Virtual power plant rapid frequency modulation control system and control method

Abstract

The invention discloses a virtual power plant rapid frequency modulation control system and control method, and the method comprises the steps: adaptively adjusting a sampling period through a dynamic sampling algorithm based on a sliding window variance; low-delay calculation of high-frequency data is realized by adopting a multi-thread queue communication technology; a multi-core ARM processor running a real-time operating system is used for executing a multi-thread control model, and data acquisition, algorithm control, fault diagnosis and communication tasks are processed in parallel; according to the control algorithm, frequency prediction feedforward signals based on an AI model and SOC state feedback are fused, and dynamic optimization of PID parameters is achieved. According to the method, millisecond-level frequency modulation response can be realized, the instruction response delay is less than 80ms, the frequency adjustment error is less than + / -0.01 Hz, the frequency modulation performance of a virtual power plant and the stability of a power grid are remarkably improved, and meanwhile, the service life of energy storage equipment is prolonged through SOC collaborative optimization. The invention aims to solve the problems of slow response and low precision of the existing frequency modulation technology.

Description

Technical Field

[0001] This invention relates to the field of power system automation and control technology, and in particular to a fast frequency regulation control system and control method for use in a virtual power plant (VPP). Background Technology

[0002] With the increasing proportion of new energy power generation in the power grid, its inherent volatility and intermittency pose a severe challenge to the frequency stability of the grid. Virtual power plants, as an effective way to aggregate distributed energy resources, energy storage systems, and controllable loads, have shown great potential in participating in grid ancillary services, particularly frequency regulation. However, existing frequency control technologies have significant shortcomings.

[0003] Existing PID and fuzzy control schemes suffer from problems such as slow response, fixed parameters, and poor adaptability to dynamic scenarios.

[0004] Furthermore, existing energy storage controllers mostly employ a one-way command transmission architecture, which cannot achieve efficient two-way information interaction with virtual power plant platforms. Their data sampling period is generally on the order of 500ms, resulting in significant delays in frequency compensation and failing to fully leverage the rapid response advantages of energy storage devices. Summary of the Invention

[0005] To address the shortcomings of the aforementioned background technology, this invention aims to provide a fast frequency regulation control system and method for virtual power plants with fast response speed, high control accuracy, and strong adaptability, so as to meet the needs of modern power systems for fast and accurate frequency regulation.

[0006] To achieve the above objectives, the present invention provides a fast frequency regulation control method for a virtual power plant, comprising the following steps: Step S1: Real-time acquisition of power grid frequency data is carried out using a dynamic sampling algorithm. The dynamic sampling algorithm is based on the time-varying characteristics of the power grid frequency deviation Δf and adaptively adjusts the sampling period Ts through the sliding window variance σ². Step S2: The acquired power grid frequency data and control tasks are processed in parallel using a multi-threaded control model, which includes at least a high-frequency data acquisition thread, a control algorithm thread, a fault diagnosis thread, and a data interaction thread. Step S3: The control algorithm thread executes a composite control algorithm combining PID and feedforward, and integrates the frequency prediction results based on the artificial intelligence model to calculate the target output command of the energy storage system; Step S4: The data interaction thread communicates bidirectionally with the virtual power plant master station through the industrial Ethernet protocol, uploads the controller status, and receives scheduling instructions.

[0007] The present invention also provides a virtual power plant fast frequency regulation control system, comprising: Data acquisition and preprocessing unit; performs digital filtering on the acquired signals and extracts the real-time power grid frequency; The central processing unit is implemented using a multi-core ARM processor and runs a real-time operating system. The communication interface unit includes a gigabit Ethernet interface and a 4G / 5G wireless communication module, which is used to support data interaction with the virtual power plant master station and energy storage devices.

[0008] Compared with the prior art, the present invention has the following significant advantages: 1. Fast response speed and high control precision: Through FPGA hardware acceleration and dynamic sampling algorithm, combined with multi-threaded parallel processing, the end-to-end latency from data acquisition to command issuance is less than 80ms, and the frequency adjustment error can be controlled within ±0.01Hz, which is nearly an order of magnitude higher than the traditional solution.

[0009] 2. High intelligence and adaptability: The introduction of AI models to predict frequency enables feedforward control and effectively suppresses disturbances. At the same time, the SOC-frequency co-optimization strategy can ensure the frequency regulation effect while taking into account the health status of the energy storage system and extending its cycle life by about 15%.

