Electric energy quality treatment method and system based on self-power supply and power electronic module cooperation

Abstract

The invention discloses a self-powered and power electronic module cooperation-based power quality treatment method and system, and the method comprises the steps: dividing an operation scene into a high-voltage scene, a low-voltage scene, a three-phase imbalance scene and a harmonic pollution scene according to the power quality problem of a transformer area, and employing different coping strategies; if the scene is a high-voltage scene, a reinforcement learning algorithm is adopted to optimize a reactive power output strategy of the SVG; if the scene is a low-voltage scene, optimizing an energy storage scheduling optimization strategy of the PPO by adopting a near-end strategy; if the scene is a three-phase imbalance scene, adopting a current transfer optimization strategy of a convolutional neural network (CNN); if the scene is a harmonic pollution scene, a harmonic compensation current generation strategy of a generative adversarial network (GAN) is adopted; through the self-powered module and the power electronic module, the problems of line loss increase, serious voltage fluctuation, three-phase imbalance and serious harmonic pollution are solved.

Description

Technical Field

[0001] The present invention relates to a method and system for power quality management, and particularly to a method and system for power quality management based on the cooperation of self-power supply and power electronic modules. Background Art

[0002] With the promotion of the realization of the goals of "carbon peak and carbon neutrality" and the deepening of the power system reform, distributed new energy has been widely used due to its environmental protection and energy-saving advantages, especially in the fields of photovoltaic and wind power. According to statistics, by 2025, the penetration rate of distributed new energy in the power grid is expected to reach more than 20%.

[0003] However, the large-scale access of distributed photovoltaic to the power grid has also brought significant power quality problems, especially in the substation area (the power supply area of the distribution transformer). During the day when the sunlight is sufficient, the full power generation of photovoltaic leads to reverse power supply, and even when the transformer is adjusted to the lowest gear, it is still difficult to eliminate the high voltage phenomenon in the substation area; while at night when the photovoltaic stops generating power and the power consumption load surges, there is generally a low voltage problem at the end of the substation area. This "high during the day and low at night" voltage instability and fluctuation phenomenon has become the main reason for the deterioration of regional power quality; the intermittent output characteristic of photovoltaic leads to an increase in line losses in the substation area, especially when the photovoltaic output fluctuates greatly, the line loss problem is more significant; the access of distributed photovoltaic often leads to unbalanced three-phase currents in the substation area, further exacerbating the line loss and equipment loss; the widespread use of power electronic devices such as photovoltaic inverters has led to an increasingly serious harmonic pollution problem in the substation area, affecting the safe operation of electrical equipment.

[0004] Although the traditional SVG can solve some power quality problems, its own loss is large during operation (full load loss ≥ 3%), resulting in a further increase in line losses in the substation area. In addition, the control strategy of SVG is single, and its dynamic perception and response ability are insufficient, making it difficult to adapt to the complex and changeable operation environment of the substation area; the transformation cost of old substations is high, the investment return period is long, and the compatibility and scalability of existing equipment are poor, making it difficult to meet the development needs of future power systems; existing power quality management devices usually rely on external power grid power supply, increasing the operation cost and dependence on the power grid, which does not conform to the development trend of green energy. Summary of the Invention

[0005] Object of the Invention: The object of the present invention is to provide a power quality management method to reduce line losses, stabilize voltage fluctuations, achieve three-phase balance and filter out harmonics. On the other hand, a power quality management system is provided.

[0006] Technical Solution: The power quality management method described in the present invention includes: According to the power quality problems in the substation area, the operation scenarios are divided into high voltage scenarios, low voltage scenarios, three-phase unbalance scenarios and harmonic pollution scenarios; In the case of a high - voltage scenario, a reinforcement learning algorithm is used to optimize the reactive power output strategy of the SVG; In the case of a low - voltage scenario, a proximal policy optimization (PPO) is used to optimize the energy storage scheduling optimization strategy; In the case of a three - phase unbalance scenario, a convolutional neural network (CNN) is used to extract features from the three - phase current time series to generate a current transfer optimization strategy; In the case of a harmonic pollution scenario, a harmonic compensation current generation strategy using a generative adversarial network (GAN) is adopted.

[0007] Preferably, the determination conditions for the operating scenarios are as follows: High - voltage scenario: V > 1.1V nominal ; Low - voltage scenario: V < 0.9V nominal ; Three - phase unbalance scenario: The three - phase current unbalance degree ΔI unbalance > 15%; Harmonic pollution scenario: The total harmonic distortion rate THDi > 5%.

