Flywheel energy storage active phase modulation control method, system, equipment and medium

Abstract

The invention discloses a flywheel energy storage active phase modulation control method, system and device and a medium, and the method comprises the following steps: collecting multi-source power data of a power grid, and obtaining preprocessed phasor data; inputting the preprocessed phasor data into a first network model, and predicting a voltage phasor trajectory, a power angle trajectory and a frequency trajectory in a future preset time window; respectively carrying out corresponding evaluation and analysis, and outputting corresponding evaluation and analysis results; and respectively calculating an active power injection instruction in a power angle support mode, a reactive power injection instruction in a voltage phase correction mode or an anti-phase power injection instruction in an oscillation suppression mode according to an evaluation analysis result, and controlling the flywheel body to output power. According to the method, the power grid phasor trajectory is predicted, risks can be pre-judged and control measures can be actively taken before instability occurs, the defect of traditional afterward response is changed, passive is changed into active, and the timeliness and effectiveness of control are improved.

Description

Technical Field

[0001] This invention relates to the field of power system control technology, and in particular to a flywheel energy storage active phase modulation control method, system, equipment and medium. Background Technology

[0002] With the large-scale grid connection of new energy sources and the widespread application of power electronic equipment, the dynamic stability challenges faced by the power system are becoming increasingly severe. Problems such as power angle instability, voltage fluctuations, and subsynchronous oscillations seriously threaten the safe operation of the power grid and have become key bottlenecks restricting the high-quality development of the power system.

[0003] Existing power grid stability control methods mainly employ passive response strategies, meaning that corresponding control measures are initiated only after a system anomaly is detected. These methods suffer from drawbacks such as response lag and limited control effectiveness, making it difficult to meet the demands of modern power systems for rapid response and proactive defense.

[0004] Flywheel energy storage systems have become an important means of frequency and voltage regulation and stability control in power grids due to their advantages such as fast response speed, high power density, and long cycle life. However, existing flywheel energy storage control strategies are mostly reactive, failing to fully utilize their rapid response characteristics and still inadequate for proactive prevention of dynamic instability in the power grid.

[0005] Synchronous phasor measurement units (STMs) can provide high-precision, time-synchronized power grid phasor data, laying the technological foundation for dynamic power grid monitoring. Combined with advanced machine learning methods, especially physical information neural networks (PINNs), accurate prediction of power grid phasor evolution trajectories can be achieved. However, there is currently no mature technical solution for organically integrating phasor trajectory prediction with flywheel energy storage control to achieve proactive sensing and modulation of power grid instability risks. Summary of the Invention

[0006] In view of the aforementioned existing problems, the present invention is proposed.

[0007] Therefore, this invention provides a flywheel energy storage active phase modulation control method, system, device and medium to solve the problems of existing power grid stability control methods having lag in response, mostly being passive and reactive, and having difficulty in predicting and actively suppressing dynamic instability problems such as power angle instability, voltage fluctuation and subsynchronous oscillation.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a flywheel energy storage active phase modulation control method, comprising the following steps: collecting multi-source power data from the power grid; preprocessing the collected multi-source power data to obtain preprocessed phasor data; inputting the preprocessed phasor data into a first network model to predict the voltage phasor trajectory, power angle trajectory, and frequency trajectory within a future preset time window; performing corresponding evaluation and analysis on the predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory, and outputting corresponding evaluation and analysis results; calculating, based on the evaluation and analysis results, active power injection commands under power angle support mode, reactive power injection commands under voltage phase correction mode, or anti-phase power injection commands under oscillation suppression mode, and controlling the flywheel body to output power; monitoring the rotational speed, temperature, and actual output power of the flywheel body; comparing the actual output power with the predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory, calculating the prediction error, and feeding the prediction error back to the first network model for parameter updates.

