Virtual Synchronous Machine Dynamic Inertia Control Method and System Based on Artificial Intelligence Technology

Abstract

The present invention provides a method and system for dynamic inertia control of a virtual synchronous machine based on artificial intelligence technology, which is applied to the control of a voltage source inverter. The main steps include: data acquisition and preprocessing, and outputting dynamic parameters through a time series neural network for the control links of the virtual synchronous machine respectively; calculating a phase compensation angle by using a convolutional neural network for phase correction of pulse width modulation, and then using a deep reinforcement learning network to output a duty cycle correction amount to adjust the pulse width modulation for collaborative control of the dynamic inertia of the virtual synchronous machine. The present invention obtains dynamically adjusted parameters through data time series processing and neural network application, and can realize adaptive adjustment of core parameters; uses a convolutional neural network to participate in the calculation of modulation signals, effectively compensates for hardware phase delay, and improves the dynamic response accuracy; and then realizes collaborative optimization of inertia control and dynamic response by combining deep reinforcement learning, reduces current overshoot and suppresses harmonics, and improves power quality and system stability.

Description

Technical Field

[0001] The present invention relates to the field of power supply technology, and specifically relates to a virtual synchronous generator dynamic inertia control method and system based on artificial intelligence technology. Background Art

[0002] With the large-scale access of distributed power sources to the power grid, the virtual synchronous generator (VSG) technology has attracted much attention because it can effectively improve the grid connection stability and power quality of distributed power sources and simulate the inertia and damping characteristics of synchronous generators.

[0003] Among them, the two-level voltage source inverter (2L-VSI) has become a common topology for VSG implementation due to its simple structure and low cost. However, the traditional 2L-VSI type VSG faces many technical problems in actual operation, specifically as follows:

[0004] Hardware characteristic constraints: The LC filter configured in it has a phase lag, and the switching frequency is limited by hardware conditions, resulting in a delay in the dynamic response of the VSG to grid frequency and voltage fluctuations and making it difficult to quickly track grid changes;

[0005] Insufficient parameter adaptability: Using fixed virtual inertia (J) and damping coefficient (D), in the face of various scenarios such as photovoltaic output fluctuations, load mutations, and weak power grids, it cannot be dynamically adjusted to meet actual needs, reducing the flexibility and stability of system operation;

[0006] Limitations of control algorithms: Based on the control strategy of linear PI regulators, in the face of nonlinear disturbances such as harmonic resonance and load mutations, it cannot effectively coordinate the relationship between inertia support and fast response, often resulting in the compromise failure of both and affecting the overall performance of the system.

[0007] In response to the above problems, existing solutions include parameter adaptive adjustment technology, predictive control solutions, and hardware compensation solutions. Although to a certain extent, it has improved problems such as the response speed and stability of the system, there are still obvious defects, specifically as follows:

[0008] Parameter adaptive adjustment scheme: Mostly rely on rule bases or fuzzy logic to adjust J and D, but these methods highly rely on manual experience to construct rules, and when facing complex and changeable actual working conditions, the generalization ability is insufficient and it is difficult to achieve precise adaptation.

[0009] Predictive control solution: Although model predictive control (MPC) can theoretically optimize control, it is necessary to establish an accurate mathematical model, and the 2L-VSI has non-ideal characteristics such as dead zone effects and device aging, resulting in model mismatch and greatly reducing the control effect.

[0010] Hardware compensation scheme: The performance is improved by adding filters or increasing the switching frequency, but this will significantly increase the system cost and power loss, violating the original intention of the 2L-VSI to pursue economy and being limited in practical applications.

[0011] In summary, the traditional 2L-VSI type VSG still requires an improved control method and corresponding system to improve the overall dynamic performance and have better multi-scenario adaptability and economy. Summary of the Invention

[0012] In view of the problems existing in the prior art, the present invention provides a system with a simple structure and low cost, and a method for collaborative control of the dynamic inertia of a virtual synchronous machine.