[0008] Preferably, the objective function of the reactive power output strategy is as follows: ; Among them, represents the voltage deviation, represents the line loss power, represents the SVG reactive power output, represents the weight coefficient; The constraint conditions are as follows: Voltage upper limit: V ≤ 1.1V nominal ; SVG capacity limit: ; Output of the self - power supply module: ; The penalty function is as follows: ; Among them, K 1 、K 2 represents the penalty coefficient.

[0009] Preferably, the reward function of the energy storage scheduling optimization strategy is as follows: ; ; Among them, represents the voltage deviation penalty, represents the line loss power penalty, represents the energy storage SOC, represents the constraint violation penalty; The constraint conditions are as follows: Lower voltage limit: V ≥ 0.9V nominal ; Energy storage charge and discharge power limit: ; Safe range of energy storage SOC: SOC ≥ 20%; SVG capacity limit: .

[0010] Preferably, the energy storage scheduling optimization strategy includes: interacting with the environment through the energy storage scheduling optimization strategy of proximal policy optimization (PPO) to collect trajectory data; using Generalized Advantage Estimation (GAE) to estimate the advantage value; limiting the policy update amplitude through a clipping mechanism to maximize the objective function; minimizing the mean squared error loss of the value network.

[0011] Preferably, the objective function of the current transfer optimization strategy is as follows: ; ; where, represents the three-phase unbalance weight, represents the line loss weight; The constraint conditions are as follows: Current transfer amount limit: ; Three-phase unbalance safety threshold: ΔI unbalance ≤ 10%; IGBT switching frequency limit: ; The penalty function is as follows: .

[0012] Preferably, the current transfer optimization strategy includes: interacting with the environment through the energy storage scheduling optimization strategy of proximal policy optimization (PPO) to collect trajectory data; using Generalized Advantage Estimation (GAE) to estimate the advantage value; limiting the policy update amplitude through a clipping mechanism to maximize the objective function; minimizing the mean squared error loss of the value network.

[0013] Preferably, the objective function of the harmonic compensation current generation strategy is as follows: ; where, represents the harmonic distortion rate weight, represents the line loss weight; The constraint conditions are as follows: Harmonic distortion rate safety threshold: THDi ≤ 5%; Compensation current bandwidth limit: ; Inverter capacity limit: ; The penalty function is as follows: 。

[0014] Preferably, the harmonic compensation current generation strategy includes: collecting harmonic spectrum data from an actual system or simulation and inputting it into a power system simulation tool; initially training a generator using the mean square error (MSE) to generate an ideal compensation current; alternately training the generator and the discriminator to gradually optimize the generation strategy; and dynamically adjusting the generator parameters through real-time data.

[0015] The power quality management system of the present invention includes: A self-power supply module for providing power to the system through photovoltaic power generation and supplying power to the power electronics module; A power electronics module for providing different coping strategies according to different operating scenarios. The strategies include a reactive power output optimization strategy for SVG optimized by a reinforcement learning algorithm, a energy storage scheduling optimization strategy for PPO optimized by a proximal policy optimization, a current transfer optimization strategy, and a harmonic compensation current generation strategy for GAN.

[0016] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: 1. The power electronics module is powered by the self-power supply module, reducing the dependence on the external power grid. At the same time, the reactive power compensation strategy is optimized to reduce line losses, and the power generation of the self-power supply module is not included in the gateway meter, further reducing the line loss rate; 2. By adjusting the phase and amplitude of the AC output voltage of the bridge rectifier circuit, reactive power is dynamically compensated, and the system voltage is stabilized near the set reference value to control voltage fluctuations; 3. By driving the IGBT, the unbalanced phase current is transferred from the large phase to the small phase to achieve three-phase balance on the grid side; 4. By generating a compensation current with the same magnitude and opposite phase to the harmonics in the system through the IGBT power converter, harmonic filtering is achieved. Description of the Drawings

[0017] Figure 1 It is a schematic diagram of the process of the present invention. Detailed Embodiments