[0009] As a preferred embodiment of the flywheel energy storage active phase modulation control method of the present invention, the multi-source power data includes grid voltage phasor, current phasor, and power phasor data; the preprocessing includes filtering and denoising, missing data interpolation, and outlier removal; the filtering and denoising uses the Kalman filtering algorithm to eliminate measurement noise; the missing data interpolation uses the cubic spline interpolation method to fill in missing data caused by communication delay or data packet loss; the outlier removal uses the 3σ criterion of sliding window statistics to identify and remove outlier data points.

[0010] As a preferred embodiment of the flywheel energy storage active phase modulation control method of the present invention, the evaluation and analysis includes power angle stability evaluation, voltage phase analysis and subsynchronous oscillation detection; The method for assessing the power angle stability is as follows: calculate the power angle difference and the rate of change of the power angle difference between adjacent generator sets based on the predicted power angle trajectory; when the absolute value of the power angle difference is greater than the power angle difference threshold or the absolute value of the rate of change of the power angle difference is greater than the rate of change of the power angle difference threshold, output a power angle instability risk signal. The voltage phase analysis method is as follows: monitor the voltage phase offset and rate of change of each node based on the predicted voltage phasor trajectory; when the voltage phase offset exceeds the voltage phase offset threshold or the voltage phase change rate exceeds the voltage phase change rate threshold, output a voltage fluctuation risk signal. The method for detecting subsynchronous oscillations is as follows: perform short-time Fourier transform on the predicted power angle trajectory and frequency trajectory to identify oscillation components; when the amplitude of the oscillation component exceeds the oscillation amplitude threshold and the duration exceeds the oscillation duration threshold, output a subsynchronous oscillation risk signal.

[0011] The beneficial effects of this preferred technical solution are as follows: by constructing a multi-dimensional risk assessment system through three means—power angle difference and its rate of change, voltage phase offset and its rate of change, and short-time Fourier transform—it can accurately identify and quantify the three types of risks: power angle instability, voltage fluctuation, and subsynchronous oscillation, enabling the system to adopt differentiated control strategies for different types of instability.

[0012] As a preferred embodiment of the flywheel energy storage active phase modulation control method of the present invention, the following steps are taken: when the evaluation and analysis result is a power angle instability risk signal, the power angle support mode is activated and the active power injection command is calculated; when the evaluation and analysis result is a voltage fluctuation risk signal, the voltage phase correction mode is activated and the reactive power injection command is calculated; when the evaluation and analysis result is a subsynchronous oscillation risk signal, the oscillation suppression mode is activated, the dominant oscillation frequency and oscillation phase are extracted, and the anti-phase power injection command is calculated.

[0013] The beneficial effects of this preferred technical solution are: it automatically switches between three control modes—power angle support, voltage phase correction, and oscillation suppression—based on the evaluation results, thereby realizing a unified processing framework for multiple types of power grid instability problems and avoiding the limitations of a single control mode under complex operating conditions.

[0014] As a preferred embodiment of the flywheel energy storage active phase modulation control method described in this invention, wherein: the loss function of the first network model Represented as: ; in, For data fitting loss, For the physical equation residuals, The equilibrium coefficient is used; the residuals of the physical equations include the generator swing equation residuals and the nodal power balance equation residuals; the generator swing equation is expressed as: ; ; in, For the angle of attack, Angular velocity, For synchronous angular velocity, The inertial time constant, For mechanical power, Electromagnetic power, The damping coefficient is; the nodal power balance equation is: and .

[0015] The beneficial effects of this preferred technical solution are as follows: by embedding the generator swing equation and the nodal power balance equation as soft constraints into the neural network loss function, the prediction model has both the generalization ability of data-driven models and the constraint of physical laws, effectively avoiding the problem that the prediction results of pure data-driven models violate the physical characteristics of the power system.

[0016] As a preferred embodiment of the flywheel energy storage active phase modulation control method described in this invention, wherein: the prediction error The calculation method is expressed as follows: ; in, The actual output power corresponds to the grid phasor measurement value. These are the corresponding values ​​of the voltage phasor trajectory, power angle trajectory, and frequency trajectory predicted by the first network model.