[0010] 3. High system reliability: The controller adopts hardware redundancy design and fast fault isolation mechanism, and the key communication links support hot-swappable switching with a switching time of less than 50ms, ensuring stable operation of the system in complex electromagnetic environments.

[0011] 4. Advanced architecture and good scalability: Based on the distributed architecture of "controller-device interface layer" and standardized communication protocol, it is easy to integrate distributed resources of different types and scales, and supports the flexible expansion of future virtual power plants. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the hardware architecture of the virtual power plant fast frequency regulation control system of the present invention.

[0013] Figure 2 This is a timing flowchart of the dynamic sampling algorithm for fast frequency regulation control of the virtual power plant in this invention.

[0014] Figure 3 This is a schematic diagram of multi-threaded task scheduling for the fast frequency regulation control of the virtual power plant in this invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] A fast frequency regulation control method for a virtual power plant includes: First, real-time acquisition of grid frequency data using a dynamic sampling algorithm. This algorithm, based on the time-varying characteristics of frequency deviation, adaptively adjusts the sampling period through a sliding window variance, increasing the sampling rate during periods of drastic frequency fluctuations and decreasing it during periods of frequency stability, thus balancing response speed and computational efficiency. Second, parallel processing of various tasks using a multi-threaded control model. This model allocates tasks such as high-frequency data acquisition, control algorithm execution, fault diagnosis, and data interaction to different threads, and binds independent CPU cores to core tasks to ensure the real-time performance of critical tasks. Simultaneously, the algorithm incorporates the state of charge (SOC) of energy storage as feedback, dynamically adjusting PID parameters to achieve coordinated optimization of frequency regulation and energy storage lifespan management. Finally, efficient and reliable bidirectional communication with the virtual power plant master station is established via a standard industrial Ethernet protocol (such as IEC 61850).

[0017] Preferably, when the sliding window variance σ² is greater than a preset fluctuation threshold, a fast sampling mode is triggered, shortening the sampling period Ts to 50ms; when σ² is less than or equal to the preset fluctuation threshold, a standard or slow sampling mode is adopted according to the absolute value of Δf.

[0018] Preferably, the high-frequency data acquisition thread has the highest running priority and is used to read preprocessed frequency, voltage and current data from the FPGA hardware logic; the energy storage state of charge is introduced as feedback to dynamically adjust the PID control parameters and realize the coordinated optimization of the energy storage state of charge and frequency, so as to avoid overcharging or over-discharging of the energy storage system.

[0019] Preferably, the frequency prediction based on the artificial intelligence model uses a long short-term memory network to predict the power grid frequency trend in the next 100 milliseconds, and the prediction result is used as a feedforward signal to be input into the composite control algorithm.

[0020] Preferably, an Internet of Things (IoT) architecture is adopted to enable two-way communication between cloud-based command issuance and controller-collected data upload. This invention also provides a virtual power plant fast frequency regulation controller for implementing the above method. The controller adopts a distributed hardware architecture, with its core comprising a data acquisition unit, a hardware preprocessing unit, a central processing unit, and a communication interface unit. The hardware preprocessing unit uses an FPGA chip, leveraging its parallel computing capabilities to perform hardware-level digital filtering and FFT transformation on the acquired raw electrical signals, calculating a high-precision real-time frequency within 1ms. The central processing unit uses a multi-core ARM processor running a real-time operating system, providing a high-performance computing platform for the reliable operation of the multi-threaded control model. The communication interface unit integrates wired and wireless communication methods, ensuring communication redundancy and reliability.