[0013] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0014] On the one hand, the present invention provides a method for controlling the dynamic inertia of a virtual synchronous machine based on artificial intelligence technology, which mainly includes the following steps:

[0015] Real-time collect multi-dimensional power grid information; preprocess the multi-dimensional power grid information and generate a time series data sequence in a preset time window;

[0016] Input the time series data sequence into a time series neural network and output dynamically adjusted virtual inertia, damping coefficient and virtual impedance;

[0017] Substitute the virtual inertia and the damping coefficient into the control module of the virtual synchronous machine, and substitute the virtual impedance into the power control voltage and current control link of the virtual synchronous machine;

[0018] Use a convolutional neural network to extract the phase-frequency characteristics of the filter of the voltage source inverter and generate a phase compensation angle; correct the phase of the pulse width modulation signal for driving the voltage source inverter through the phase compensation angle;

[0019] Train a deep reinforcement learning network based on the power quality comprehensive optimization reward function, use the system state parameters of the voltage source inverter and the power grid operation state parameters as the input of the deep reinforcement learning network, and output the duty cycle correction amount of the pulse width modulation signal for driving the voltage source inverter; input the duty cycle correction amount into the pulse width modulation strategy of the voltage source inverter to achieve collaborative control of the dynamic inertia of the virtual synchronous machine.

[0020] Optionally, the time series neural network is synchronously updated through an online update strategy;

[0021] The online update strategy includes the following steps:

[0022] Periodically collect the multi-dimensional power grid information and historical inertia parameter sequences, and form a current operating state dataset;

[0023] Apply a sliding window forgetting factor to the current operating state dataset to obtain a training dataset; in the training dataset, assign an updated weight to the latest data, and reduce the weight of the historical data in the dataset by the exponential decay method to generate weighted training data;

[0024] Periodically perform incremental updates on the weights of the time series neural network based on the weighted training data.

[0025] Optionally, the time series neural network is a long short-term memory network and adopts a two-layer long short-term memory structure;

[0026] In the two-layer long short-term memory structure, the bottom layer has 64 neurons and the top layer has 32 neurons.

[0027] Optionally, the time series neural network is optimized through transfer learning after being deployed in the corresponding scenario;

[0028] The transfer learning includes source domain pre-training based on multi-scenario simulation data and target domain fine-tuning using the measured data of the target scenario to update the top network parameters.

[0029] Optionally, the multi-dimensional power grid information includes power grid frequency deviation, active power fluctuation, DC bus voltage change rate, and output current signal;

[0030] The preprocessing at least includes noise filtering processing of the collected multi-dimensional power grid information through a filtering algorithm.

[0031] Optionally, the convolutional neural network is a one-dimensional convolutional neural network, adopts an asymmetric convolutional kernel structure, and the kernel weight distribution is obtained through phase difference supervised training.

[0032] Optionally, the deep reinforcement learning network is a deep Q network;

[0033] The parameters of the power quality comprehensive optimization reward function are at least provided with a frequency deviation reward term, a current overshoot penalty term, and a harmonic distortion penalty term, and the index weights of each item are dynamically set.

[0034] On the other hand, the present invention also provides a virtual synchronous machine dynamic inertia control system based on artificial intelligence technology for implementing the virtual synchronous machine dynamic inertia control method based on artificial intelligence technology;

[0035] The system includes a data acquisition layer and a control layer;

[0036] The data acquisition layer is provided with multiple types of sensors for collecting current power grid status information; the data acquisition layer is connected to the control layer via an analog-to-digital conversion unit;

[0037] The control layer is provided with a heterogeneous multi-processor embedded computing platform for calculating control signals and transmitting the control signals to the voltage source inverter.

[0038] Optionally, the heterogeneous multi-processor embedded computing platform is a local collaborative architecture, including a digital signal processor and a field programmable gate array;

[0039] The digital signal processor is used for control and algorithm execution;

[0040] The field programmable gate array is used for hardware acceleration.