[0018] As Figure 1 shown, according to the power quality problems in the distribution area, the operating scenarios are divided into high voltage scenarios, low voltage scenarios, three-phase imbalance scenarios, and harmonic pollution scenarios; Scenario determination conditions: High voltage scenario: V > 1.1V nominal , the photovoltaic power generation is greater than the load demand of the distribution area; during the daytime when the sunlight is sufficient (such as 10:00 - 15:00); Low voltage scenario: V < 0.9V nominal , during the night when the load surges and the photovoltaic power generation stops; during the peak load period at night (such as 18:00 - 22:00); Three-phase unbalanced scenario: The unbalance degree of three-phase current ΔI unbalance > 15%, during the period when the load distribution in the substation area is uneven and the load fluctuates greatly; Harmonic pollution scenario: The total harmonic distortion rate THDi > 5%, during the period when power electronic devices are intensive and the photovoltaic output fluctuates greatly.

[0019] In the case of a high-voltage scenario, the reinforcement learning algorithm is used to optimize the reactive power output strategy of the SVG. By real-time feedback of voltage changes, the reactive power compensation amount is dynamically adjusted. Through continuous learning, the system can find the optimal reactive power compensation strategy in a complex and changeable operating environment.

[0020] Algorithm process: State space: Substation area voltage, photovoltaic output, load demand; Action space: SVG reactive power output; Reward function: The smaller the voltage deviation, the higher the reward; The lower the line loss, the higher the reward; The objective function is as follows: ; Among them, represents Voltage deviation , represents the line loss power, represents the SVG reactive power output, represents the weight coefficient; The constraint conditions are as follows: Voltage upper limit: V ≤ 1.1V nominal ; SVG capacity limit: ; Output of the self-powered module: ; The penalty function is as follows: ; Among them, K 1 , K 2 represents the penalty coefficient.

[0021] Solution method: Deep Q Network DQN: (1) Neural network structure: Input layer: 4 nodes (corresponding to the state space dimension); Hidden layer: 2 fully connected layers (24 nodes, ReLU activation); Output layer: 3 nodes (corresponding to the Q values of 3 discrete actions); (2) Experience replay: Buffer size: 2000 pieces of experience (state, action, reward, next state, termination flag); Sampling strategy: Randomly select batches (batch_size = 32) for training to break data correlation; (3) Greedy strategy: Initial exploration rate ϵ = 1.0; Minimum exploration rate: ϵmin = 0.01; Decay rate: ϵdecay = 0.995ϵdecay = 0.995, gradually reduce the exploration rate to balance exploration and exploitation.

[0022] (4)Target network: Update frequency: Copy the main network parameters to the target network every 100 steps to stabilize the training process.

[0023] In the case of a low-voltage scenario, adopt the proximal policy optimization (PPO) energy storage scheduling optimization strategy. When the night load surges and causes the substation area voltage to be lower than the rated value, it is necessary to improve the voltage through energy storage discharge and SVG capacitive reactive power compensation, while minimizing line losses and energy storage life losses.

[0024] State space: Substation area voltage , load demand , energy storage SOC ϵ[0.2, 1.0] (20% - 100%), SVG reactive power output (kVar); Action space: Continuous action, energy storage charge and discharge power (negative value for charging, positive value for discharging); Reward function: ; ; Among them, represents the voltage deviation penalty (target voltage 0.95V nominal ), represents the line loss power penalty, represents the energy storage SOC change penalty (encouraging stable charge and discharge), represents the constraint violation penalty (such as voltage over-limit, SOC over-limit); The constraint conditions are as follows: Voltage lower limit: V ≥ 0.9V nominal ; Energy storage charge and discharge power limit: ; Energy storage SOC safety range: SOC ≥ 20%; SVG capacity limit: .

[0025] PPO optimizes the energy storage scheduling strategy through importance sampling and policy clipping while ensuring the stability of training; Neural network structure: Policy network (Actor), which outputs the mean and standard deviation (Gaussian distribution) of the energy storage charge and discharge power; Input layer with 4 nodes (corresponding to the state space); Two fully connected hidden layers (64 nodes, ReLU activation); Output layer with 1 node (mean ) + 1 node (standard deviation ); Value network (Critic) to evaluate the state value: Input layer with 4 nodes; Two fully connected hidden layers (64 nodes, ReLU activation); Output layer with 1 node (state value V(s)).