[0021] Reference Figure 1The virtual power plant fast frequency regulation controller described in this invention adopts a distributed architecture of "controller-device interface layer". The core of the controller mainly consists of an FPGA chip (such as the Xilinx Zynq UltraScale+ series) and an ARM Cortex-A53 quad-core processor. The FPGA, as a hardware preprocessing unit, is responsible for high-speed sampling of the grid bus voltage and current through a 24-bit high-precision ADC, and performs FIR digital filtering and 1024-point FFT operations using its internal integrated hardware logic unit. The hardware processing latency is less than 1ms, thereby quickly and accurately extracting the fundamental frequency (50Hz) and harmonic components and calculating the real-time frequency deviation Δf. The ARM processor, as the central processing unit, runs the RT-Linux real-time operating system and carries the subsequent multi-threaded control model. The two communicate via an AXI4-Stream high-speed bus with a bandwidth of up to 1Gbps. The controller is also equipped with 4GB of DDR4 memory (with ECC check) for caching process data. The communication interface layer includes a Gigabit Ethernet PHY chip (such as the TI DP83867) supporting the IEC 61850-9-2LE protocol and a 5G / 4G dual-mode wireless module, enabling wired / wireless dual-backup communication with the virtual power plant master station. In addition, it integrates an optocoupler-isolated DI / DO module for direct control of the energy storage PCS (Power Conversion System).

[0022] Software Algorithm: The core control method of this invention is implemented in software on an ARM processor, mainly including a dynamic sampling algorithm and a multi-threaded control model.

[0023] Dynamic sampling algorithm: refer to Figure 2 The implementation process of this algorithm is as follows: 1. Data preprocessing (FPGA execution): Perform 32nd order FIR low-pass filtering on the raw sampled data to remove high-frequency noise, and then calculate the real-time frequency deviation Δf = f_measured - 50Hz using the FFT algorithm.

[0024] 2. Sliding Window Variance Detection (ARM Execution): The software maintains a sliding window containing the most recent N (e.g., N=10) Δf values ​​and calculates the variance σ² of the data within the window. When σ² is greater than a preset threshold (e.g., 0.01Hz²), it indicates severe fluctuations in the power grid frequency. At this time, the system triggers a fast sampling mode, dynamically adjusting the ADC sampling trigger signal through the ARM's PWM timer to adjust the sampling period Ts to 50ms. If σ² is lower than the threshold, the system recovers to the standard sampling period (e.g., 100ms) or a slower period based on the absolute value of Δf to reduce the system's computational load.

[0025] Multithreaded control model: Reference Figure 3To ensure real-time control, the system adopts a multi-threaded parallel processing architecture and binds different tasks to different CPU cores.

[0026] Thread 1 (CPU0, priority 98): High-frequency data acquisition thread. Responsible for reading processed data such as Δf, voltage, and current from the FPGA through a lock-free circular buffer and updating them to shared memory.

[0027] Thread 2 (CPU1, priority 95): Control algorithm thread. This thread is triggered by a semaphore. Once new data is ready, it immediately executes the PID + feedforward composite control algorithm to calculate the energy storage output command P_ref. Its control equation is: P_ref = Kp *Δf + Ki * ∫Δf dt + Kd * d(Δf) / dt + P_feedforward. The feedforward term P_feedforward is provided by the AI ​​prediction model.

[0028] Thread 3 (CPU2, priority 90): Fault diagnosis thread. Real-time monitoring of sensor and communication link status, achieving rapid fault location (<30ms) and redundancy switching through methods such as "heartbeat detection + feature value comparison".

[0029] Thread 4 (CPU3, priority 85): Data interaction thread. Responsible for bidirectional communication with the virtual power plant platform via the IEC 61850 MMS protocol, uploading status data and receiving remote dispatch commands.

[0030] In the diagram, H represents constant power mode (Hold); J represents planned value mode (Joint); D represents dispatch command mode (Dispatch); Ts represents sampling time or control time step; FPGA stands for Field Programmable Gate Array. In energy storage / power systems, it is commonly used for high-speed sampling, real-time control, PWM driving, and communication interface processing. p_ref represents reference active power command; P_actual represents actual active power; PCS represents energy storage converter. The virtual power plant fast frequency regulation controller of this invention adopts a distributed architecture of "controller-device interface layer". Each controller interacts with lower-level devices through a dedicated bus and protocol interface to ensure low latency, high reliability, and real-time performance. The overall system design has strong anti-interference capabilities and is suitable for complex power environments. The specific product architecture includes: 1. Controller 1.1 Core Components: 1.1.1. FPGA chip (Xilinx Zynq UltraScale+): integrates approximately 280,000 logic units (LUTs), supports 1024-point FFT hardware acceleration (transformation time <1μs) and FIR filtering function.

[0031] 1.1.2 ARM Cortex-A53 quad-core processor: 1.8GHz clock speed, running the real-time operating system RT-Linux, supporting multi-threaded priority scheduling and task isolation.

