Power distribution transformer group reconstruction and optimization method and system based on intelligent soft switch

Abstract

The application discloses a power distribution transformer group reconstruction and optimization method and system based on intelligent soft switch, and belongs to the field of power grid optimization control; the method deploys a flexible intelligent soft switch with an energy storage unit on the low-voltage side of a distribution transformer; through load prediction and real-time identification of loss parameters, the optimal economic operation interval and dynamic overload margin of the distribution transformer are dynamically calculated; a multi-objective optimization model with the minimum total loss, the minimum overload margin variance, the minimum voltage deviation and the optimal protection setting sensitivity as the targets is constructed, the power setting value is solved to execute inductive load transfer, and the system has an active compensation treatment function for reverse power, the system performs sandbox verification through local virtual protection logic, executes safe rollback and degradation control in case of failure, and the application realizes efficient inductive load transfer of the distribution transformer group, significantly improves the operation economy and overload prevention capability of the power grid, and guarantees the reliability of relay protection and the operation safety of the system.

Description

Technical Field

[0001] This invention belongs to the field of power grid optimization control, and more specifically, relates to a method and system for reconfiguration and optimization of distribution transformer groups based on intelligent soft switching. Background Technology

[0002] With the large-scale integration of new loads such as distributed photovoltaic power and electric vehicle charging piles, the spatiotemporal distribution of load rates of distribution transformers in the power distribution network is severely unbalanced, leading to the following technical problems:

[0003] The overall losses of distribution transformers consist of iron losses and copper losses, and there exists an optimal economic operating range (typically 30%-80% load rate). In traditional distribution networks, the load rate of distribution transformers is entirely determined by the load they carry and cannot be actively adjusted. Statistics show that approximately 40%-60% of distribution transformers deviate from their optimal economic operating range for extended periods, with an average annual overall loss rate of about 3.5%-4.5%, resulting in significant energy waste.

[0004] The load distribution of adjacent distribution transformers is severely uneven, often resulting in "one transformer being overloaded and the other being lightly loaded." Traditional methods involve manually closing tie switches to transfer the load, but this has a slow response time (30 minutes to several hours), requires a short power outage (5-10 seconds) during the switching process, and can only achieve a coarse adjustment of "fully open" or "fully closed," without being able to precisely control the amount of power transferred.

[0005] During peak solar power output periods, solar power generation may exceed local load power, leading to power backflow in distribution transformers and causing problems such as reduced efficiency and voltage exceeding limits. The traditional solution is to limit solar power generation ("curtailment"), resulting in a waste of clean energy.

[0006] Existing technologies typically employ fixed rated capacity constraints (load rate ≤ 100%), failing to fully utilize the dynamic overload capacity of distribution transformers. This results in premature load shedding during peak load periods, impacting power supply reliability.

[0007] Finally, the investment cost of existing flexible DC distribution networks is 3-5 times that of traditional distribution networks; the response time of centralized optimized control is 5-15 minutes; and the load transfer process generates 5%-10% voltage fluctuations, which can be clearly perceived by the user side.

[0008] Therefore, there is an urgent need for a low-cost, fast-response, and sensorless control technology for optimizing the operation of distribution transformer groups. Summary of the Invention

[0009] To address the above shortcomings, this invention provides a method for reconfiguration and optimization of distribution transformer groups based on intelligent soft switching, comprising the following steps:

[0010] S1. Deploy a flexible intelligent soft switch on the low-voltage side of an adjacent distribution transformer. The flexible intelligent soft switch adopts a bidirectional converter topology and has a parallel energy storage unit on the DC side.

[0011] S2. Rolling forecast of future load for each distribution transformer. The forecast is based on the load change rate judgment logic and is reassessed every hour or when the load change rate exceeds the preset threshold.

[0012] S3. Real-time identification of loss parameters of each distribution transformer, and dynamic calculation of the optimal economic operating range based on operating conditions;

[0013] S4. Establish a thermal model of the distribution transformer and calculate the dynamic overload margin;

[0014] S5. Construct a multi-objective optimization model. The objective functions include minimizing the total loss of the distribution transformer group, minimizing the overload margin variance, minimizing the voltage deviation, and optimizing the sensitivity of the protection setting. The constraints include dynamic efficiency range constraints, power flow constraints, capacity constraints, voltage constraints, and overload margin constraints.

[0015] S6. Solve the model using a multi-objective optimization algorithm and output the power setting value of each flexible intelligent soft switch;

[0016] S7. Generate a power regulation curve based on the optimization results, execute phase synchronization control and dual-machine voltage vector matching, and achieve seamless load transfer through flexible intelligent soft switching output power regulation and instantaneous power compensation of energy storage unit;

[0017] S8. Real-time monitoring of the power direction of the distribution transformer. When power backfeed is detected, the flexible intelligent soft switch is controlled to transfer the backfeed power while performing active harmonic compensation and control, and injecting the backfeed power into the adjacent distribution transformer.

[0018] S9. Monitor the status of the flexible intelligent soft switch in real time, execute a safe backoff path before blocking the faulty branch, pull back the load to the original transformer's load-bearing range, and then block the flexible intelligent soft switch and switch to degraded control mode.

[0019] Furthermore, in step S1, the flexible intelligent soft switch adopts a back-to-back bidirectional AC-DC-AC converter topology, and the energy storage unit is a supercapacitor.

[0020] In step S2, the future load of each distribution transformer is predicted in a rolling manner based on a neural network. The input dimensions include historical load curves, weather forecasts, and date types.

[0021] In step S3, the recursive least squares method is used to identify the iron loss and copper loss parameters of each distribution transformer in real time, and the upper and lower limits of the optimal economic operating range are dynamically calculated based on the transformer aging coefficient, temperature correction coefficient, and harmonic distortion rate influence coefficient.

[0022] Furthermore, in step S3, the lower limit of the optimal economic operating range is:

[0023] Equation (1)

[0024] The upper limit of the optimal economic operating range is:

[0025] Equation (2)

[0026] in, This is the lower limit of the baseline efficiency range. The upper limit of the baseline efficiency range, The aging factor is... For the service life of the distribution transformer, This is a temperature correction factor. The deviation between the measured temperature and the rated temperature of the distribution transformer. The harmonic distortion rate influence coefficient. This represents the total harmonic distortion rate of the transformer current.

[0027] Furthermore, in step S4, a thermal model of the distribution transformer is established based on the IEEE C57.91 standard, and the formula for calculating the dynamic overload margin is as follows:

[0028]

[0029] in, For distribution transformers exist Dynamic overload margin at any given time. This refers to the maximum allowable temperature of the transformer winding. For distribution transformer Temperature is measured in real time. Ambient temperature;

[0030] The power flow constraints in step S5 include:

[0031] Output power constraints of distribution transformers:

[0032]

[0033] End-load node voltage constraints:

[0034]

[0035] in, For distribution transformers The output active power, For the active power of the load, For power transmission in flexible intelligent soft switches, For line current, The equivalent line resistance from the distribution transformer to the user, The equivalent line reactance from the distribution transformer to the user, For the low-voltage side bus voltage of the distribution transformer, This refers to the voltage at the end load node. The power factor.

[0036] Furthermore, the objective function of the multi-objective optimization model in step S5 is:

[0037]

[0038]

[0039]

[0040]

[0041] in, For the total loss of the transformer group, For distribution transformers At load rate The overall power loss is as follows. For the loss weighting coefficient of flexible intelligent soft switching, For flexible intelligent soft switches Self-consumption For overload margin variance, For a moment Dynamic overload margin of each distribution transformer The variance; The sum of squares of voltage deviations For nodes Voltage deviation, For sensitivity evaluation function, For fixed parameters Probability of protection malfunction Sensitivity coefficient;

[0042] In step S6, the NSGA-III multi-objective evolutionary algorithm is used to solve the multi-objective optimization model, output the Pareto optimal solution set, and select the optimal solution based on the real-time electricity price and load tension.

[0043] Furthermore, in step S7:

[0044] The power regulation curve is a power ramp-up curve, and the ramp-up rate is adaptively adjusted according to the voltage deviation.

[0045] The energy storage unit automatically charges and discharges according to the DC bus voltage deviation to compensate for the instantaneous power gap;

[0046] When voltage fluctuations exceed a preset threshold, the ramp rate is reduced to achieve inductive load transfer with voltage fluctuations of less than 2%.

[0047] Furthermore, the triggering condition for the reverse power control in step S8 is:

[0048] When the active power of the distribution transformer is detected to be negative, it is determined to be power backfeeding;

[0049] The system automatically calculates the absolute value of the reverse power and generates a reverse power transfer command;

[0050] The flexible intelligent soft switch injects back power into the adjacent distribution transformer with the highest load rate and overload margin that meets safety requirements.

[0051] Furthermore, the specific strategy for the degradation control mode in step S9 is as follows:

[0052] After the fault-tolerant intelligent soft-locking is closed, the system automatically switches to the traditional radial operating topology;

[0053] The optimization objective has been downgraded from minimizing the total loss of the transformer group to avoiding transformer overload and voltage exceeding limits.