[0026] Algorithm steps: (1) Interact with the environment through the current policy to collect trajectory data; (2) Use GAE (Generalized Advantage Estimation) to estimate the advantage value ; (3) Limit the policy update amplitude through a clipping mechanism to maximize the objective function: ; (4) Update the value function to minimize the mean squared error loss of the value network; (5) Hyperparameter settings: Discount factor = 0.99, GAE parameter = 0.95, clipping threshold ϵ = 0.2, learning rate , batch size 64, number of training epochs 1000.

[0027] Training steps: (1) Initialize the environment: Set the initial voltage, load demand, and energy storage SOC; (2) Collect trajectory data (state, action, reward, next state) generated by the current policy; (3) Use GAE to estimate the advantage value of each state ; (4) Update the Actor and Critic networks through the PPO clipping mechanism; (5) Repeat steps 2 - 4 until convergence.

[0028] Key implementation details: Action mapping, linearly map the output action to the actual charge and discharge power: ; Add a penalty term for SOC violation to the reward function: 。

[0029] Reward function adjustment: Increase the reward for the voltage recovery speed to avoid excessive discharge of the energy storage; Exploration strategy: Add noise at the beginning of training to improve the diversity of the strategy; Parallel training: Use multiple environments to collect data in parallel to accelerate the training process.

[0030] Taking a low-voltage area in Jiangsu as an example: Initial state: Voltage 0.88 p.u . Load 300 kW, energy storage SOC = 50%; Results after training: Voltage recovered to 0.95 p.u . Line loss reduced by 15%; The energy storage SOC is dynamically adjusted, and the lower limit is not triggered (SOC > 25%); Comparison with traditional strategies: The voltage recovery time of the PPO strategy is shortened by 30%, and the energy storage life loss is reduced by 20%.

[0031] Through the PPO algorithm, the system can intelligently dispatch the charging and discharging power of the energy storage and the reactive power output of the SVG, optimize the line loss and the energy storage life while improving the voltage; In the next step, combined with the photovoltaic output prediction, multi-time scale optimization can be realized to further improve the economy and reliability.

[0032] In the case of a three-phase unbalanced scenario, features are extracted from the three-phase current time series through a convolutional neural network CNN to generate an optimized current transfer strategy.

[0033] In a three-phase unbalanced scenario, the uneven load distribution leads to significant differences in the currents of each phase. It is necessary to transfer the high-phase current to the low-phase through the IGBT drive circuit to achieve three-phase balance while minimizing the line loss.

[0034] State space: Instantaneous values of three-phase currents: (Time series data, sampling frequency 1 kHz); Current unbalance degree: ; Line loss power: ; Action space: Continuous action, inter-phase current transfer amount (Positive value means transfer from high phase to low phase, negative value means reverse transfer); Objective function: ; ; Among them, represents the weight of the three-phase unbalance degree, represents the weight of the line loss.

[0035] The constraints are as follows: Current transfer amount limit: ; Safety threshold of three-phase unbalance degree: ΔI unbalance ≤10%; IGBT switching frequency limit: ; The penalty function is as follows: ; K 1 = 10, K 2 = 50.

[0036] Convolutional neural network CNN design: CNN is used to extract features from the three-phase current time series and generate the optimal current transfer strategy.

[0037] (1) Input data preprocessing: Input shape T×3 (time step T = 100, three-phase current); Normalize the current data by phase (mean 0, variance 1); (2) Network structure: Input layer (100, 3), Convolutional layer Conv1D(32, kernel_size = 5, activation='relu'): Extract local time features, MaxPooling1D(2) for downsampling to reduce the computational amount; Fully connected layer Dense(64, activation='relu'): Feature fusion, Dense(1, activation='tanh'): Output the current transfer amount (range [-1, 1], which needs to be linearly mapped to the actual value); (3) Loss function: ; MSE represents minimizing the mean square error between the predicted current transfer amount and the target value, and Penalty represents the penalty term for constraint violation.

[0038] Training and optimization: (1) Data preparation: Use power system simulation software (such as MATLAB / Simulink or OpenDSS) to simulate different load distribution scenarios, generate the input three-phase current time series (T = 100), and output the optimal current transfer amount (calculated by traditional optimization algorithms or expert strategies); Data augmentation, adding noise (±5% current fluctuation) and random load mutations to improve the robustness of the model; (2) Hyperparameter tuning: Learning rate 0.001 (Adam optimizer), batch size 32, number of training epochs 200, convolutional kernel size 5 (balancing local features and computational efficiency); (3) Performance verification metrics: Imbalance reduction rate: ; Line loss reduction rate: 。

[0039] Taking a three-phase unbalanced power distribution area as an example: Initial state: Three-phase current , Imbalance , Line loss ; After CNN optimization: Current transfer amount (transferring from phase A to phase C), Balanced current , Imbalance , Line loss (reduced by 36%).