[0032] 1.1.3 DDR4 memory (4GB, with ECC check): ensures no storage errors during high-frequency data processing and supports 1 hour of continuous data caching.

[0033] 1.2 Technical Features: 1.2.1. FPGA hardware acceleration for FFT and FIR filtering is employed, along with multi-threaded computation using an ARM multi-core processor, to achieve low-latency data processing. The FPGA and ARM communicate at high speed via the AXI4-Stream bus, with a data bandwidth ≥1Gbps, hardware preprocessing latency ≤1ms, and hardware timer accuracy reaching 10 ns, thereby ensuring real-time capture of grid frequency changes and rapid power regulation.

[0034] 1.2.2. Integrate temperature and voltage monitoring modules to achieve fault self-diagnosis with a diagnostic coverage rate of ≥95%.

[0035] 1.2.3. The sampling period and power adjustment strategy are adaptively adjusted according to the grid frequency fluctuations. The sliding window variance detection triggers fast sampling, which ensures the system responds quickly when the frequency fluctuations are severe, and reduces the sampling frequency in a stable state to reduce the computational load and energy consumption.

[0036] 2. Communication Interface Layer 2.1 Core Components: 2.1.1. Gigabit Ethernet PHY chip (TI DP83867): Supports IEC 61850-9-2LE protocol, end-to-end transmission latency <20ms.

[0037] 2.1.2.4G / 5G Dual-Mode Module (Huawei ME909s-821): Supports 5G SA mode, downlink speed ≥300Mbps, used for wireless backup communication.

[0038] 2.1.3. Optocoupler isolated DI / DO module: provides 8 digital inputs (for acquiring the status of the energy storage PCS) and 8 digital outputs (for controlling the switching of energy storage charging and discharging), with an isolation voltage ≥2500V.

[0039] 2.2 Technical Features: 2.2.1. The Ethernet interface supports IEEE 1588 PTP precise synchronization, and the wireless communication module supports VPN encrypted transmission (AES-256 algorithm).

[0040] 2.2.2. All external interfaces are protected by surge protectors (compliant with IEC 61000-4-5 standard) to withstand ±4kV differential mode and ±8kV common mode surges.

[0041] 3. Topology Description: 3.1. Sensing layer → Computing layer: ADC sampling and filtering data are transmitted to the FPGA via LVDS bus (1Gbps rate), and the FPGA clock is synchronized with the timing module via PTP protocol.

[0042] 3.2. Inside the computing layer: The data preprocessed by the FPGA is transmitted to the ARM processor through the AXI4-Stream bus. The ARM and DDR4 memory interact through a 64-bit bus (bandwidth 25.6GB / s) to achieve high-speed data processing.

[0043] 3.3. Computing Layer → Communication Layer: The ARM sends IEC 61850 protocol data to the virtual power plant platform through the Ethernet interface; the wireless module and DI / DO module are directly connected to the ARM through USB / GPIO to enable low-latency control commands to be sent to the energy storage PCS.

[0044] 3.4. Communication Layer → Cloud: Gigabit Ethernet and 5G modules are accessed in parallel to achieve wired / wireless dual backup data upload.

[0045] The reliability design of the virtual power plant fast frequency regulation controller of the present invention includes: 1. Anti-interference design: The sensing layer adopts a differential input circuit and is equipped with a metal shielding shell to suppress common-mode interference, ensuring a common-mode rejection ratio (CMRR) ≥ 120 dB. The power module integrates an EMI filter conforming to EN 61000-6-2 standard, effectively reducing the impact of grid noise on system sampling and data processing, thereby improving the system's anti-interference capability.

[0046] 2. Hardware-level real-time performance: The FPGA has a built-in hardware timer with a timing accuracy of 10 ns, ensuring that the data preprocessing delay is ≤1 ms. The ARM processor responds to the sampling trigger signal first through the interrupt binding (IRQ) mechanism, realizing fast acquisition and real-time processing of sampled data.

[0047] 3. Redundancy Design: Key sensors and communication links employ a dual-backup configuration, supporting hot-swappable switching with a switching time of <50 ms, thus ensuring uninterrupted critical data acquisition and communication. The power supply module adopts a 1+1 redundant power supply structure to ensure continuous and stable system operation.