[0054] Only the power distribution of the non-faulty flexible intelligent soft switches is optimized, while the faulty branches are restored to traditional operation;

[0055] After the fault is resolved, the system automatically returns to normal optimization mode.

[0056] This invention also discloses a distribution transformer group reconfiguration and optimization system based on intelligent soft switching, used to implement the above-mentioned optimization method, including:

[0057] The flexible intelligent soft-switching module is deployed on the low-voltage side of the adjacent distribution transformer, adopting a bidirectional converter topology, with energy storage units connected in parallel on the DC side;

[0058] The data acquisition module is installed in each distribution transformer to collect electrical parameters in real time;

[0059] The load forecasting module predicts the future load of each distribution transformer;

[0060] The parameter identification module identifies the loss parameters of the distribution transformer in real time and dynamically calculates the optimal economic operating range.

[0061] The overload margin calculation module calculates the dynamic overload margin of each distribution transformer in real time.

[0062] The multi-objective optimization module solves the multi-objective optimization model that minimizes the total loss of the distribution transformer group, the overload margin variance, and the voltage deviation, and outputs the power setpoints of each flexible intelligent soft switch.

[0063] The sensorless control module coordinates the flexible intelligent soft switch and energy storage unit based on the optimization results to achieve sensorless load transfer;

[0064] The reverse power control module monitors the power direction of the distribution transformer in real time, and automatically triggers a reverse power transfer command when reverse power is detected.

[0065] The fault diagnosis and degradation control module monitors the operating status of the flexible intelligent soft switch in real time, and automatically locks and switches to degradation control mode when a fault occurs.

[0066] Furthermore, the flexible intelligent soft-switching module adopts a back-to-back bidirectional AC-DC-AC converter topology, and the energy storage unit is a supercapacitor energy storage unit with a single-sided rated capacity of 50-200kVA and a response time of less than 10 milliseconds.

[0067] The load forecasting module uses an LSTM neural network to perform rolling forecasts of the load of each distribution transformer for the next 15 minutes.

[0068] The parameter identification module uses the recursive least squares method to identify the iron loss and copper loss parameters of the distribution transformer in real time.

[0069] The overload margin calculation module calculates the dynamic overload margin of each distribution transformer in real time based on the IEEE C57.91 standard.

[0070] The multi-objective optimization module uses an improved NSGA-III algorithm to solve the multi-objective optimization model, and automatically adjusts the reference point weights based on historical scheduling decisions and actual results through the Q-Learning reinforcement learning algorithm.

[0071] The fault diagnosis and degradation control module performs sandbox verification through local virtual protection logic. When a fatal abnormal signal is detected, it first calls the energy storage unit to smoothly reduce the transmission power of the branch where the faulty flexible intelligent soft switch is located to zero, and then executes the contactor action and IGBT blocking logic.

[0072] Compared with the prior art, the present invention has the following advantages:

[0073] 1. Significantly reduce the overall loss of distribution transformer groups

[0074] By using flexible intelligent soft switches to achieve continuous and precise load transfer between distribution transformers, each distribution transformer can operate within its dynamic optimal economic operating range. This can reduce the average annual comprehensive loss rate of the distribution transformer group from 3.5%-4.5% to 2.5%-3.0%, a reduction of 28%-44%.

[0075] 2. Balance the overload margin of distribution transformers to improve power supply reliability.

[0076] By minimizing the overload margin variance as the optimization objective, the standard deviation of the overload margin of each distribution transformer in the distribution transformer group was reduced from 25%-35% to 8%-15%, a reduction of 57%-68%. The number of overload trips of distribution transformers decreased by 83%, and the average lifespan was extended by 25%-40%.

[0077] 3. Achieve non-inductive load transfer with voltage fluctuations of less than 2%.

[0078] Through power ramping curves, adaptive ramping rates, and instantaneous power compensation from supercapacitors, voltage fluctuations during load transfer are controlled within 2%, and are imperceptible to the user.

[0079] 4. Actively absorb the power fed back by distributed photovoltaic power

[0080] By actively transferring power back to the grid, the photovoltaic (PV) grid connection capacity is increased from 30%-50% of the rated capacity of the distribution transformer to 50%-80%, an increase of 40%-100%. The light-load time of the distribution transformer is reduced by 80%, and the number of voltage over-limit incidents is reduced by 80%.

[0081] 5. Make full use of the dynamic overload capacity of the distribution transformer.

[0082] By calculating the dynamic overload margin in real time, the overall load acceptance capacity of the distribution transformer group is increased by 15%-25%, avoiding the investment in new distribution transformers.

[0083] 6. Automatic fault diagnosis and degradation control

[0084] Through the fault diagnosis and degradation control module, the system reliability level reaches over 99.5%. In the event of a fault, it automatically switches to traditional radial operation to ensure power supply continuity. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating the optimization method in this invention.

[0086] Figure 2 This is a diagram showing the internal topology and control block diagram of the flexible intelligent soft switch (SOP) in this invention.

[0087] Figure 3 This is a comparison chart of the load rate before and after the optimization of the distribution transformer in Example 2.

[0088] Figure 4 This is a comparison chart of the overload margin of the distribution transformer before and after optimization in Example 2.

[0089] Figure 5 This is a graph showing the voltage fluctuation curve of the distribution transformer T2 bus during the load transfer process of the flexible intelligent soft switch 1 in Example 2. Detailed Implementation

[0090] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0091] Example 1

[0092] This embodiment is implemented based on a power distribution substation in an industrial park in a certain city. The substation comprises three distribution transformers, all model S11-M-315 / 10, with a rated capacity of 315kVA, rated voltage of 10kV / 0.4kV, and rated frequency of 50Hz. Distribution transformer 1 mainly supplies power to the machining workshop, with a peak load of 280kW (load rate 89%) and an off-peak load of 50kW (load rate 16%). Distribution transformer 2 mainly supplies power to the office building and canteen, with a peak load of 180kW (load rate 57%) and an off-peak load of 30kW (load rate 10%). Distribution transformer 3 mainly supplies power to the warehouse and auxiliary facilities, with a peak load of 120kW (load rate 38%) and an off-peak load of 20kW (load rate 6%). The rated voltage of the low-voltage busbar of the distribution transformers is 380V (three-phase four-wire system), and the allowable voltage deviation range is -7% to +7% of the rated voltage (i.e., 353.4V to 406.6V). The distances between the distribution transformers are as follows: 80 meters between distribution transformer 1 and distribution transformer 2, 120 meters between distribution transformer 2 and distribution transformer 3, and 150 meters between distribution transformer 1 and distribution transformer 3.

[0093] This embodiment provides a distribution transformer group reconfiguration and optimization system based on intelligent soft switching, including the following modules:

[0094] Flexible intelligent soft-switching module (i.e., flexible intelligent soft-switching adopts a back-to-back bidirectional AC-DC-AC converter topology, including rectifier side, DC bus, inverter side, supercapacitor energy storage unit, control unit, and protection unit):

[0095] Flexible intelligent soft switch SOP-1: Deployed between distribution transformer 1 and distribution transformer 2, with a single-sided rated capacity of 100kVA (100kVA each for the rectifier and inverter sides). Specific hardware configuration is as follows:

[0096] Rectifier side: A three-phase voltage-type PWM rectifier is adopted. The AC side is connected to the low-voltage bus (380V) of the distribution transformer 1. The rated capacity of the rectifier side is 100kVA. The switching device is an IGBT module (model Infineon FF300R12ME4, rated voltage 1200V, rated current 300A, switching frequency 10kHz). Two IGBT modules are connected in parallel for each phase (a total of 6 IGBTs). The AC side filter inductor is 1.5mH (iron powder core inductor, rated current 200A).

[0097] DC bus: Rated voltage 750V, capacitance 10mF (using 10 1000μF / 900V film capacitors in parallel), bus voltage fluctuation range is 700-800V.

[0098] Inverter side: A three-phase voltage-source PWM inverter is adopted. The AC side is connected to the low-voltage bus (380V) of the distribution transformer 2. The rated capacity of the inverter side is 100kVA. The switching devices are the same as those on the rectifier side (6 IGBT modules). The AC side filter inductance is 1.5mH.

[0099] Supercapacitor energy storage unit: Connected in parallel to the DC bus, with a total capacity of 50F (using 10 5F / 400V supercapacitor modules connected in series, with a total withstand voltage of 4000V, stepped down to 750V operating voltage through a voltage equalization circuit), rated energy storage of 14kJ, providing instantaneous power compensation for 5 seconds with a peak power of 30kW. The supercapacitor is connected to the DC bus via a bidirectional DC-DC converter (Buck-Boost topology, switching frequency 20kHz).

[0100] Control unit: It adopts TI's TMS320F28379D dual-core DSP (200MHz clock frequency, 800MFLOPS floating-point operation capability), with a control algorithm cycle of 100μs. The communication interfaces include Ethernet interface (RJ45, communication rate 1Gbps), Modbus-RTU interface (RS485, baud rate 115200bps), and CAN bus interface (baud rate 500kbps).

