, power factor deviation, and reactive power change gradient of the current transformer, and construct a standard fuzzy input vector in combination with the regional collaborative instructions. Map the standard fuzzy input vector into fuzzy language variables through a dynamically updated fuzzy rule base, and infer and output the initial switching strategy. The initial switching strategy is related to the capacitor switching amount, thyristor trigger phase, and collaborative enable flag. Step 3: Compare and analyze the proportional deviation and power factor deviation using a hardware comparator to determine the triggering conditions and the existence of faults; Step 4: Based on the triggering conditions and the existence of faults, and combined with the regional reactive power topology map, dynamically optimize the initial switching strategy to achieve control of JKW.

2. The JKW control method based on adaptive fuzzy control according to claim 1, characterized in that, Step 1 includes: Real-time monitoring of the reactive power change rate of the local power grid, and matching it with the change rate-cycle reference table to obtain the interaction cycle; Based on the interaction cycle, the reactive power compensation status of the JKW device in the adjacent area and the compensation capacity and power factor of the local JKW device are collected to obtain interaction data, and the interaction data is encrypted to construct a cross-domain reactive power status dataset. Based on the cross-domain reactive power status dataset and combined with the node connection relationship, line impedance parameters and equipment priority of the local power grid, a regional reactive power topology map is dynamically constructed. The node characteristics of the regional reactive power topology map are related to reactive power deficit, equipment health and voltage deviation, and the edge characteristics of the regional reactive power topology map are related to the calculated reactive power transmission loss and transmission delay between nodes. With the optimization goals of reactive power balance, minimum transmission loss, and balanced equipment lifespan, and in conjunction with the supply and demand relationship of the regional reactive power topology map, regional coordination instructions are generated. These regional coordination instructions are related to the regional status symbol, the compensation amount of the JKW device, and the instruction priority.

3. The JKW control method based on adaptive fuzzy control according to claim 2, characterized in that, Encrypting the interactive data includes: Obtain the node property association and physical distance between local nodes and any neighboring nodes in the reactive power topology map of the region, and determine the potential association granularity; Task category matching is performed on the first task set of the local node and the second task set of any neighboring node, and the potential transmission granularity is obtained by combining the network transmission paths of the local node and any neighboring node. Based on the potential association granularity and the potential transmission granularity, a fuzzy relationship between the local node and any of the neighboring nodes is constructed to obtain the corresponding data granularity; The interactive data is divided into real-time sensitive slices, device feature slices, and topology-related slices according to the data granularity, and an initial security array is generated based on the slice attributes of different slices. Obtain the working log of each node in the reactive power topology map of the region, and construct the security dimension array of the corresponding node; A first-stage encryption program matching the initial security array is loaded from the secure storage medium, and a second-stage encryption program matching the security dimension array is loaded from the secure storage medium simultaneously. The first-stage encryption procedure is executed to obtain the storage offset parameter of each second-stage encryption procedure. The storage offset parameter is the offset of the starting address of the corresponding second-stage encryption procedure relative to the starting address of the first-stage encryption procedure. It is constrained by the minimum erasable unit size of the real-time sensitive processing unit, device feature processing unit, and topology association unit compatible with the encryption procedure, the communication time slot period, and the transmission delay tolerance and phase difference tolerance associated with the distance deviation in the regional reactive topology. Based on the interaction direction between neighboring nodes in adjacent regions and local nodes in the local power grid, and combined with the corresponding offset, a new dimension array is obtained by performing a security upgrade analysis on the corresponding security dimension array in the same dimension. A secure sequence is obtained based on the initial secure array and all new dimension arrays, and the interactive data is encrypted.

4. The JKW control method based on adaptive fuzzy control according to claim 1, characterized in that, Step 2 includes: Constructing a four-dimensional fuzzy input vector The dimensions are then standardized and mapped to the interval between -1 and 1 to obtain the standard fuzzy input vector X0, where ΔK is the proportional deviation of the mutual inductor and Δcosφ is the power factor deviation. The reactive power gradient, ΔQ c For regional coordination commands; The standard fuzzy input vector is mapped to fuzzy language variables based on a dynamically updated fuzzy rule base, and the initial switching strategy is inferred and output. Among them, capacitor switching amount Where m1 represents the number of rules in the dynamically updated fuzzy rule base; w i Let represent the weight of the i-th rule; fi(X0) represents the output function of the i-th rule; The thyristor trigger phase θ0 is the optimal trigger angle dynamically calculated based on the voltage zero-crossing point; The collaboration enable flag is activated when the set standard conditions are met.