[0041] Optionally, a server is also included;

[0042] The server is connected to the control layer and is used for deploying and online updating of the temporal neural network.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The system structure of the present invention is simple, and the control and construction costs are low. Dynamically adjusted parameters are obtained through data timing processing and neural network application, which can realize adaptive adjustment of core parameters and solve the problem that traditional fixed parameters are difficult to cope with complex scenarios. Convolutional neural networks are used to participate in modulation signal calculations to effectively compensate for hardware phase delays and improve dynamic response accuracy. Deep reinforcement learning is then combined to achieve coordinated optimization of inertia control and dynamic response, reduce current overshoot and suppress harmonics, improve power quality and system stability, and significantly enhance the adaptability of virtual synchronous machines to dynamic changes in the power grid and the comprehensive control performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 is a system diagram in a specific embodiment of the present invention;

[0047] Figure 2 This is a diagram of a temporal neural network structure in a specific embodiment of the present invention;

[0048] Figure 3 is a structural diagram of a convolutional neural network in a specific embodiment of the present invention;

[0049] Figure 4 It is a simulation and experimental comparison diagram of the comprehensive optimization reward function of power quality in a specific embodiment of the present invention. Specific Embodiments

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0052] In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0053] It is worth noting that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products unless otherwise specified, and their sources are not specifically limited.

[0054] On the one hand, this embodiment provides a virtual synchronous machine dynamic inertia control method based on artificial intelligence technology, which mainly includes the following steps:

[0055] First, as Figure 1 shown in the system, the application scenario of this embodiment is a two-level voltage source inverter (2L-VSI) and a virtual synchronous generator (SVG) of the topology, which is used to realize dynamic inertia collaborative control.

[0056] Multidimensional grid information is collected in real time through sensors. The information includes grid frequency deviation, active power fluctuation, DC bus voltage change rate, and output current signal. The collected multidimensional grid information is preprocessed through a filtering algorithm for noise filtering, etc., to generate a time series data sequence of a preset time window.

[0057] Input the time series data sequence into the time series neural network optimized by transfer learning. The time series neural network is a two-layer long short-term memory network (LSTM), with 64 neurons in the bottom layer and 32 neurons in the top layer. Its transfer learning process includes: pre-training in the source domain based on multi-scenario simulation data of the IEEE 39-bus system, such as multi-scenario simulation data of photovoltaic fluctuations, load switching, short-circuit faults, etc., to complete the initial LSTM model; in the target scenario, such as when deploying an island microgrid, freeze the weights of the bottom layer network, and only fine-tune the parameters of the top fully connected layer using the measured data of the target scenario. For example, using 1% of the measured data in the target scenario to complete rapid fine-tuning, the error convergence speed can be increased by 60%.

[0058] Thus, construct the time series neural network structure as shown in Figure 2 which is specifically as follows:

[0059] (1) Input layer (Input);

[0060] Input features: time window data, length = 10ms; grid frequency deviation ; power fluctuation ; DC bus voltage change rate ; historical inertia parameter sequence ， ; Dimension: [batch_size, timesteps = 10, features = 4].

[0061] (2) LSTM layer;

[0062] The first LSTM layer (Layer1): number of neurons: 64; return sequence , that is, transfer the complete time series output to the next layer; activation function: tanh; transfer learning mark: freeze the weights, locked after pre-training in the source domain.

[0063] The second LSTM layer (Layer2): number of neurons: 32; return sequence: , that is, only output the result of the last time step; activation function: tanh; transfer learning mark: can be fine-tuned, update the weights when adapting to the target domain.

[0064] (3) Fully connected layer;

[0065] Feature fusion layer: number of neurons: 32 → 16; activation function: ReLU; input: output of the LSTM layer + real-time grid impedance Z grid , input from the outside;

[0066] Output layer: number of neurons: 3, corresponding to J , D , ; Activation function: linear, without an activation function, directly outputting the parameter value.

[0067] (4) Output layer (Output layer);

[0068] Output parameters: virtual inertia J , with the unit of kg•m²; damping coefficient D , with the unit of N•m•s / rad; virtual impedance , with the unit of Ω.

[0069] Through the above-mentioned time series neural network, dynamically adjustable virtual inertia J , damping coefficient D and virtual impedance can be output.