[0101] Protection units include: overcurrent protection (current transformer sampling, protection threshold 150% of rated current, operating time <5ms), overvoltage protection (voltage sensor sampling, protection threshold 850V, operating time <10ms), undervoltage protection (protection threshold 650V), overtemperature protection (IGBT junction temperature sensor sampling, protection threshold 125℃), and short circuit protection (current change rate di / dt detection, protection threshold 10kA / μs).

[0102] Flexible intelligent soft switch SOP-2: Deployed between distribution transformer 2 and distribution transformer 3, with the same hardware configuration as SOP-1.

[0103] Data acquisition module:

[0104] The data acquisition module is deployed on the low-voltage side of each distribution transformer, responsible for collecting real-time operating parameters of the distribution transformers and uploading them to the edge computing nodes. The hardware configuration is as follows:

[0105] Smart meter: The Wasion DTSD341-MB3 three-phase smart energy meter (accuracy class 0.2S) is used. It is installed at the low-voltage side of the distribution transformer to measure three-phase voltage, current, active power, reactive power, power factor, energy and frequency. The sampling frequency is 1 second / time.

[0106] Temperature sensor: A PT100 platinum resistance temperature sensor is used, which is installed on the surface of the winding and core of the distribution transformer (3 sensors per distribution transformer). The measurement range is -50℃ to +200℃, the accuracy is ±0.5℃, and the sampling frequency is 10 seconds / time.

[0107] Power quality analyzer: A Fluke 435-II power quality analyzer was used and installed on the low-voltage side bus of the distribution transformer to measure voltage harmonic distortion rate (THD_U), current harmonic distortion rate (THD_I), voltage unbalance, and frequency deviation. The sampling frequency was 1 second / time.

[0108] Communication gateway: Utilizing Advantech's EKI-7659C industrial Ethernet switch, it integrates Modbus-RTU to Ethernet protocol conversion, aggregating data from smart meters, temperature sensors, and power quality analyzers before uploading it to the edge computing node via Ethernet. The communication gateway supports the IEC 61850 standard (Power System Communication Protocol), with a communication latency of <10ms.

[0109] The data acquisition module collects data once per second (the sampling frequency for high-frequency parameters such as current and voltage is 1kHz, but the frequency of uploading to the edge computing node is 1Hz). The data format is JSON, which includes timestamp, distribution transformer number, measurement point name, measured value, and unit.

[0110] Load forecasting module:

[0111] The load forecasting module is deployed on edge computing nodes and uses an LSTM (Long Short-Term Memory) neural network to perform rolling load forecasts for each distribution transformer for the next 15 minutes. It includes the following sub-modules:

[0112] The data preprocessing submodule receives historical load data (time window: the past 24 hours) from the data acquisition module, performs missing value imputation (using linear interpolation), outlier removal (based on the 3σ criterion, data points exceeding the mean ± 3 standard deviations are considered outliers), and data normalization (Min-Max normalization to the [0,1] interval). It also integrates with a weather forecast API (obtaining temperature, humidity, wind speed, and light intensity forecast data for the next 15 minutes from a municipal meteorological bureau).

[0113] Feature Engineering Submodule: Constructs input feature vectors, including historical load sequences (past 24 hours, time resolution 1 minute, total 1440 data points), weather forecast data (temperature, humidity, wind speed, light intensity, total 4 features), and time features (hour, day of the week, whether it is a holiday, total 3 features, using One-Hot encoding).

[0114] LSTM Neural Network: The network structure consists of a 3-layer LSTM (layer 1 with 128 neurons, layer 2 with 64 neurons, and layer 3 with 32 neurons) + a fully connected layer (16 neurons) + an output layer (15 neurons, corresponding to the predicted load value for the next 15 minutes). The activation function is tanh, the loss function is mean squared error (MSE), and the optimizer is Adam (learning rate 0.001). The model training data consists of historical load data from the past 6 months (approximately 2.6 million sample points), and the model parameters are saved after training.

[0115] The prediction execution submodule performs a prediction every minute, taking the current feature vector as input and outputting the load prediction value for the next 15 minutes and a 95% confidence interval (calculated based on the historical statistical distribution of prediction errors). The prediction results are then transmitted to the multi-objective optimization module.

[0116] In this embodiment, the load forecasting module has a mean absolute percentage error (MAPE) of 4.2% on the test set and a forecasting time of <50ms.

[0117] Parameter identification module:

[0118] The parameter identification module is deployed on edge computing nodes and uses the recursive least squares (RLS) method to determine the iron loss of each distribution transformer. Copper loss coefficient Real-time identification is performed. The comprehensive power loss model for distribution transformers is as follows:

[0119]

[0120] The module uses the load rate measured in real time by the data acquisition module. Calculate the actual losses based on the input and output power of the distribution transformer. The iron loss and copper loss coefficients are updated in real time using the recursive least squares method (the forgetting factor is set to 0.98 and updated once per minute).

[0121] The identified parameters are used to calculate the dynamic optimal economic operating range. Based on the aging factor, temperature correction factor, and harmonic distortion rate influence factor of the distribution transformer, the formula for calculating the dynamic efficiency range is as follows:

[0122] The lower limit of the optimal economic operating range is:

[0123] Equation (1)

[0124] The upper limit of the optimal economic operating range is:

[0125] Equation (2)

[0126] in, This is the lower limit of the baseline efficiency range. The upper limit of the baseline efficiency range, The aging factor is... For the service life of the distribution transformer, This is a temperature correction factor. The deviation between the measured temperature and the rated temperature of the distribution transformer. The harmonic distortion rate influence coefficient. For the total harmonic distortion rate of the distribution transformer current, detailed reference efficiency range. aging coefficient / year, temperature correction factor Harmonic distortion rate influence coefficient ;

[0127] That is, we get:

[0128]

[0129]

[0130] in, The service life of the distribution transformer. For winding temperature, The total harmonic distortion of the current;

[0131] In this embodiment, the distribution transformer 1 is identified. , The dynamic efficiency range is [30.5%, 79.2%]; the distribution transformer 2 was identified as follows. , The dynamic efficiency range is [30.0%, 79.4%].

[0132] Multi-objective optimization module:

[0133] The multi-objective optimization module is deployed on edge computing nodes. Based on load forecasting results, dynamic efficiency range, and dynamic overload margin, it constructs a multi-objective optimization model to solve for the optimal power allocation scheme for each flexible smart soft switch, including:

[0134] Decision variables: Power setpoints for each flexible intelligent soft switch (A positive value indicates power transmission from distribution transformer i to distribution transformer i+1, and a negative value indicates reverse transmission.)

[0135] Objective function: Minimize the total loss of the distribution transformer group Minimum overload margin variance Minimum voltage deviation The weighting coefficients are adaptively adjusted based on real-time electricity prices and load conditions.

[0136] Constraints: Dynamic efficiency range constraint, power flow constraint (based on distribution network power flow equations), capacity constraint, voltage constraint, minimum overload margin constraint;

[0137] In addition, the formula for calculating dynamic overload margin is:

[0138]

[0139] in, For distribution transformers exist Dynamic overload margin at any given time. This refers to the maximum allowable temperature of the transformer winding. For distribution transformer Temperature is measured in real time. Ambient temperature;

[0140] The solution algorithm employs an improved NSGA-III (Non-Dominated Sorting Genetic Algorithm-III), introducing a Q-Learning reinforcement learning mechanism to adaptively adjust the reference point weights. The algorithm outputs a Pareto optimal solution set, selecting the optimal solution based on real-time electricity prices and load tension. In this embodiment, the multi-objective optimization module performs optimization every 5 minutes (the optimization frequency can be adjusted according to load fluctuations), with each optimization taking approximately 50 seconds. The output optimal solution can reduce the total loss of the distribution transformer group by 28-32% and the overload margin variance by 75-85%.

[0141] Sensorless control module:

[0142] The seamless control module is deployed on edge computing nodes. Based on the optimization results of the multi-objective optimization module, it coordinates the flexible intelligent soft switch and energy storage unit to achieve seamless load transfer with voltage fluctuations of less than 2%, as detailed below:

[0143] Power ramp-up curve generation: Generate a power ramp-up curve based on the optimization results, with the initial ramp-up rate set to 10kW / s.

[0144] Adaptive ramp rate adjustment: Real-time monitoring of the bus voltage of the distribution transformer. When the voltage deviation exceeds 3%, the ramp rate is automatically reduced to 5kW / s. When the voltage deviation falls back to within 2%, the ramp rate is restored.

[0145] Supercapacitor Instantaneous Power Compensation: During power ramp-up, the output adjustment of the distribution transformer has a 5-10 second lag, resulting in an instantaneous power gap. The supercapacitor energy storage unit automatically charges and discharges by monitoring the DC bus voltage deviation to compensate for the instantaneous power gap. The control strategy is PI control (proportional coefficient). Integral coefficient );

[0146] In this embodiment, the maximum voltage fluctuation peak-to-peak value during the seamless load transfer process is less than 5% and the duration is less than 1 second, so the user side is basically unaware of it.