[0017] As a preferred embodiment of the flywheel energy storage active phase modulation control method of the present invention, the parameter update adopts the stochastic gradient descent algorithm, and the update formula is expressed as: ; in, These are the parameters of the first network model before the update. Here are the updated parameters of the first network model, and η is the learning rate. This is the gradient of the loss function with respect to the prediction error.

[0018] Secondly, the present invention provides a flywheel energy storage active phase modulation control system, comprising: a synchronous phasor measurement unit for acquiring multi-source power data, preprocessing it, and outputting preprocessed phasor data; The phasor trajectory prediction unit is used to predict the voltage phasor trajectory, power angle trajectory, and frequency trajectory based on the preprocessed phasor data. The instability risk assessment unit is used to perform assessment and analysis based on voltage phasor trajectory, power angle trajectory and frequency trajectory, and output the assessment and analysis results. The phase modulation control law calculation unit is used to calculate the active power injection command, reactive power injection command, or anti-phase power injection command based on the evaluation and analysis results, and control the flywheel body to output power. The state monitoring and feedback unit is used to calculate the prediction error and feed it back to the phasor trajectory prediction unit for parameter updates of the first network model.

[0019] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the flywheel energy storage active phase modulation control method.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the flywheel energy storage active phase modulation control method.

[0021] Compared with existing technologies, the beneficial effects of this invention are as follows: By predicting the phasor trajectory of the power grid, this invention can anticipate risks and proactively take control measures before instability occurs, overcoming the shortcomings of traditional reactive responses and transforming passivity into proactivity, thus improving the timeliness and effectiveness of control. It integrates three control modes: phasor support, voltage phase correction, and oscillation suppression, thereby addressing different types of instability risks with appropriate strategies and adapting to changing circumstances. This constructs a unified processing framework for various power grid stability problems, avoiding the limitations of a single control mode under complex operating conditions. Through a state monitoring and feedback module, the actual execution effect is compared with the predicted value, continuously optimizing the prediction model parameters, enabling the system to have adaptive learning capabilities. It can improve daily as the power grid operating conditions change, maintaining prediction accuracy. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the overall process of the flywheel energy storage active phase modulation control method according to an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 As an embodiment of the present invention, a flywheel energy storage active phase modulation control method is provided, comprising the following steps: S100. Collect multi-source power data from the power grid, preprocess the collected multi-source power data, and obtain preprocessed phasor data.

[0026] S200. Input the preprocessed phasor data into the first network model to predict the voltage phasor trajectory, power angle trajectory and frequency trajectory within the preset time window.

[0027] S300 performs corresponding evaluation and analysis on the predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory, and outputs the corresponding evaluation and analysis results.

[0028] S400: Based on the evaluation and analysis results, calculate the active power injection command in the power angle support mode, the reactive power injection command in the voltage phase correction mode, or the anti-phase power injection command in the oscillation suppression mode, and control the flywheel body to output power.

[0029] S500 monitors the rotational speed, temperature, and actual output power of the flywheel body, compares the actual output power with the predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory, calculates the prediction error, and feeds the prediction error back to the first network model for parameter updates.

[0030] It should be noted that with the large-scale grid connection of new energy sources and the widespread application of power electronic equipment, the dynamic stability challenges faced by power systems are becoming increasingly severe. Problems such as power angle instability, voltage fluctuations, and subsynchronous oscillations seriously threaten the safe operation of the power grid. These problems occur rapidly and evolve in complex ways, placing extremely high demands on the response speed and predictive capabilities of control systems. Traditional power grid stability control methods mainly adopt passive response strategies, that is, corresponding control measures are only initiated after system anomalies are detected. This approach suffers from problems such as response lag and limited control effectiveness. At the same time, due to the complex and variable operating conditions of the power grid, a single control mode is insufficient to cope with multiple types of instability risks, and purely data-driven prediction models are prone to producing prediction results that violate the physical laws of the power system, affecting the reliability of control. Therefore, proactive prevention and control of dynamic instability in the power grid are of paramount importance.