[0070] Among them, the input of the time series neural network is specifically the grid frequency deviation , power fluctuation , DC bus voltage change rate within a 10ms time window, as well as the historical inertia parameter sequence. The 10ms short sequence is suitable for the embedded platform to avoid excessive model complexity. The main considerations are real-time requirements: the system is implemented based on the embedded platform and requires low-latency processing, such as a 10ms window; grid frequency characteristics: 50Hz fundamental wave, and the dynamic fluctuations are usually within the sub-second level. A 1kHz sampling rate is sufficient to capture key harmonics and transient processes. Therefore, a 1kHz sampling rate, that is, 10 points, is a reasonable choice.

[0071] In addition, the time series neural network maintains adaptability through an online update strategy: periodically collecting multi-dimensional grid information and the historical inertia parameter sequence to form the current operating state dataset; applying a sliding window forgetting factor, such as selecting λ = 0.95 to process the dataset, giving the latest data a higher update weight, that is, a higher weight; while the historical data reduces the weight according to the exponential decay method to generate weighted training data; incrementally updating the network weights every 5 minutes based on the weighted training data to adapt to long-term changes such as equipment aging and seasonal characteristic drift.

[0072] Substitute the virtual inertia J and the damping coefficient D into the control module of the virtual synchronous machine to implement the core logic of active-frequency droop and reactive-voltage regulation; substitute the virtual impedance into the voltage-current control link of the virtual synchronous machine.

[0073] Under the parallel operation conditions of the above dynamic parameter operations, a convolutional neural network is used to extract the phase-frequency characteristics of the LC filter of the voltage source inverter, and a non-linear delay compensation mechanism is realized. The CNN adopts an asymmetric convolutional kernel structure (the kernel size of the previous layer is 5×1, and the kernel size of the latter layer is 3×1). The kernel weight distribution is determined by supervised training of the Hilbert transform phase difference. After generating the phase compensation angle, the pulse width modulation (PWM) signal is phase-corrected to compensate for the phase lag of the filter. Then, the harmonic content (THD) of the filtered voltage is monitored in real time. When THD>3%, it switches to the high-precision compensation mode, that is, a 0.1° resolution is adopted.

[0074] Furthermore, in this embodiment, a one-dimensional convolutional neural network (1D-CNN) is adopted, and the LC filter phase lag compensation network architecture formed by training of the 1D-CNN is as Figure 3 shown. From left to right in the figure, it is divided into four parts: input layer, convolutional layer, pooling / full connection layer, and output layer. Specifically, the architecture is as follows:

[0075] 1. Input layer;

[0076] 1.1 Input data:

[0077] Waveform: The difference waveform between the inverter output voltage and the filtered voltage , 256 points / cycle;

[0078] Dimension: 1×256 single-channel time series.

[0079] 2. Convolutional layer;

[0080] 2.1 The first convolutional layer (Conv1D-1): Number of convolutional kernels (kernel-1): 5; Kernel size: 5 lengths; Stride: 2; Activation function: ReLU; Output dimension: 5×126 feature maps;

[0081] 2.2 The second convolutional layer (Conv1D-2): Number of convolutional kernels (kernel-2): 10; Kernel size: 3 lengths. Stride: 1; Activation function: ReLU; Output dimension: 10×124.

[0082] 3. Pooling and full connection layer;

[0083] 3.1 Max pooling layer (MaxPool1D): Pooling size: 2; Stride: 2; Output dimension: 10×62; Flatten layer: Output dimension: 1×620 (10×62); Full connection layer (Dense1,2): Number of neurons: 64→16→1; Activation function: The last layer is linear, and the output phase compensation angle θ comp .

[0084] 4. Output Layer;

[0085] 4.1 Output: Phase Compensation Angle θ comp , in degrees, with an accuracy of 0.1°, which is used for leading compensation.

[0086] The data stream is labeled as follows:

[0087] a. Input → Conv1D-1: Waveform local feature extraction (such as high-frequency harmonics, phase mutation).

[0088] b. Conv1D-1 → Conv1D-2: Deep feature abstraction (such as phase delay trend).

[0089] c. Conv1D-2 → MaxPool: Dimensionality reduction to retain significant features.

[0090] d. Flatten → Dense: Mapping to the phase compensation amount.

[0091] The weight distribution of the above one-dimensional convolutional neural network is supervised and trained by the Hilbert transform phase difference to calculate the actual phase difference in real time, as the true value label for CNN training, ensuring that the compensation error ≤ 0.5.