[0048] The virtual power plant fast frequency regulation control method of the present invention includes the following process: 1. Implementation process of dynamic sampling algorithm 1.1. Core Logic: Based on the time-varying characteristics of the power grid frequency deviation (Δf), the sampling period (Ts) is adaptively adjusted through the sliding window variance. The specific steps are as follows: 1.2. Data Preprocessing (FPGA Execution): 1.2.1. Perform FIR low-pass filtering (cutoff frequency 50Hz, order 32) on the original sampled data to remove high-frequency noise.

[0049] 1.2.2. The fundamental frequency component (50Hz±0.5Hz) is extracted using the FFT algorithm, and the real-time frequency deviation Δf = f_measured - 50Hz is calculated.

[0050] 1.3. Sliding Window Variance Detection (ARM Execution): Maintain a sliding window containing the 10 most recent Δf values ​​and calculate the window variance σ²:

[0051] Where N is the sliding window length (number of sample points), Δf i Let i be the change at the i-th sampling point. This represents the average value of Δf within the window. When σ² > 0.01Hz² (i.e., Δf fluctuates wildly), a fast sampling mode is triggered (Ts = 50ms); otherwise, the mode is switched based on the absolute value of Δf.

[0052] 1.4. Sampling period switching mechanism: The ADC sampling trigger signal is dynamically adjusted through the ARM's PWM timer, with a switching delay of ≤10ms; "data zero-padding interpolation" is used during switching to avoid signal breakage and ensure control continuity.

[0053] 2. Multi-threaded control model

[0054] H = Hold mode, J = Joint mode, D = Dispatch mode, Ts: usually represents sampling time or time step. FPGA: Field Programmable Gate Array. In energy storage / power systems, it is often used for high-speed sampling, real-time control, PWM driving, and communication interface processing. p_ref: generally represents reference active power command, P_actual: actual active power; PCS = energy storage converter.

[0055] This invention employs the following mechanism: 1. High-speed sampling mechanism 1.1. Sampling Object: 1.1.1. Grid-side data: bus voltage, current, real-time frequency, power factor.

[0056] 1.1.2. Energy storage system data: battery SOC (State of Charge), SOH (State of Health), terminal voltage, current, and temperature.

[0057] 1.1.3. PCS (Converter) Data: Active Power, Reactive Power, Output Current Waveform.

[0058] 1.2. Data collected: 1.2.1. Instantaneous quantities: Instantaneous waveforms of voltage and current, used for frequency and harmonic analysis.

[0059] 1.2.2. Statistics: Active / Reactive Power, Frequency Deviation, Load Fluctuation.

[0060] 1.2.3. State variables: battery SOC, SOH, charge / discharge capacity.

[0061] 1.3. High-speed performance: 1.3.1. At the hardware level, sampling is performed using FPGA, with a minimum sampling period of 50ms.

[0062] 1.3.2. Data is preprocessed in real time within the ARM multi-core processor (filtering, noise reduction, fast Fourier transform FFT).

[0063] 1.3.3. Ensure that the end-to-end delay from sampling to data entry / processing is ≤10ms.

[0064] 2. Multi-source data fusion mechanism: The FPGA performs spatiotemporal alignment (based on hardware timestamp) of sensor data (frequency, voltage) with scheduling instructions (such as SOC limits) issued by the virtual power plant platform, and eliminates measurement noise through Kalman filtering. The accuracy of the fused data is improved to ±0.003Hz.

[0065] 3. SOC-Frequency Co-optimization: The control algorithm introduces energy storage state of charge (SOC) feedback and dynamically adjusts PID parameters: when SOC < 20%, the Kp gain is reduced by 20% to avoid over-discharge and extend cycle life by about 15%.

[0066] 4. Rapid fault isolation: Thread 3 uses "heartbeat detection + feature value comparison" (e.g., Δf mutation rate > 0.5Hz / s to determine sensor fault) to achieve fault location time < 30ms, and restores control capability after switching to redundant modules.

[0067] 5. AI prediction mechanism 5.1. Prediction Method: 5.1.1. Time series data obtained through high-speed sampling (such as frequency deviation Δf, current fluctuations, etc.) are input into AI prediction models (LSTM / RNN / lightweight Transformer).

[0068] 5.1.2. The model predicts frequency trends and load trends for the next 0.5–2 seconds.