[0147] Backfeed power control module:

[0148] The reverse power control module is deployed on edge computing nodes to monitor the power direction of each distribution transformer in real time. When reverse power transmission (power flowing from the low-voltage side to the high-voltage side) is detected, the module will respond accordingly. When this occurs, a reverse power transfer command is automatically triggered.

[0149] Based on the load rate and overload margin of adjacent distribution transformers, the module prioritizes injecting backfeed power into the distribution transformer with the highest load rate and overload margin that meets safety requirements (≥30%), while also considering the capacity constraints of the flexible intelligent soft switch. The backfeed power transfer process adopts the same ramp control strategy as the sensorless control module, keeping voltage fluctuations within ±2%.

[0150] In this embodiment, the reverse power control function can effectively solve the problem of reverse power transmission of the distribution transformer caused by distributed photovoltaic access, avoid the bus voltage exceeding the limit, and reduce the photovoltaic curtailment rate from 15.3% to 2.1%.

[0151] Fault diagnosis and degradation control module:

[0152] The fault diagnosis and degradation control module is deployed on the edge computing node to monitor the operating status of the flexible intelligent soft switch in real time (DC bus voltage, IGBT junction temperature, AC side current harmonic distortion rate, AC side short circuit current). When an abnormality is detected, fault diagnosis is triggered.

[0153] Fault confirmation: The fault signal is confirmed a second time (the fault signal is detected in 3 consecutive sampling cycles) to confirm that the fault actually exists.

[0154] Fault Lockout: The flexible intelligent soft switch is immediately locked out upon fault confirmation (turning off all IGBT switching transistors and disconnecting the AC side contactor), with a lockout time of <20ms.

[0155] Degraded control mode: After the faulty flexible intelligent soft switch is locked, the system automatically switches to the degraded control mode, restores the traditional radial operation topology, and the optimization objective is downgraded from "minimizing the total loss of the distribution transformer group" to "avoiding distribution transformer overload and voltage over-limit", only optimizing the non-faulty flexible intelligent soft switch.

[0156] Fault recovery: The system monitors the fault status in real time. After the fault is repaired and confirmed by the operator, the system automatically returns to the normal optimization mode and re-enables the fault-flexible intelligent soft switch by adopting a smooth transition strategy (reducing the ramp rate to 5kW / s).

[0157] In this embodiment, the fault diagnosis and degradation control functions passed the test, the fault response time was <20ms, and the power supply continuity was 100% under degradation control mode.

[0158] It should be noted that the core of the fault diagnosis and degradation control module lies in establishing local virtual protection logic for sandbox verification. Before the system executes a specific fault degradation strategy, the proposed degradation scheme is simulated and verified offline or online through the built-in virtual protection logic unit. This logic reads real-time data of the current power grid operating conditions and builds an equivalent circuit model in the local processor to simulate the electrical response after executing the safe backoff path, including voltage drop and current surge. If the simulation results show that there is a risk of overload or protection maloperation, the system will automatically adjust the backoff curve parameters until it passes the sandbox verification. Through the verification of the local virtual protection logic, it is ensured that when the system detects a fatal abnormal signal, it first calls the energy storage unit to smoothly reduce the transmission power of that branch to zero before triggering the contactor action and IGBT blocking logic, effectively avoiding grid operation overvoltage caused by sudden disconnection.

[0159] Example 2

[0160] This embodiment applies to the system of Embodiment 1 and conducts actual operation tests in a distribution transformer area in an industrial park in a certain city. The distribution transformer area includes three 315kVA oil-immersed distribution transformers (T1, T2, and T3), which supply power to the machining workshop, office building and canteen, and warehouse area, respectively. A flexible intelligent soft switch SOP-1 is deployed between distribution transformers T1 and T2, and a flexible intelligent soft switch SOP-2 is deployed between T2 and T3, forming a flexible interconnection topology. The test time is from 14:00 to 16:30 on August 15, 2023, the weather is sunny, the ambient temperature is 32℃, and the humidity is 65%.

[0161] At the start of the test, the distribution transformer group was operating as follows:

[0162] Distribution transformer T1: Output power 280kW, load rate 88.9% (near full load), winding temperature 92℃, bus voltage 375V. High-power equipment in the machining workshop is starting up simultaneously, causing the load to continuously increase, posing a risk of overload.

[0163] Distribution transformer T2: Output power 120kW, load rate 38.1% (light load), winding temperature 68℃, bus voltage 385V. The office building air conditioning load is stable, the canteen's electricity usage is about to end, iron loss is too high, and operating efficiency is low.

[0164] Distribution transformer T3: Output power 95kW, load rate 30.2% (light load), winding temperature 65℃, bus voltage 383V. The lighting and temperature control equipment in the storage area are under light load, close to the lower limit of the low efficiency range.

[0165] This shows that:

[0166] 1) The load rate of distribution transformer T1 is 88.9%, which is close to full load. The winding temperature is 92℃, which is close to the warning value. The dynamic overload margin is only 17.8% (lower than the minimum overload margin constraint of 30%), which poses a risk of overload tripping.

[0167] 2) The load rates of distribution transformers T2 and T3 are 38.1% and 30.2% respectively, which are far from the center of the dynamic efficiency range. The iron loss ratio is too high, resulting in low operating efficiency.

[0168] 3) The overload margins of the three distribution transformers are 17.8%, 50.7%, and 54.8% respectively, which are severely unevenly distributed (variance 274.5), making them unable to cope with sudden load fluctuations;

[0169] like Figure 1-2 As shown, the optimization process is as follows:

[0170] S1, System Initialization

[0171] S101, Flexible Intelligent Soft Switch Initialization:

[0172] SOP-1 is deployed between the low-voltage sides of distribution transformers T1 and T2, and SOP-2 is deployed between the low-voltage sides of T2 and T3.

[0173] Both SOP-1 and SOP-2 adopt a back-to-back bidirectional AC-DC-AC converter topology, with a supercapacitor energy storage unit connected in parallel on the DC side (rated energy storage 140kJ, rated capacity 100kVA).

[0174] Power-on self-test: The DC bus voltage is established to 750V by the pre-charge circuit, the IGBT junction temperature is normal (55℃), the insulation resistance is qualified (>100MΩ), the AC contactors on the rectifier side and inverter side are closed, and the system enters the ready state.

[0175] The supercapacitor energy storage unit has an initial state of charge (SOC) of 85% and a stable DC bus voltage of 750V.

[0176] S102. Initialization of each module:

[0177] Data acquisition module: Completes communication handshake with each distribution transformer, with a sampling period set to 1 second.

[0178] Load forecasting module: Loads LSTM neural network weight parameters. Input dimensions include historical load curve (past 24 hours), weather forecast (sunny, temperature 28℃), and date type (weekday).

[0179] Parameter identification module: Initialize the recursive least squares covariance matrix P(0) as an identity matrix, and the forgetting factor λ=0.98.

[0180] Multi-objective optimization module: Loads the NSGA-III algorithm population parameters (population size 200, crossover probability 0.9, mutation probability 0.1), and loads the Q-value table for the Q-Learning algorithm.

[0181] The sensorless control module, the reverse power control module, and the fault diagnosis and degradation control module are ready.

[0182] Step S2: Load Forecasting (The load forecasting module uses an LSTM neural network to perform rolling load forecasts for each distribution transformer over the next 15 minutes)

[0183] S201, Input Features:

[0184] Historical load curve: Power data of each distribution transformer over the past 24 hours (sampling interval of 5 minutes, total of 288 data points).

[0185] Weather forecast: Sunny from 14:00 to 14:15, temperature 28℃, humidity 62%.

[0186] Date type: Weekday (coded as 1), not public holidays.

[0187] S202, LSTM neural network structure: input layer dimension 288 (historical load curve) + 3 (weather features) + 1 (date type) = 292, hidden layer contains 2 layers of LSTM units (128 neurons per layer), output layer is a fully connected layer (outputs the load prediction value of each distribution transformer in the next 15 minutes).

[0188] S202. Prediction Results (Predicted Load Values ​​for Each Distribution Transformer at 14:15):

[0189] Distribution transformer T1: Predicted load 318kW, load rate There is a risk of overload. Reason: Three high-power CNC machine tools (50kW each, totaling 3) in the machining workshop will be started successively at 14:10, causing the load to continuously increase.

[0190] Distribution transformer T2: Predicted load 102kW, load rate Reason: The office building's air conditioning load is stable (80kW), and the cafeteria's electricity usage is about to end (reducing from 40kW to 22kW).

[0191] Distribution transformer T3: Predicted load 109kW, load rate Reason: Increased lighting load in the storage area (supplementary lighting before dusk), and activation of temperature control equipment.

[0192] The load forecasting module transmits the forecast results to the multi-objective optimization module, and simultaneously triggers the parameter identification module and the overload margin calculation module to perform real-time calculations.