[0092] Based on the Deep Q-Network (DQN) algorithm for multi-objective collaborative optimization, its network structure is a 6-dimensional input layer → hidden layer (64 neurons → 32 neurons, ReLU) → 5-dimensional Q-value output layer, and the training strategy of this network is: Prioritized Experience Replay (PER) + Double Q-Network (Double DQN) to avoid overestimation of Q values.

[0093] Furthermore, a corresponding reward function is set in the DQN training stage. Specifically, the reward function of the DQN set in this embodiment is:

[0094] ;

[0095] This reward function is the training function of the comprehensive power quality optimization reward function. The complete comprehensive power quality optimization reward function combines a frequency deviation reward term, a current overshoot penalty term, a harmonic distortion penalty term, and a storage health reward term. The comprehensive power quality optimization reward function formula is:

[0096] ;

[0097] In the formula, is the frequency deviation reward term, α is the frequency stability reward weight coefficient, which determines the contribution amplitude of the frequency deviation to the reward. Its role is in a weak power grid, such as when the short-circuit capacity ratio SCR < 2 or the frequency fluctuates violently. Increasing α can preferentially encourage the system to maintain frequency stability; β is the frequency deviation sensitivity coefficient, which controls the attenuation rate of the reward with the frequency deviation. When β > 0 and the value is larger, the negative impact of the frequency deviation on the reward is more significant, prompting the system to quickly suppress the frequency offset; is the current overshoot penalty term, is the output current overshoot in the dq coordinate system, that is, the deviation relative to the rated current, which characterizes the current fluctuation in the dynamic response. γ is the current overshoot penalty weight coefficient, which adjusts the degree of attention to current stability. Its role is when high-load switching or power mutation occurs, increasing its value can suppress the current peak and avoid the risk of device overcurrent; is the harmonic distortion penalty term, THD is the Total Harmonic Distortion, which measures the harmonic content of the output voltage / current. It can be set to trigger the penalty when THD > 3%. δ is the harmonic distortion penalty weight coefficient, which controls the optimization priority of the power quality. For example, in the grid-connected scenario, by increasing δ, it is ensured that THD meets the standard and harmonic pollution of the power grid is avoided; is the energy storage health reward term, is the energy storage health, which can be designed as a piecewise linear function or a Gaussian function to encourage the SOC to be maintained in a reasonable range, such as 30% - 90%, to avoid the impact of deep charge and discharge on the battery life, is the energy storage health weight coefficient, which adjusts the optimization priority of the energy storage life, and its typical value range is, for example, 0.1 - 0.3.

[0098] Among them, the weight of each index of the comprehensive power quality optimization reward function is dynamically adjusted according to the grid strength and load characteristics. Further, the aforementioned dynamic adjustment is the implementation method of scenario adaptation. Weak power grid: increase α, such as from 0.8 → 1.2, to strengthen the frequency stability priority and suppress low-frequency oscillation; High load: increase γ, such as from 0.5 → 0.8, to reduce the stress on the device caused by current overshoot; Harmonic-sensitive scenario: increase δ, such as from 0.3 → 0.6, to ensure that THD meets the standard. Refer to THD < 2.5% when connecting to the PV grid. The specific implementation method of automatic dynamic adjustment can refer to Pareto front optimization: obtain the Pareto Optimal optimal solution under different weight combinations through offline simulation, preset a multi-group (α, β, γ, δ) parameter library, and dynamically call it by DQN according to the real-time grid state (such as SCR value, load rate) to avoid manual parameter adjustment. Furthermore, conduct convergence verification on the comprehensive power quality optimization reward function of this embodiment, such as Figure 4 shown, draw the function convergence curve under the simulation environment and experimental environment conditions respectively. The experimental steady-state value is 0.48, and the simulation steady-state value is 0.3.