[0069] 5.1.3. The prediction results are used to allocate energy storage output in advance to achieve feedforward regulation.

[0070] 5.2 Combination method: 5.2.1. After the sampled data enters the preprocessing module, it is divided into a real-time control channel (millisecond-level response) and an AI prediction channel (second-level trend prediction).

[0071] 5.2.2. Real-time data is used to directly trigger edge control actions; predictive data serves as an optimization signal to adjust the control strategy and power allocation.

[0072] 5.3. Multi-threaded collaborative AI prediction: 5.3.1. Sampling Thread: Responsible for data acquisition from different devices, ensuring parallel processing and non-blocking.

[0073] 5.3.2. Preprocessing thread: Performs filtering, FFT and outlier removal.

[0074] 5.3.3. AI Prediction Thread: Runs a lightweight AI model independently to predict future frequency and power requirements.

[0075] 5.3.4. Control Thread: Dynamically issues adjustment instructions by combining real-time data and prediction results.

[0076] With its multi-threaded architecture, the system can complete the control loop in milliseconds while running prediction and optimization tasks in parallel, avoiding latency buildup.

[0077] 5.4. Edge control collaborative AI prediction: 5.4.1. Edge-related work content: Data sampling and preprocessing are performed at the edge (FPGA / ARM layer).

[0078] Real-time alarms and policy enforcement at the edge (local EMS controller).

[0079] Fast frequency modulation and power distribution at the edge (based on SOC / SOH + prediction signal) 5.4.2. Relationship with AI prediction: AI predictions provide lead time information (trend power demand) as feedforward input for edge control strategies.

[0080] Edge control executes real-time actions (feedback control) in milliseconds and combines with AI prediction to form a feedforward + feedback hybrid control.

[0081] Experimental data: index Background Technology (PID Control) This invention Increase Frequency adjustment error (Hz) 0.082 0.009 89%↓ Response latency (ms) 210 73 65%↓ Energy storage cycle efficiency 92.5% 96.8% 4.3%↑ Theoretical advantages: Dynamic sampling reduces redundant data by 70%, lowering the load on edge computing.

[0082] Hardware-level timestamp synchronization (±1μs accuracy) solves the problem of multi-source data alignment.

[0083] It supports SOC-frequency co-optimization, extending energy storage life by approximately 15%.

[0084] Improve the speed and accuracy of power grid frequency regulation response.

[0085] Through intelligent power distribution and multi-device collaboration, millisecond-level frequency regulation response is achieved, significantly improving grid stability and the capacity for renewable energy absorption.

[0086] Enhance the intelligence and automation level of virtual power plants.

[0087] It supports automatic device discovery, configuration, data collection, and alarms, reducing operation and maintenance costs and improving the system's intelligence level.

[0088] Ensure data security and system reliability.

[0089] Multi-level storage, encrypted communication, and single-instance operation mechanisms ensure data integrity and system security.

[0090] Supports large-scale device access and flexible expansion.

[0091] The dynamic table structure and modular design support the access of multiple types of devices and functional expansion, adapting to the future development needs of virtual power plants.

[0092] Achieve efficient synchronization between cloud and local policies.

[0093] Cloud-based policies and alarms can be synchronized to the local device in real time, enhancing the remote management and intelligent scheduling capabilities of the virtual power plant.

[0094] In summary, this invention achieves high efficiency, intelligence, security, and scalability of a virtual power plant fast frequency control controller through multi-level time-series data storage, intelligent frequency regulation algorithm, asynchronous queue processing, and cloud synchronization mechanism, demonstrating significant technological innovation and application value.

[0095] AI Prediction Mechanism: The system utilizes a dedicated AI prediction thread to run a lightweight LSTM or RNN model. This model uses historical frequency deviation time-series data to predict frequency trends for the next 0.5-2 seconds. The prediction result is fed forward as a feedforward signal P_feedforward into the control algorithm thread, enabling the controller to respond in advance to impending frequency fluctuations, thus achieving a "prediction-control" closed loop.

[0096] 2. SOC-Frequency Co-optimization: When calculating P_ref, the control algorithm thread obtains the SOC value of the energy storage system in real time. When the SOC is below the safety threshold (e.g., 20%), the proportional gain Kp in the PID controller is dynamically reduced to decrease the depth of discharge; when the SOC is above the safety threshold (e.g., 90%), the charging power is reduced. This co-optimization strategy effectively avoids overcharging and over-discharging of the energy storage system, and tests have shown that it can extend its cycle life by approximately 15%.