[0040] High-precision feature extraction. CNN captures the local temporal features of three-phase current through convolutional kernels, which is more adaptable to dynamic load changes than traditional FFT analysis; Real-time response ability, with a single inference time < 1ms, meeting the 10kHz switching frequency requirement. The model can adapt to new load distribution patterns through online learning without re-designing the control logic; Multi-objective coordination. The line loss and imbalance are integrated into the loss function to achieve dual optimization of economy and power quality.

[0041] S5. If it is a harmonic pollution scenario, adopt the harmonic compensation current generation strategy of the generative adversarial network GAN.

[0042] In a harmonic pollution scenario, the intensive use of power electronic devices (such as photovoltaic inverters, frequency converters) leads to a total harmonic distortion rate (THDi > 5%). It is necessary to generate compensation current ( ) that is equal in magnitude and opposite in phase to the harmonic to achieve harmonic filtering and minimize line loss.

[0043] State space: Harmonic current spectrum (characteristic harmonic amplitudes such as 3rd, 5th, 7th harmonics), meeting the requirements of opposite phase and matching amplitude.

[0044] Objective function: ; Among them, represents the harmonic distortion rate weight, represents the line loss weight; 。

[0045] The constraint conditions are as follows: Harmonic distortion rate safety threshold: THDi ≤ 5%; Compensation current bandwidth limit: ; Inverter capacity limit: ; The penalty function is as follows: ; where, , represents the penalty coefficient.

[0046] Generative Adversarial Network (GAN) design: The GAN generates the compensated current spectrum through the Generator, and the Discriminator distinguishes the real harmonic data from the generated data, gradually optimizing the generation strategy.

[0047] (1) Generator: Input the random noise vector and the real-time harmonic spectrum , and output the compensated current spectrum (N is the harmonic order).

[0048] (2) Network structure: Fully connected layer (128 nodes, ReLU), fully connected layer (64 nodes, ReLU), output layer (NN nodes, tanh activation); Output mapping, linearly map the tanh output (range [-1,1]) to the actual compensated current amplitude: ; represents the maximum allowable compensated current for each harmonic; Fully connected layer (64 nodes, LeakyReLU), fully connected layer (32 nodes, LeakyReLU), output layer (1 node, sigmoid activation).

[0049] (3) Loss function: Generator loss: (The first term deceives the discriminator, and the second term penalizes the constraint violation); Discriminator loss: .

[0050] (4) Training process: Data acquisition, collect the harmonic spectrum data IharmonicIharmonic from the actual system or simulation; Generator pre-training, initially train the generator using the Mean Squared Error (MSE) to generate a rough compensated current; Adversarial training, alternately train the generator and the discriminator to gradually optimize the generation strategy; Online fine-tuning, dynamically adjust the generator parameters through real-time data after deployment.

[0051] (5) Training and optimization: Use power system simulation tools (such as MATLAB / Simulink or PLECS) to simulate different harmonic scenarios and generate the input harmonic spectrum \(I_{harmonic}\), and the output ideal compensation current \(I_{comp}^{target}\) (calculated through FFT analysis); perform data augmentation by adding random harmonic perturbations (±10% amplitude change) and noise (Gaussian white noise) to improve the robustness of the model; Hyperparameter settings: Noise dimension 100, number of training epochs 500, batch size 32, learning rate for the generator (0.0001), discriminator (0.0003), penalty weight λ = 10.

[0052] Performance metrics: THDi reduction rate: ; Line loss reduction rate: ; Taking a harmonic - polluted sub - station in an industrial park as an example: Initial state: ; THDi = 8.5%, line loss .

[0053] After GAN compensation: Generate the compensation current ; After compensation, THDi = 3.2%, line loss (reduced by 33%).

[0054] GAN generates compensation currents that highly match the real harmonics through adversarial training, which is superior to the traditional FFT + PI control; the online fine - tuning mechanism enables the model to adapt to harmonic spectrum changes (such as load mutations) in real - time; at the same time, it compensates for characteristic harmonics such as the 3rd, 5th, and 7th harmonics, avoiding the lag of successive filtering.