[0099] Thus, system state parameters of the voltage source inverter, such as the DC bus voltage, output current, etc., and grid operation state parameters, such as frequency deviation, power fluctuation, etc., are used as the inputs of the aforementioned trained deep learning network. By discretizing the action space, for example, dividing the continuous PWM duty cycle correction amount ΔD into a five-level action space of -5%, -2.5%, 0%, +2.5%, +5% with a step resolution of 0.5%; thus calculating and generating the PWM duty cycle correction amount, inputting it into the inverter PWM modulation strategy, and realizing the coordinated control of dynamic inertia and response speed.

[0100] On the other hand, this embodiment also provides Figure 1 a virtual synchronous machine dynamic inertia control system based on artificial intelligence technology as shown, which is used to implement the above control method, mainly including a data acquisition layer, a control layer and a server.

[0101] Among them, the data acquisition layer: Deploy multiple types of sensors, such as Hall voltage / current sensors: measure the DC bus voltage Vdc, the inverter output current iabc, the grid voltage vgrid, and a frequency detection module: real-time collect the grid frequency fgrid. Through the above sensors, the state information such as grid frequency, voltage, current, and DC bus voltage is collected in real time. After the sensor signals are converted into digital signals by a high-precision analog-to-digital conversion unit (ADC), they are transmitted to the control layer.

[0102] The control layer: Adopt a heterogeneous multi-processor embedded computing platform, and the DSP+FPGA local cooperation architecture adopted in this embodiment. Among them, the digital signal processor (DSP), such as TMS320F28335, is used to execute the core algorithm: run a two-layer LSTM network to generate dynamic virtual inertia, damping coefficient, and virtual impedance; optimize the PWM duty cycle correction amount through the DQN algorithm; realize the core control logic of the virtual synchronous machine. The field programmable gate array (FPGA), such as Xilinx Zynq, is responsible for hardware acceleration: preprocess the collected data, including normalization and sliding window processing; calculate the phase compensation angle in real time through a fixed 1D-CNN hardware accelerator; generate a PWM signal and embed the dead time, and integrate an overcurrent / overvoltage protection circuit to achieve nanosecond-level fault response. After receiving the data from the data acquisition layer, the FPGA performs preprocessing and caching, and realizes data transmission with the DSP through direct memory access (DMA).

[0103] The execution layer is the application object of this embodiment, and its main body is a two-level voltage source inverter (2L-VSI). The execution layer also includes a PWM module, an LC filter, and a topology. Among them, the topology also includes three-phase bridge arms (IGBT) and DC bus capacitors. The execution layer interacts with the grid / load and collects the connection point voltage through sensors v PCCSum current i grid 。

[0104] Server: Connected to the control layer through a high-speed communication interface, used to deploy the initial LSTM model and support online updates: Store multi-scenario simulation datasets for source domain pre-training; Receive the measured data of the target scenario uploaded by the control layer, and perform fine-tuning of the top-level network parameters; Periodically send the updated model weights to the DSP to achieve cross-scenario adaptation and long-term maintenance of the time-series neural network.

[0105] During system operation, the data acquisition layer provides real-time input for the control layer. The DSP and FPGA cooperate to complete AI algorithm inference and hardware control. The server supports model training and updates, forming a closed loop of "data acquisition - algorithm optimization - hardware execution - model iteration" to achieve dynamic inertia collaborative control of the virtual synchronous machine in a complex power grid environment.

[0106] Comparative Example 1: VSG control of photovoltaic energy storage system;

[0107] Application scenario: In a photovoltaic energy storage system, a 2L-VSI is connected to a 380V low-voltage distribution network as a virtual synchronous machine, and it needs to cope with photovoltaic output fluctuations, load mutations, and power grid frequency disturbances, while meeting the grid connection harmonic standard of THD < 3%.

[0108] Among them, the system configuration and hardware parameters of the system applying the method of the present invention in this comparative example are as shown in Table 1 below:

[0109] Table 1

[0110]

[0111] In addition, a traditional PI control group is set in the same environment, and simulation field tests are respectively carried out through two groups of events. The events are: Event 1, sudden drop in photovoltaic power (cloud cover), the light intensity drops by 70% within 0.5s, and the photovoltaic output drops from 50kW to 15kW; Event 2, power grid frequency drop (simulated fault), the power grid frequency drops from 50Hz to 48.5Hz and lasts for 200ms.