[0193] S3. Parameter Identification and Dynamic Efficiency Range Calculation

[0194] S301, Parameter Identification Results (Based on data from the past hour):

[0195] T1 0.850 0.0105 8 +12 4.2% T2 0.820 0.0098 6 +8 3.5% T3 0.810 0.0095 5 +5 2.8%

[0196] S302, Dynamic efficiency interval calculation:

[0197] Based on the results of the lower limit of the optimal economic operating range - Equation (1) and the upper limit of the optimal economic operating range - Equation (2) in Example 1, it can be concluded that:

[0198] Distribution transformer T1:

[0199] Lower limit:

[0200] Maximum:

[0201] The current load rate of 88.9% exceeds the upper limit of the dynamic efficiency range of 79.9%, resulting in excessive copper loss, low operating efficiency, and a risk of overload.

[0202] Distribution transformer T2:

[0203] Lower limit:

[0204] Maximum:

[0205] The current load factor of 38.1% is within the range [ Within the range, but biased towards the lower limit, with iron loss accounting for too high a proportion;

[0206] Distribution transformer T3:

[0207] Lower limit:

[0208] Maximum:

[0209] The current load rate of 30.2% is lower than the lower limit of the range of 31.2%, which is close to the low efficiency range. The iron loss ratio is too high. This dynamic efficiency range is then transmitted to the multi-objective optimization module.

[0210] S4, Dynamic Overload Margin Calculation

[0211] A thermal model of distribution transformers was established based on the IEEE C57.91 standard, and the dynamic overload margin of each distribution transformer was calculated in real time based on equation (3):

[0212]

[0213] in This refers to the maximum allowable temperature of the distribution transformer winding (oil-immersed distribution transformer). For distribution transformer Temperature is measured in real time. Ambient temperature;

[0214] Calculation results of dynamic overload margin for each distribution transformer:

[0215] Distribution transformer T1:

[0216]

[0217] Approaching the minimum overload margin constraint of 30%, there is a risk of overload.

[0218] Distribution transformer T2:

[0219]

[0220] Meets safety requirements;

[0221] Distribution transformer T3:

[0222]

[0223] Meets safety requirements;

[0224] The overload margin variance is:

[0225]

[0226] Substituting, we get:

[0227]

[0228]

[0229] The overload margin variance is as high as 173.0, indicating that the overload margin distribution of each distribution transformer is seriously uneven. The overload margin of distribution transformer T1 is seriously insufficient, while the overload margins of T2 and T3 are abundant. This result is transmitted to the multi-objective optimization module.

[0230] S5. Construct a multi-objective optimization model

[0231] The objective functions include minimizing the total loss of the distribution transformer group, minimizing the overload margin variance, and minimizing the voltage deviation. The constraints include dynamic efficiency range constraints, power flow constraints, capacity constraints, voltage constraints, and overload margin constraints.

[0232] Decision variables: [ ], range [-100, +100] kW (positive values ​​indicate power transmission from the former to the latter, negative values ​​indicate reverse transmission);

[0233] Objective function:

[0234] 1) Minimize the total loss of the distribution transformer group:

[0235]

[0236] in, For distribution transformers At load rate The overall power loss is as follows. For the loss weighting coefficient of flexible intelligent soft switching, For flexible intelligent soft switches Power loss (including converter switching loss, conduction loss and line loss).

[0237] The formula for calculating the loss of flexible intelligent soft switching is as follows:

[0238]

[0239] in, The switching loss factor is... The conduction loss coefficient is... For flexible intelligent soft switches The transmission power;

[0240] 2) Minimum overload margin variance:

[0241]

[0242] in, This represents the average overload margin of each distribution transformer.

[0243] 3) Minimal voltage deviation:

[0244]

[0245] in, For the low-voltage side bus voltage of distribution transformer i, Rated voltage;

[0246] Constraints:

[0247] 1) Dynamic efficiency interval constraints:

[0248] To ensure that the load rate of each distribution transformer is within the optimal economic operating range;

[0249] 2) Current constraints:

[0250]

[0251]

[0252] in, The output active power of distribution transformer i For the active power of the load, Power is transmitted to the flexible intelligent soft switch (positive for inflow and negative for outflow). For line current, The equivalent line resistance from the distribution transformer to the user, The equivalent line reactance from the distribution transformer to the user, For the low-voltage side bus voltage of the distribution transformer, This refers to the voltage at the end load node. Power factor;

[0253] 3) Capacity constraints:

[0254] ;in, Rated capacity;

[0255] 4) Voltage constraint:

[0256] ,Right now ;

[0257] 5) Minimum overload margin constraint:

[0258] Ensure that each distribution transformer has sufficient overload capacity.

[0259] The objective function of the multi-objective optimization model also includes a sensitivity evaluation function. This is used to quantify the impact of parameter fluctuations on system security. Specifically, the sensitivity assessment formula is constructed as follows:

[0260]

[0261] in, For fixed parameters Probability of protection malfunction The sensitivity coefficient is used to quantify the impact of parameter fluctuations on system safety. Introducing this function into optimization ensures that the optimization strategy improves economy while maintaining the reliability of system operation.

[0262] After the multi-objective optimization model is constructed, it is passed to step S6 for solution.

[0263] S6. Solve the multi-objective optimization model

[0264] The multi-objective optimization model is solved using a multi-objective optimization algorithm to obtain the power setting value of each flexible intelligent soft switch. The multi-objective optimization model constructed in step S5 is then solved using an improved NSGA-III algorithm.

[0265] Weighting coefficients are adaptively adjusted:

[0266] At 14:00, the real-time electricity price was 0.82 yuan / kWh (industrial electricity price), and the load factor was:

[0267]

[0268] Since the dynamic overload margin of distribution transformer T1 is 30.4%, which just meets the minimum overload margin constraint of 30%, but considering that the load forecast for the next 15 minutes shows that T1 will be overloaded to 100.9%, the system will adjust the load stress to... (Higher);

[0269] The Q-Learning algorithm selects the objective function weight coefficients based on the current state (flat electricity price, high load pressure). = [0.2, 0.5, 0.3];

[0270] Target weight of total loss of distribution transformer group (Reduced weighting; given the current flat electricity price, energy efficiency is given lower priority).

[0271] Overload margin variance target weight (Increase weighting; given high load pressure, prioritize balancing the overload margin of each distribution transformer).

[0272] Voltage deviation target weight (Maintain a moderate weighting to ensure voltage quality);

[0273] Solving using the NSGA-III algorithm:

[0274] The population size is 200, the crossover probability is 0.9, and the mutation probability is 0.1. Simulated binary crossover (SBX) and polynomial mutation operators are used. The iteration reaches the 87th generation and meets the termination condition (the change in the hypervolume index is <0.01 for 10 consecutive generations), outputting the Pareto optimal solution set (containing 32 non-dominated solutions).

[0275] Based on the weighting coefficients [ Given [0.2, 0.5, 0.3], calculate the weighted objective function value:

[0276]

[0277] The optimal solution is selected from the Pareto optimal solution set with the minimum weighted objective function value: SOP-1 power setting value +70kW (transferring 70kW load from distribution transformer T1 to T2), SOP-2 power setting value +30kW (transferring 30kW load from distribution transformer T2 to T3). The optimization algorithm takes about 48 seconds to solve.

[0278] Optimization effect expected (e.g.) Figures 3-5 As shown in the attached diagram, the abbreviation for distribution transformer is "distribution transformer," and in... Figure 5 During the load transfer process in SOP-1 from 14:01:05 to 14:01:13, when the voltage of the distribution transformer T2 bus dropped to 368V at 14:01:11 (voltage deviation reached 3.16%, exceeding the 3% preset threshold), the system automatically triggered adaptive ramp rate adjustment, reducing the ramp rate from 10kW / s to 5kW / s, causing the voltage of the distribution transformer T2 bus to quickly recover and stabilize at 373V (voltage deviation reduced to 1.84%). The entire load transfer process took 8 seconds, with the maximum voltage fluctuation peak-to-peak value less than 5% and the duration less than 1 second, achieving a virtually imperceptible load transfer on the user side.

[0279] After implementing the optimization plan, the expected changes in the operating parameters of each distribution transformer are as follows:

[0280] T1 280 210 88.9% 66.7% 30.4% 42.1% T2 120 160 38.1% 50.8% 56.5% 51.3% T3 95 125 30.2% 39.7% 59.8% 54.2%

[0281] The load factor of distribution transformer T1 decreased from 88.9% to 66.7%, entering the middle position of the dynamic efficiency range [32.3%, 79.9%], and the dynamic overload margin increased from 30.4% to 42.1%, significantly reducing the risk of overload.

[0282] The load factor of distribution transformer T2 increased from 38.1% to 50.8%, which is closer to the center of the dynamic efficiency range, and the operating efficiency improved by about 8%.

[0283] The load factor of distribution transformer T3 increased from 30.2% to 39.7%, entering the dynamic efficiency range [31.2%, 79.9%], far from the lower limit of the low efficiency range, and the proportion of iron loss decreased;

[0284] The overload margin variance decreased from 173.0 to 49.5, a reduction of 71.4%, and the overload margin distribution of each distribution transformer was significantly more balanced (42.1%, 51.3%, and 54.2%).