5. The JKW control method based on adaptive fuzzy control according to claim 1, characterized in that, Step 3 includes: The proportional deviation and power factor deviation of the current transformer are connected to a dual-channel high-speed hardware comparator, and a threshold comparison mechanism is configured for each channel to determine the decision result. The dual-channel high-speed hardware comparator has a built-in temperature compensation circuit, and the threshold dynamic adjustment logic is linked to the local power grid frequency. A fault code is generated based on the judgment result, and the fault code is used to locate the fault.

6. The JKW control method based on adaptive fuzzy control according to claim 5, characterized in that, Based on the judgment result, a fault code is generated, and fault location is performed based on the fault code, including: The judgment results, the cumulative number of anomalies in the timing logic judgments, and the offset of real-time monitored equipment operating parameters are aligned to construct a fault feature vector; The fault feature vector is matched with the pre-constructed 16-bit fault code, and the fault category is locked according to the TypeCode and SeverityCode. The fault propagation path is calculated by combining the LocationCode, TimeStamp and the shortest path algorithm of the topology graph to determine the most likely fault source. A fault cause-effect graph is constructed. When a compound fault is detected, the main fault source is determined by Bayesian network inference. The node weights of the fault cause-effect graph are taken from the device health of the regional reactive power topology graph, and the fault propagation probability of the edges is obtained by statistical analysis of historical fault data from the past m0 months.

7. The JKW control method based on adaptive fuzzy control according to claim 1, characterized in that, Step 4 includes: Convert the trigger signal output by the hardware comparator into a trigger strength coefficient Kt; The fault is encoded using 3-bit binary code to generate the fault impact weight Wf. Extract the node correlation degree R from the reactive power topology map of the region. i1j1 And the reactive power transmission loss coefficient γ=0.01+0.05L·(1+δ1·(T(t)-T0)+δ2·μ(t)), where R i1j1 δ1 represents the node correlation degree between local node i1 and neighboring node j1; L represents the line length; T(t) represents the real-time ambient temperature; T0 represents the reference temperature; δ1 represents the temperature influence coefficient of 0.002 / ℃; μ(t) represents the real-time line load rate; δ2 represents the load rate influence coefficient. The initial switching strategy is dynamically optimized: Under normal operating conditions, the following optimizations are performed: in, The optimized capacitor switching amount; m2 represents the total number of neighboring nodes; ΔQ j1 The reactive power deficit of the j1-th neighboring node; is the health decay coefficient of the j1th neighboring node; For time period correction factors; in, The optimized thyristor trigger phase; When the operating condition is faulty, the following optimizations are performed: If an abnormality is detected in the current transformer, the switching amount will be recalculated based on the data from the redundant current transformers. (1-Wf); If a system-level fault is triggered, the collaborative enable flag is forcibly activated: B0=1, and the preset percentage of switching amount is allocated to the neighboring nodes according to the regional reactive power topology map, where B0 is the result of the collaborative enable flag; If the reactive power topology map of the region fails due to communication interruption, a preset neighborhood compensation protocol is activated, and the topology recovery status is monitored in real time. Once the topology is restored, the system switches to dynamic optimization mode.

8. The JKW control method based on adaptive fuzzy control according to claim 7, characterized in that, Also includes: Collect the optimized power factor and neighborhood compensation response, and calculate the optimization effect evaluation coefficient E; Wherein, cosφ new For the optimized power factor, ΔQ sy The optimized compensation response; ΔQ ra The rated reactive power is Nc; the number of switching operations is τ1, τ2, and τ3 are weighting coefficients, and τ1 + τ2 + τ3 = 1. When the optimization effect evaluation coefficient E is greater than the preset evaluation coefficient, the fuzzy rule base is dynamically updated.

Images