[0097] Implementation Case: This invention was applied in a virtual power plant equipped with a 2MW / 4MWh lithium iron phosphate energy storage system. When a frequency disturbance of ±0.5Hz occurs in the grid, the controller's initial sampling period is 100ms. When the detected frequency deviation Δf reaches 0.25Hz and the sliding window variance exceeds the threshold, the controller automatically switches the sampling period to 50ms. The AI ​​prediction model predicts a further frequency drop 0.8 seconds in advance. The control algorithm thread combines the prediction result and the real-time Δf to calculate the target power P_ref=1.35MW and sends a command to the PCS through the DI / DO module. The entire process has a response delay of 73ms, and the frequency regulation error eventually stabilizes at 0.009Hz. After the regulation task is completed, the controller automatically returns to the 100ms baseline sampling period.

[0098] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A virtual power plant fast frequency regulation control method, characterized in that, The method comprises the following steps: Step S1: Real-time collection of power grid frequency data using a dynamic sampling algorithm, which is based on the time-varying characteristics of power grid frequency deviation Δf and adaptively adjusts the sampling period Ts through sliding window variance σ²; Step S2: Parallel processing of the collected power grid frequency data and control tasks using a multi-thread control model, which at least includes a high-frequency data collection thread, a control algorithm thread, a fault diagnosis thread, and a data interaction thread; Step S3: The control algorithm thread executes a compound control algorithm combining PID and feedforward, and fuses the frequency prediction results based on an artificial intelligence model to calculate the target output instruction of the energy storage system; Step S4: The data interaction thread communicates with the virtual power plant master station through an industrial Ethernet protocol for bidirectional communication, uploads the controller status, and receives scheduling instructions.

2. The frequency modulation control method according to claim 1, characterized by, In step S1, when the sliding window variance σ² is greater than a preset fluctuation threshold, a fast sampling mode is triggered, and the sampling period Ts is shortened to 50 ms; when σ² is less than or equal to the preset fluctuation threshold, a standard or slow sampling mode is used according to the absolute value of Δf.

3. The frequency modulation control method according to claim 1, characterized by, In step S2, the high-frequency data collection thread has the highest running priority and is used to read the preprocessed frequency, voltage, and current data from the FPGA hardware logic; the energy storage state of charge is introduced as feedback to dynamically adjust the PID control parameters, realizing the coordinated optimization of the energy storage state of charge and the frequency to avoid overcharging or overdischarging of the energy storage system.

4. The frequency modulation control method according to claim 1, characterized by, In step S3, the frequency prediction based on the artificial intelligence model uses a long short-term memory network to predict the future 100 ms trend of the power grid frequency and inputs the prediction results as feedforward signals to the compound control algorithm.

5. The frequency modulation control method according to claim 1, characterized by, An IOT architecture is used to realize the bidirectional communication function of cloud instruction issuance and controller data collection uploading.

6. A virtual power plant fast frequency regulation control system for performing the virtual power plant fast frequency regulation control method according to any one of claims 1 to 5, characterized by, It comprises: a data collection and preprocessing unit; digital filtering of the collected signals to extract real-time power grid frequency; a central processing unit implemented using a multi-core ARM processor and running a real-time operating system; a communication interface unit including a gigabit Ethernet interface and a 4G / 5G wireless communication module for supporting data interaction with the virtual power plant master station and energy storage devices.

7. The virtual power plant fast frequency regulation control system of claim 6, wherein The algorithm is automatically converted into C code using MATLAB Coder and transplanted into an embedded system for running to complete real-time signal data preprocessing.

8. The virtual power plant fast frequency regulation control system of claim 6, wherein An AI model file is generated after training a high-frequency, real-time, and unstable time series model using artificial intelligence.

9. The virtual power plant fast frequency regulation control system of claim 8, wherein The AI model file is called by a frequency modulation program to make an advance prediction of power grid frequency changes, realizing more timely and smooth frequency modulation.

10. The virtual power plant fast frequency regulation control system of claim 6, wherein The communication interface unit supports the IEC 61850 communication protocol and uses the IEEE 1588 precise time synchronization protocol to realize time synchronization of the entire system.

Images