[0285] The multi-objective optimization module will select the optimal solution. The data is transmitted to the sensorless control module, and then proceeds to step S7 for execution.

[0286] S7, Seamless Control Execution

[0287] Based on the optimization results, a power regulation curve is generated. The flexible intelligent soft switch adjusts the output power according to the regulation curve, and the energy storage unit compensates for instantaneous power gaps, realizing seamless load transfer. The seamless control module coordinates the flexible intelligent soft switch and the energy storage unit based on the optimization results to achieve seamless load transfer with voltage fluctuations of less than 2%.

[0288] Power ramp-up curve generation:

[0289] The sensorless control module uses the optimal solution [ [and current state] Generate a power ramp-up curve:

[0290] SOP-1: Climb from 0kW to +70kW, initial climbing rate 10kW / s, estimated climbing time 7 seconds;

[0291] SOP-2: Climbing from 0kW to +30kW, initial climbing rate 10kW / s, estimated climbing time 3 seconds;

[0292] At 14:01:05, the sensorless control module simultaneously sends power ramp-up commands to SOP-1 and SOP-2.

[0293] Adaptive climb rate adjustment:

[0294] SOP-2 successfully completed the ramp-up within 3 seconds (from 14:01:05 to 14:01:08). During the ramp-up, the voltage of the distribution transformer T2 busbar dropped from 385V to 380V (voltage deviation). The voltage on the T3 bus of the distribution transformer dropped from 383V to 379V (voltage deviation). The voltage fluctuation is less than 2%, which meets the requirements for sensorless control.

[0295] During the ramp-up process in SOP-1, when the power increases from 0kW to 60kW (at 14:01:11, 6 seconds into the ramp-up), the voltage on the distribution transformer T2 bus drops to 368V, a voltage deviation from the rated voltage of 380V. If the voltage deviation exceeds a preset threshold of 3% (in this embodiment, the preset threshold is set to 3%), the sensorless control module detects the excessive voltage deviation and immediately triggers adaptive ramp rate adjustment: reducing the ramp rate of SOP-1 from 10kW / s to 5kW / s. After the ramp rate is reduced, the voltage on the T2 bus of the distribution transformer rises back to 373V, and the voltage deviation decreases to [missing value]. This satisfies the voltage constraint.

[0296] At 14:01:13, SOP-1 completed its ramp-up and stabilized its output at +70kW. The total load transfer process took 8 seconds (from 14:01:05 to 14:01:13).

[0297] Supercapacitor instantaneous power compensation:

[0298] During the ramp-up process of SOP-1, the output power of distribution transformer T1 needs to be reduced (from 280kW to 210kW), but the mechanical inertia of the distribution transformer causes a 5-10 second lag in output adjustment. The supercapacitor energy storage unit connected in parallel on the DC side of SOP-1 monitors the DC bus voltage deviation in real time. When the DC bus voltage drops from 750V to 730V (voltage deviation 2.67%), the supercapacitor automatically discharges to compensate for the instantaneous power gap.

[0299] Between 14:01:05 and 14:01:13, the actual output power of distribution transformer T1 decreased only from 280kW to 250kW (regulation lag of 40kW), resulting in an instantaneous power shortfall of approximately 40kW. The supercapacitor accumulated a discharge energy of 13.3kJ, accounting for 9.5% of the rated energy storage of 140kJ. After the supercapacitor compensated for the instantaneous power shortfall, the voltage fluctuation of the distribution transformer T2 bus was controlled within ±2% (except for the instantaneous 3.16% at 14:01:11 that triggered adaptive adjustment, the voltage deviation was <2% at all other times), which was basically imperceptible to the user side.

[0300] At 14:01:20, the output of distribution transformer T1 was adjusted to 210kW, the DC bus voltage was restored to 750V, and the supercapacitor stopped discharging.

[0301] Seamless control effect:

[0302] After the load transfer was completed, the operating status of each distribution transformer improved significantly:

[0303] The load factor of distribution transformer T1 decreased from 88.9% to 66.7%, the expected winding temperature decreased from 92℃ to about 78℃, and the dynamic overload margin increased from 30.4% to 42.1%, significantly reducing the risk of overload.

[0304] The load factor of distribution transformer T2 increased from 38.1% to 50.8%, which is closer to the center of the dynamic efficiency range, and the operating efficiency improved by about 8%.

[0305] The load factor of distribution transformer T3 increased from 30.2% to 39.7%, entering the dynamic efficiency range and moving away from the lower limit of the low efficiency range, thus reducing the proportion of iron loss.

[0306] The overload margin distribution is significantly balanced (42.1%, 51.3%, 54.2%), and all three distribution transformers have sufficient overload capacity to cope with sudden load fluctuations.

[0307] The maximum voltage fluctuation during the load transfer process was 3.16% (occurring only momentarily at 14:01:11, immediately triggering adaptive adjustment), while the voltage deviation at other times was less than 2%, which was basically imperceptible to the user.

[0308] The sensorless control module transmits the load transfer completion signal to the data acquisition module, and the system enters the steady-state operation monitoring phase.

[0309] S8, Reverse Power Control

[0310] At 14:30:00, the weather was clear, with the solar irradiance reaching a peak of 950W / m², and the output of the 100kW distributed photovoltaic system connected to distribution transformer T2 reached a peak of 95kW. Simultaneously, the office building's air conditioning load decreased, the cafeteria's electricity usage ended, and the local load on distribution transformer T2 decreased to 32kW.

[0311] Backfeed power detection:

[0312] The system monitors the power direction of each distribution transformer in real time. When power backfeed is detected, the backfeed power is injected into an adjacent distribution transformer via a flexible intelligent switch. The backfeed power control module monitors the power direction of each distribution transformer in real time. The data acquisition module measures the operating parameters of distribution transformer T2.

[0313] Photovoltaic output: ;

[0314] Local load: ;

[0315] SOP-1 transmission power: (Transmission from T1 to T2);

[0316] Actual output power of distribution transformer T2:

[0317] ;

[0318] Since the photovoltaic output of 95kW is greater than the sum of the local load of 32kW and the 70kW consumed by SOP-1, there is a 7kW surplus power in distribution transformer T2. If backfeed power control is not implemented, this surplus power will be fed back to the medium voltage side through distribution transformer T2, which may lead to an increase in bus voltage and photovoltaic curtailment.

[0319] The backfeed power control module determined that there is a potential backfeed risk in distribution transformer T2 (surplus power 7kW).

[0320] Reverse power transfer scheme generation:

[0321] The reverse power control module queries the operating status of adjacent distribution transformers (T1 and T3):

[0322] Distribution transformer T1: Load rate 66.7%, dynamic overload margin 42.1%, still has capacity to accept loads, but the load rate is already high.

[0323] Distribution transformer T3: Load rate 39.7%, dynamic overload margin 54.2%, with sufficient capacity to accommodate loads.

[0324] The reverse power control module decides to prioritize injecting reverse power into distribution transformer T3 (where the load rate is low and the overload margin meets safety requirements), generating a reverse power transfer command to adjust the SOP-2 power setpoint from +30kW to... (Increase transmission power by 7kW);

[0325] Seamless control execution:

[0326] At 14:30:05, the backfeed power control module sends a power adjustment command to the sensorless control module. The sensorless control module employs the same ramp-up control strategy as in step S7, smoothly ramping the SOP-2 power from +30kW to +37kW. The ramp-up process takes approximately 1.4 seconds (ramp-up rate 5kW / s, from 14:30:05 to 14:30:06.4), during which the voltage on the distribution transformer T2 bus rises from 373V to 376V (voltage deviation). The voltage on the T3 bus of the distribution transformer dropped from 379V to 377V (voltage deviation). The voltage fluctuation is less than 2%, which meets the requirements for sensorless control.

[0327] Backfeed power control effect:

[0328] At 14:30:06.4, SOP-2 completed its ramp-up to +37kW, and the excess power of distribution transformer T2 was absorbed by SOP-2, eliminating the backfeeding phenomenon. Operating status after the backfeeding power transfer:

[0329] Distribution transformer T2: Output power 7kW, photovoltaic output 95kW is fully absorbed by local load 32kW, SOP-1 transmission 70kW, and distribution transformer T2's own losses and line losses of approximately 5kW, with no backfeed power. Bus voltage 376V, voltage deviation 1.05%, meeting voltage constraints.

[0330] Distribution transformer T3: Output power increased from 125kW to 132kW, load rate increased from 39.7% to 41.9%, still within the dynamic efficiency range [31.2%, 79.7%], and dynamic overload margin decreased from 54.2% to approximately 52.1%, meeting safety requirements.

[0331] After the backfeed power was eliminated, the bus voltage of distribution transformer T2 stabilized at 376V from the previously possible overvoltage state (where photovoltaic backfeed caused the bus voltage to rise above 390V), thus preventing voltage exceedance. The photovoltaic curtailment rate decreased from the previously possible... The photovoltaic utilization rate has been reduced to 0%, resulting in a significant increase in photovoltaic utilization.