[0112] Thus, the experimental comparison results of traditional PI control and the control method of the present invention are obtained, as shown in Table 2 below:

[0113] Table 2

[0114]

[0115] Through comparison, it can be seen that the method and system of the present invention have the following advantages:

[0116] Multi-time scale collaboration: LSTM is responsible for generating inertia parameters at the 10 ms level, DQN optimizes the dynamic response at the 1 ms level, and CNN compensates for hardware delays at the μs level. Economy-performance balance: Based on a low-cost 2L-VSI, it achieves dynamic performance close to that of a 3L-NPC, saving more than 30% in hardware costs. Strong robustness: Combining transfer learning and online fine-tuning to adapt to non-ideal factors such as photovoltaic power output fluctuations and device aging.

[0117] Furthermore, the method of the present invention can be further extended to multiple scenarios such as wind power and energy storage. By adjusting the training data set and the weights of the reward function to adapt to different application requirements, it also shows that the method of the present invention has good multi-scenario generality.

[0118] Comparative example 2: Adaptive regulation of VSG under a weak grid;

[0119] Application scenario: A 2L-VSI virtual synchronous machine is connected to a weak grid with a short-circuit ratio SCR < 2, such as a remote island microgrid or a high-proportion new energy access area, which needs to cope with risks such as sudden changes in grid impedance, voltage fluctuations, and subsynchronous oscillations, while maintaining grid connection stability and power quality.

[0120] Among them, the system configuration and hardware parameters of the system applying the method of the present invention in this comparative example are as shown in Table 3 below:

[0121] Table 3

[0122]

[0123] In addition, a fixed group of traditional VSGs is set in the same environment, and both are simulated and field-tested through two groups of events respectively. The events are: Event 1, sudden change in grid impedance, such as line switching, the grid impedance steps from 2 mH to 8 mH to simulate line fault isolation; Event 2, background harmonics cause subsynchronous oscillation, and the amplitude of the 5th harmonic voltage in the grid background rises to 8%, triggering a 90 Hz subsynchronous oscillation.

[0124] Thus, the fixed of the traditional VSG and the experimental comparison results controlled by the method of the present invention are obtained, as shown in Table 4 below:

[0125] Table 4

[0126]

[0127] Through comparison, it can be seen that the method and system of the present invention have the following advantages:

[0128] Impedance self-adaptive ability: Dynamically adjust through LSTM , at Z gridIt remains stable when the change is ±300% without manual re-parameterization. Resonant active suppression: Integrate CNN harmonic lead compensation detection and notch filtering, and the subsynchronous oscillation decay time < 100 ms. Cross-scenario compatibility: Adversarial training enables the model to adapt to unseen impedance topologies, such as looped microgrids and radial distribution networks.

[0129] Furthermore, the method of the present invention can be further extended to the scenario of multiple VSGs in parallel with a weak grid. Through the Consensus Algorithm, it coordinates each unit's and J to avoid circulating current and oscillation.

[0130] In summary, through the comparison of ratios, the method of the present invention has the following improvements in terms of performance: the dynamic response speed is increased by ≥ 30%, and the error of the inertia support effect is ≤ 5%; and under the same THD requirement, the volume of the LC filter is reduced by 20%. Further, in terms of economy, it realizes the performance of the multilevel topology based on the low-cost 2L-VSI hardware, and through the lightweight design of the model, it does not require an additional AI acceleration chip. Moreover, in terms of generalization ability, transfer learning supports cross-scenarios, such as photovoltaic, wind power, and energy storage, and can be quickly adapted, reducing the training data requirement by 80%.