Claims

1. A method for power quality management based on the collaboration of self-power supply and power electronic module, characterized in that: include: According to the power quality issues in the substation, the operation scenarios are divided into high voltage scenario, low voltage scenario, three-phase unbalance scenario and harmonic pollution scenario; In high-voltage scenarios, a reinforcement learning algorithm is used to optimize the reactive power output strategy of SVG; In low-voltage scenarios, the proximal strategy is used to optimize the PPO energy storage scheduling optimization strategy; In the case of a three-phase unbalanced scenario, the convolutional neural network (CNN) is used to extract features from the three-phase current time series and generate a current transfer optimization strategy. If it is a harmonic pollution scenario, the harmonic compensation current generation strategy of the generative adversarial network GAN is adopted.

2. The power quality control method according to claim 1, characterized in that: The judgment conditions of the operation scenario are as follows: High voltage scenario: V>1.1V nominal ; Low voltage scenario: V<0.9V nominal ; Three-phase unbalanced scenario: three-phase current imbalance ΔI unbalance >15%; Harmonic pollution scenario: total harmonic distortion rate THDi>5%.

3. The power quality control method according to claim 1, characterized in that: The reactive power output strategy objective function is as follows: ； in, Indicates voltage deviation, Indicates line loss power, Indicates SVG reactive output. represents the weight coefficient; The constraints are as follows: Voltage upper limit: V≤1.1V nominal ; SVG capacity limitations: ; Self-powered module output: ; The penalty function is as follows: ； Among them, K1 and K2 represent penalty coefficients.

4. The power quality control method according to claim 1, characterized in that: The energy storage scheduling optimization strategy reward function is as follows: ； ； in, represents the voltage deviation penalty, Indicates line loss power penalty, Indicates energy storage SOC, indicates a constraint violation penalty; The constraints are as follows: Voltage lower limit: V ≥ 0.9V nominal ; Energy storage charging and discharging power limit: ; Energy storage SOC safety range: SOC ≥ 20%; SVG capacity limitations: .

5. The power quality control method according to claim 1, characterized in that: The energy storage scheduling optimization strategy includes: optimizing the PPO energy storage scheduling optimization strategy through the proximal strategy to interact with the environment and collect trajectory data; using GAE to estimate the advantage value; limiting the strategy update amplitude through the clipping mechanism to maximize the objective function; and minimizing the mean square error loss of the value network.

6. The power quality control method according to claim 1, characterized in that: The objective function of the current transfer optimization strategy is as follows: ； ； in, represents the three-phase unbalance weight, represents line loss weight; The constraints are as follows: Current transfer limit: ; Three-phase unbalance safety threshold: ΔI unbalance ≤10%; IGBT switching frequency limit: ; The penalty function is as follows: 。 7. The power quality control method according to claim 1, characterized in that: The current transfer optimization strategy includes: normalizing the current data by phase; simulating different load distribution scenarios through power system simulation software and inputting three-phase current time series; outputting the optimal current transfer amount through traditional optimization algorithms or expert strategies; adding noise and random load mutations; and integrating the loss function of MSE and constraint penalty.

8. The power quality control method according to claim 1, characterized in that: The objective function of the harmonic compensation current generation strategy is as follows: ； in, represents the harmonic distortion weight, represents line loss weight; The constraints are as follows: Harmonic distortion safety threshold: THDi≤5%; Compensation current bandwidth limitation: ; Inverter capacity limitation: ; The penalty function is as follows: 。 9. The power quality control method according to claim 1, characterized in that: The harmonic compensation current generation strategy includes: collecting harmonic spectrum data from the actual system or simulation and inputting it into the power system simulation tool; using the mean square error (MSE) to preliminarily train the generator to generate an ideal compensation current; alternately training the generator and the discriminator to gradually optimize the generation strategy; and dynamically adjusting the generator parameters through real-time data.

10. A power quality management system based on self-power supply and power electronic module collaboration, characterized in that: include: A self-powered module, which is used to provide power to the system through photovoltaic power generation and to power the power electronic module; The power electronics module is used to provide different response strategies according to different operating scenarios, including the reactive power output strategy of SVG optimized by reinforcement learning algorithm, the energy storage scheduling optimization strategy of proximal strategy optimization PPO, the current transfer optimization strategy and the harmonic compensation current generation strategy of generative adversarial network GAN.

Images