[0332] Step S9: Fault Diagnosis and Degradation Control

[0333] At 15:45:30, an anomaly occurred during the operation of SOP-1: the DC bus voltage suddenly dropped from 750V to 680V, ​​which is below the undervoltage protection threshold of 700V.

[0334] Fault diagnosis:

[0335] The system monitors the operational status of the flexible intelligent soft switch in real time. When an anomaly is detected, fault diagnosis is triggered. After confirming the fault, the flexible intelligent soft switch is locked and switched to degraded control mode. The fault diagnosis and degraded control module monitors the operational status of the flexible intelligent soft switch in real time. The SOP-1 control unit monitors the DC bus voltage in real time. If the DC bus voltage is detected to be below 700V for three consecutive sampling cycles (15:45:30, 15:45:31, 15:45:32), the fault is confirmed to exist.

[0336] The fault diagnosis module analyzes the cause of the fault based on the SOP-1 operating data:

[0337] A. The three-phase voltage and current on the rectifier side and inverter side are normal, with no short circuit or overcurrent.

[0338] B. The voltage across the DC bus capacitor is unbalanced (upper bridge arm capacitor voltage 360V, lower bridge arm capacitor voltage 320V, total 680V), suggesting aging of the DC bus capacitor or a fault in the voltage equalization circuit.

[0339] C. The IGBT junction temperature is normal (68℃), with no overheating.

[0340] D. The supercapacitor's SOC is 45%, which is within the normal range.

[0341] The fault diagnosis module determined that aging of the SOP-1 DC bus capacitor or a fault in the voltage equalization circuit could lead to DC bus voltage instability, affecting power transmission stability. To ensure the power supply safety of the distribution transformer group, the fault diagnosis module triggered a lockout command.

[0342] Fault lockout:

[0343] At 15:45:33, the fault diagnosis and degradation control module sends a lockout command to SOP-1:

[0344] 1. Immediately block all IGBT drive signals on the rectifier and inverter sides of SOP-1 to stop power transmission;

[0345] 2. Disconnect the AC contactors on the rectifier side and inverter side of SOP-1 to electrically isolate SOP-1 from distribution transformers T1 and T2;

[0346] 3. The DC bus voltage is discharged from 680V to a safe voltage below 50V through the discharge resistor (discharge time is about 15 seconds).

[0347] At 15:45:48, SOP-1 completed the interlocking, the DC bus voltage dropped to 42V, and the system entered a safe shutdown state.

[0348] Degradation control mode switching:

[0349] After SOP-1 is locked, the flexible interconnection capability between distribution transformers T1 and T2 is lost, and the system automatically switches to degraded control mode.

[0350] Degradation control strategy:

[0351] 1. Topology Degradation: Distribution transformers T1 and T2 are restored to the traditional radial operating topology, each supplying power independently, and load transfer is no longer performed through SOP-1;

[0352] 2. Optimization target downgrade: The optimization target is downgraded from "minimum total loss of distribution transformer group, minimum overload margin variance, and minimum voltage deviation" to "avoid distribution transformer overload and voltage exceeding limits" (single target optimization, only ensuring power supply safety).

[0353] 3. Optimize only the non-faulty branch: SOP-2 (between T2 and T3) is still operating normally. The system only optimizes the power distribution of SOP-2, and the SOP-1 branch is restored to traditional operation.

[0354] The degradation control module re-executes the optimization calculation (at 15:45:50):

[0355] Distribution transformer T1: Current load rate is 66.7%, dynamic overload margin is 42.1%, which meets safety requirements and requires no adjustment.

[0356] Distribution transformer T2: Current load factor is 50.8%. Due to SOP-1 lockout, it loses 70kW power support from T1, and the load factor is expected to decrease to [missing value]. It enters a period of low efficiency;

[0357] Distribution transformer T3: Current load rate is 41.9%, dynamic overload margin is 52.1%, which meets safety requirements.

[0358] Degradation control module decision: Adjust the SOP-2 power setting from +37kW to +55kW (transfer 55kW load from T2 to T3) to improve the load rate of distribution transformer T2 and prevent it from entering the low efficiency range.

[0359] At 15:45:55, SOP-2 completed power adjustment, increasing the load factor of distribution transformer T2 from 28.6% to [missing value]. It has entered the vicinity of the lower limit of the dynamic efficiency range, basically meeting the operational requirements.

[0360] Degradation control effect:

[0361] After SOP-1 fault lockout, the system ensures power supply continuity through degraded control mode:

[0362] No overload or voltage exceeding the limit occurred in distribution transformers T1, T2, and T3, and the power supply continuity was 100%.

[0363] The load rate of distribution transformer T2 decreased from 50.8% before the fault to 33.3%, and the operating efficiency decreased slightly, but still met the basic requirements.

[0364] The load factor of distribution transformer T3 increased from 41.9% to 50.8%, improving its operating efficiency;

[0365] The total loss of the distribution transformer group increased from 10.2kW before the fault to 11.5kW (an increase of 12.7%, due to the loss of the load balancing capability of SOP-1), but it was still better than the 15.1kW before optimization.

[0366] Fault recovery:

[0367] At 15:55:00, maintenance personnel arrived on site and replaced the faulty capacitor module SOP-1. At 16:10:00, SOP-1 completed repairs. After power-on self-test, the DC bus voltage returned to 750V. The fault diagnosis and degradation control module detected that SOP-1 had returned to normal, automatically released the lockout, and the system returned to normal optimization mode.

[0368] Summary of the performance of this embodiment:

[0369] During the test period (14:00 to 16:30, a total of 2.5 hours), the running results are as follows:

[0370] 1) The total loss of the distribution transformer group was significantly reduced:

[0371] Total losses of the distribution transformer group before optimization: 15.07 kWh (cumulative power consumption in 2.5 hours);

[0372] The total loss of the distribution transformer group after optimization is 10.81 kWh (cumulative power consumption over 2.5 hours, including degraded control during SOP-1 fault).

[0373] Loss reduction rate: ;

[0374] 2) The overload margin distribution is significantly balanced:

[0375] Unoptimized overload margin variance: 173.0;

[0376] Optimized overload margin variance: 49.5;

[0377] Variance reduction rate: 71.4%;

[0378] After optimization, the overload margin distribution of each distribution transformer is balanced (42.1%, 51.3%, 54.2%), and all meet the minimum overload margin constraint of 30%.

[0379] 3) Significantly improved voltage quality:

[0380] Before optimization, the voltage of the distribution transformer T1 bus was 375V (voltage deviation 1.3%), the voltage of T2 bus was 385V (voltage deviation 1.3%), and the voltage of T3 bus was 383V (voltage deviation 0.8%).

[0381] After optimization, the voltage of the distribution transformer T1 bus is 378V (voltage deviation 0.5%), the voltage of T2 bus is 376V (voltage deviation 1.05%), and the voltage of T3 bus is 377V (voltage deviation 0.79%).

[0382] Sum of squared voltage deviations from Reduce to , a decrease of 51.4%;

[0383] 4) Improved photovoltaic utilization rate:

[0384] From 14:30 to 15:30, the 100kW photovoltaic system connected to distribution transformer T2 generated a cumulative power of 82kWh.

[0385] Before optimization (without backfeed power control): the photovoltaic curtailment rate was approximately 7.4%, and the curtailed power was 6.1 kWh;

[0386] After optimization (with reverse power control enabled): PV curtailment rate 0%, curtailed power 0 kWh;

[0387] The photovoltaic utilization rate is increased to 100%, saving approximately [amount missing] in electricity costs. Yuan;

[0388] 5) Seamless load transfer effect:

[0389] From 14:01:05 to 14:01:13, SOP-1 completed the transfer of 70kW load, with a maximum voltage fluctuation of 3.16% (occurring only momentarily at 14:01:11, immediately triggering adaptive adjustment), and voltage deviations at other times were all <2%.

[0390] During the period from 14:30:05 to 14:30:06.4, SOP-2 completed the transfer of 7kW backfeed power with a maximum voltage fluctuation of 1.05%, which meets the requirements of sensorless control (<2%).

[0391] No complaints about voltage fluctuations were received from users, and the power supply quality satisfaction rate was 100%.

[0392] 6) Fault response capability:

[0393] At 15:45:33, SOP-1 detected a DC bus voltage fault. The fault diagnosis and degradation control module completed fault confirmation, blocking command transmission, and degradation control mode switching within 3 seconds.

[0394] During the SOP-1 fault lockout period (15:45:33 to 16:10:00), the system ensures 100% power supply continuity through the degradation control mode, with no overload or voltage exceeding the limit.

[0395] After the fault is resolved, the system will automatically unlock and return to normal optimization mode without manual intervention.