[0131] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than limiting the protection scope of the present invention. Any simple modification or equivalent replacement of the technical solution of the present invention by those of ordinary skill in the art does not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A dynamic inertia control method for a virtual synchronous machine based on artificial intelligence technology, which is applied to a voltage source inverter control system, and is characterized in that: It includes the following steps: Collect multi-dimensional power grid information in real time; preprocess the multi-dimensional power grid information and generate a time series data sequence with a preset time window; Input the time series data sequence into a time series neural network, and output dynamically adjusted virtual inertia, damping coefficient and virtual impedance; Substitute the virtual inertia and the damping coefficient into the control module of the virtual synchronous machine, and substitute the virtual impedance into the voltage and current control link of the virtual synchronous machine; Use a convolutional neural network to extract the phase-frequency characteristics of the filter of the voltage source inverter and generate a phase compensation angle; correct the phase of the pulse width modulation signal used to drive the voltage source inverter through the phase compensation angle; Train a deep reinforcement learning network based on an integrated power quality optimization reward function, use the system state parameters of the voltage source inverter and the grid operation state parameters as the input quantities of the deep reinforcement learning network, and output the duty cycle correction amount of the pulse width modulation signal used to drive the voltage source inverter; input the duty cycle correction amount into the pulse width modulation strategy of the voltage source inverter to co-control the dynamic inertia of the virtual synchronous machine.

2. The dynamic inertia control method of the virtual synchronous machine based on artificial intelligence technology according to claim 1, wherein: The time series neural network is synchronously updated through an online update strategy; The online update strategy includes the following steps: Periodically collect the multi-dimensional power grid information and the historical inertia parameter sequence, and form a current operating state data set; Apply a sliding window forgetting factor to the current operating state data set and obtain a training data set; In the training data set, assign an update weight to the latest data, and reduce the weight of the historical data in the data set by the exponential decay method to generate weighted training data; Periodically perform incremental updates on the weights of the time series neural network based on the weighted training data.

3. The dynamic inertia control method of the virtual synchronous machine based on artificial intelligence technology according to claim 1, characterized in that: The time series neural network is a long short-term memory network and adopts a double-layer long short-term memory structure; In the double-layer long short-term memory structure, the bottom layer has 64 neurons and the top layer has 32 neurons.

4. The virtual synchronous machine dynamic inertia control method based on artificial intelligence technology according to claim 1 or 3, characterized in that: The time series neural network is optimized through transfer learning after being deployed in the corresponding scenario; The transfer learning includes source domain pre-training based on multi-scenario simulation data and target domain fine-tuning using the measured data of the target scenario to update the top network parameters.

5. The dynamic inertia control method of the virtual synchronous machine based on artificial intelligence technology according to claim 1, characterized in that: The multi-dimensional power grid information includes grid frequency deviation, active power fluctuation, DC bus voltage change rate and output current signal; The preprocessing at least includes noise filtering processing of the collected multi-dimensional power grid information through a filtering algorithm.

6. The dynamic inertia control method of a virtual synchronous machine based on artificial intelligence technology according to claim 1, characterized in that: The convolutional neural network is a one-dimensional convolutional neural network, adopts an asymmetric convolutional kernel structure, and the kernel weight distribution is obtained through phase difference supervised training.

7. The dynamic inertia control method of a virtual synchronous machine based on artificial intelligence technology according to claim 1, wherein: The deep reinforcement learning network is a deep Q network; The parameters of the integrated power quality optimization reward function are at least provided with a frequency deviation reward term, a current overshoot penalty term and a harmonic distortion penalty term, and the index weights of each item are dynamically set.

8. A virtual synchronous machine dynamic inertia control system based on artificial intelligence technology, characterized in that: For implementing the dynamic inertia control method of the virtual synchronous machine based on artificial intelligence technology according to any one of claims 1-7; The system includes a data acquisition layer and a control layer; The data acquisition layer is provided with multiple types of sensors for collecting the status information of the current power grid; the data acquisition layer is connected to the control layer through an analog-to-digital conversion unit; The control layer is provided with a heterogeneous multi-processor embedded computing platform for calculating control signals and transmitting the control signals to the voltage source inverter.

9. The virtual synchronous machine dynamic inertia control system based on artificial intelligence technology according to claim 8, characterized in that: The heterogeneous multi-processor embedded computing platform is a local collaborative architecture, including a digital signal processor and a field programmable gate array; The digital signal processor is used for control and algorithm execution; The field programmable gate array is used for hardware acceleration.

10. The virtual synchronous machine dynamic inertia control system based on artificial intelligence technology according to claim 8 or 9, characterized in that: It also includes a server; The server is connected to the control layer and is used for deploying and online updating the temporal neural network.

Images