[0396] In summary, this embodiment achieves seamless topology reconfiguration and economical operation optimization of the distribution transformer group by deploying flexible intelligent soft switches within the distribution transformer group and combining techniques such as load forecasting, parameter identification, dynamic overload margin calculation, multi-objective optimization, sensorless control, reverse power control, fault diagnosis, and degradation control. Compared to before optimization, the total loss of the distribution transformer group is reduced by 28.3%, the overload margin variance is reduced by 71.4%, the sum of squared voltage deviations is reduced by 51.4%, photovoltaic utilization is increased to 100%, and power supply continuity is 100%. This method significantly improves the operating efficiency, power supply reliability, and economy of the distribution transformer group, demonstrating significant practical value.

[0397] It should be noted that the structure described in this invention can be implemented in many different forms and is not limited to the embodiments described. Any equivalent transformations made by those skilled in the art based on the description and drawings of this invention, or direct or indirect applications in other related technical fields, are included within the protection scope of this invention.

Claims

1. A method for reconfiguration and optimization of distribution transformer groups based on intelligent soft switching, characterized in that, Includes the following steps: S1. Deploy a flexible intelligent soft switch on the low-voltage side of an adjacent distribution transformer. The flexible intelligent soft switch adopts a bidirectional converter topology and has a parallel energy storage unit on the DC side. S2. Predict the future load of each distribution transformer; S3. Real-time identification of loss parameters of each distribution transformer, and dynamic calculation of the optimal economic operating range based on the operating conditions of the distribution transformer; S4. Establish a thermal model of the distribution transformer and calculate the dynamic overload margin of each distribution transformer. S5. Construct a multi-objective optimization model. The objective functions include minimizing the total loss of the distribution transformer group, minimizing the overload margin variance, minimizing the voltage deviation, and optimizing the sensitivity of the protection setting. The constraints include dynamic efficiency range constraints, power flow constraints, capacity constraints, voltage constraints, and overload margin constraints. S6. Solve the multi-objective optimization model using a multi-objective optimization algorithm to obtain the power setting value of each flexible intelligent soft switch; S7. Generate a power regulation curve based on the optimization results, execute phase synchronization control and dual-machine voltage vector matching, and achieve seamless load transfer through flexible intelligent soft switching output power regulation and instantaneous power compensation of energy storage unit; S8. Real-time monitoring of the power direction of each distribution transformer. When power backfeed is detected in the distribution transformer, the flexible intelligent soft switch is controlled to transfer the backfeeding power while performing active harmonic compensation and control, and injecting the backfeeding power into the adjacent distribution transformer. S9. Monitor the operating status of the flexible intelligent soft switch in real time, execute a safe backoff path before blocking the faulty branch, pull back the load to the original transformer's load-bearing range, and then block the flexible intelligent soft switch and switch to the degraded control mode.

2. The optimization method as described in claim 1, characterized in that: In step S1, the flexible intelligent soft switch adopts a back-to-back bidirectional AC-DC-AC converter topology, and the energy storage unit is a supercapacitor. In step S2, the future load of each distribution transformer is predicted in a rolling manner based on a neural network. The input dimensions include historical load curves, weather forecasts, and date types. In step S3, the recursive least squares method is used to identify the iron loss and copper loss parameters of each distribution transformer in real time, and the upper and lower limits of the optimal economic operating range are dynamically calculated based on the transformer aging coefficient, temperature correction coefficient, and harmonic distortion rate influence coefficient.

3. The optimization method as described in claim 1, characterized in that: In step S3, the lower limit of the optimal economic operating range for: The upper limit of the optimal economic operating range for: in, This is the lower limit of the baseline efficiency range. The upper limit of the baseline efficiency range, The aging factor is... For the service life of the distribution transformer, This is a temperature correction factor. The deviation between the measured temperature and the rated temperature of the distribution transformer. The harmonic distortion rate influence coefficient. This represents the total harmonic distortion rate of the transformer current.

4. The optimization method as described in claim 3, characterized in that: In step S4, a thermal model of the distribution transformer is established based on the IEEE C57.91 standard, and the formula for calculating the dynamic overload margin is as follows: in, For distribution transformers exist Dynamic overload margin at any given time. This refers to the maximum allowable temperature of the transformer winding. For distribution transformer Temperature is measured in real time. Ambient temperature; The power flow constraints in step S5 include: Output power constraints of distribution transformers: End-load node voltage constraints: in, For distribution transformers The output active power, For the active power of the load, For power transmission in flexible intelligent soft switches, For line current, The equivalent line resistance from the distribution transformer to the user, The equivalent line reactance from the distribution transformer to the user, For the low-voltage side bus voltage of the distribution transformer, This refers to the voltage at the end load node. The power factor.

5. The optimization method as described in claim 1, characterized in that: The objective function of the multi-objective optimization model in step S5 is: in, For the total loss of the transformer group, For distribution transformers At load rate The overall power loss is as follows. For the loss weighting coefficient of flexible intelligent soft switching, For flexible intelligent soft switches Self-consumption For overload margin variance, For a moment Dynamic overload margin of each distribution transformer The variance; The sum of squares of voltage deviations For nodes The square of the voltage deviation, For sensitivity evaluation function, For fixed parameters Probability of protection malfunction Sensitivity coefficient; In step S6, the NSGA-III multi-objective evolutionary algorithm is used to solve the multi-objective optimization model, output the Pareto optimal solution set, and select the optimal solution based on the real-time electricity price and load tension.

6. The optimization method as described in claim 1, characterized in that: In step S7, the non-inductive load transfer process includes: the flexible intelligent soft switch locks the voltage phase of the adjacent distribution transformer bus through a phase-locked loop, and realizes the parallel operation of the two machines through voltage feedforward control, ensuring that the voltage vector angle fluctuation of the two transformer bus is less than 5° and the voltage fluctuation is less than 2% during the entire power transfer process.

7. The optimization method as described in claim 1, characterized in that: The triggering condition for the reverse power control in step S8 is: When the active power of the distribution transformer is detected to be negative, it is determined to be power backfeeding; The system automatically calculates the absolute value of the reverse power and generates a reverse power transfer command; The flexible intelligent soft switch controls the total harmonic distortion rate of the low-voltage side of the distribution transformer to below 5% through harmonic current detection and command compensation in the dq coordinate system.

8. The optimization method as described in claim 1, characterized in that: The specific strategy for the degradation control mode in step S9 is as follows: After the fault-tolerant intelligent soft-locking is closed, the system automatically switches to the traditional radial operating topology; The optimization objective has been downgraded from minimizing the total loss of the transformer group to avoiding transformer overload and voltage exceeding limits. Only the power distribution of the non-faulty flexible intelligent soft switches is optimized, while the faulty branches are restored to traditional operation; After the fault is resolved, the system automatically returns to normal optimization mode.

9. A distribution transformer group reconfiguration and optimization system based on intelligent soft switching, used to implement the method of any one of claims 1 to 8, characterized in that, include: The flexible intelligent soft-switching module is deployed on the low-voltage side of the adjacent distribution transformer, adopting a bidirectional converter topology, with energy storage units connected in parallel on the DC side; The data acquisition module is installed in each distribution transformer to collect electrical parameters in real time; The load forecasting module predicts the future load of each distribution transformer; The parameter identification module identifies the loss parameters of the distribution transformer in real time and dynamically calculates the optimal economic operating range. The overload margin calculation module calculates the dynamic overload margin of each distribution transformer in real time. The multi-objective optimization module solves the multi-objective optimization model that minimizes the total loss of the distribution transformer group, the overload margin variance, and the voltage deviation, and outputs the power setpoints of each flexible intelligent soft switch. The sensorless control module coordinates the flexible intelligent soft switch and energy storage unit based on the optimization results to achieve sensorless load transfer; The reverse power control module monitors the power direction of the distribution transformer in real time, and automatically triggers a reverse power transfer command when reverse power is detected. The fault diagnosis and degradation control module monitors the operating status of the flexible intelligent soft switch in real time, and automatically locks and switches to degradation control mode when a fault occurs.

10. The optimization system as described in claim 9, characterized in that: The flexible intelligent soft-switching module adopts a back-to-back bidirectional AC-DC-AC converter topology, and the energy storage unit is a supercapacitor energy storage unit with a single-sided rated capacity of 50-200kVA and a response time of less than 10 milliseconds. The load forecasting module uses an LSTM neural network to perform rolling forecasts of the load of each distribution transformer for the next 15 minutes. The parameter identification module uses the recursive least squares method to identify the iron loss and copper loss parameters of the distribution transformer in real time. The overload margin calculation module calculates the dynamic overload margin of each distribution transformer in real time based on the IEEE C57.91 standard. The multi-objective optimization module uses an improved NSGA-III algorithm to solve the multi-objective optimization model, and automatically adjusts the reference point weights based on historical scheduling decisions and actual results through the Q-Learning reinforcement learning algorithm. The fault diagnosis and degradation control module includes a safety back-off controller, which, when a fatal abnormal signal is detected, prioritizes calling the energy storage unit to smoothly reduce the transmission power of the branch where the faulty flexible intelligent soft switch is located to zero, and then executes the contactor action and IGBT blocking logic.

Images