[0031] Therefore, to address the aforementioned issues of response lag and prediction reliability, steps S100-S500 are used to collect and preprocess multi-source power data from the power grid. The first network model is then used to predict the phasor evolution trajectory of the power grid, enabling proactive prediction of risks such as power angle instability, voltage fluctuations, and subsynchronous oscillations. Based on different types of instability risks, the system automatically switches between power angle support, voltage phase correction, or oscillation suppression control modes, calculates corresponding power injection commands to control the flywheel body for power output, and simultaneously optimizes the prediction model parameters online based on a state monitoring and error feedback mechanism, enabling the system to adapt to dynamic changes in the power grid's operating conditions.

[0032] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a flywheel energy storage active phase modulation control method is provided.

[0033] In this embodiment, step S100 collects multi-source power data from the power grid and preprocesses the collected multi-source power data to obtain preprocessed phasor data. Specifically, the multi-source power data includes power grid voltage phasor, current phasor, and power phasor data. Data is collected using synchronous phasor measurement units (PMUs) deployed at key nodes of the power grid. The PMUs are deployed at key nodes such as generator terminals, important substation busbars, and both ends of tie lines, with a sampling frequency set to 100-200Hz. All PMUs achieve microsecond-level time synchronization via GPS or BeiDou satellites, ensuring spatial comparability and temporal consistency of phasor data at each node.

[0034] When preprocessing multi-source power data, the preprocessing process includes three steps: filtering and noise reduction, missing data interpolation, and outlier removal.

[0035] The filtering and denoising process employs an adaptive Kalman filter algorithm to eliminate measurement noise. Kalman filtering establishes a state-space model and utilizes the system's dynamic and observation equations to perform optimal estimation of noisy measurement data, effectively eliminating measurement noise and high-frequency interference introduced during PMU acquisition and improving the signal-to-noise ratio of phasor data.

[0036] Missing data interpolation employs cubic spline interpolation to fill in missing data caused by communication delays or packet loss. Cubic spline interpolation constructs a piecewise cubic polynomial function to smoothly interpolate missing data points while ensuring the continuity of the first and second derivatives of the interpolation curve. This maintains good continuity and smoothness in the interpolated data sequence, preventing incomplete input to subsequent prediction models due to missing data.

[0037] Outlier removal employs the 3σ criterion of sliding window statistics to identify and remove outlier data points. Specifically, a sliding window of a certain length is set, and the mean of the data within the window is calculated. and standard deviation When the deviation of a data point from the mean exceeds At that time, that is If an outlier is detected, the data point is identified as an outlier and removed. This method can effectively identify abnormal data caused by equipment failure, electromagnetic interference, etc., ensuring the quality of phasor data input to the prediction model.

[0038] In this embodiment of the application, S200 inputs the preprocessed phasor data into the first network model to predict the voltage phasor trajectory, power angle trajectory and frequency trajectory within a future preset time window.

[0039] Specifically, the first network model employs a Physical Information Neural Network (PINN), whose network structure includes an input layer, multiple hidden layers, and an output layer. The input layer receives preprocessed historical phasor sequence data, the hidden layers use a fully connected network with residual connections, the activation function is the tanh function, and the output layer provides predicted values ​​for voltage amplitude, voltage phase angle, power angle, and frequency within a preset time window. The preset time window can be set according to actual control requirements, typically ranging from 100ms to 1s.

[0040] Loss function of the first network model Represented as: ; in, For data fitting loss, For the physical equation residuals, This is the balance coefficient; The physical equation residuals include the generator swing equation residuals and the nodal power balance equation residuals; and the generator swing equations are expressed as: ; ; in, For the angle of attack, Angular velocity, For synchronous angular velocity, The inertial time constant, For mechanical power, Electromagnetic power, The damping coefficient; The node power balance equation is and This means that the active and reactive power of each node satisfies the balance condition.

[0041] By embedding the aforementioned physical equations into the loss function in the form of residuals, the neural network not only learns the statistical regularities of the data during training, but also needs to satisfy the physical constraints of the power system, thereby improving the physical consistency and reliability of the prediction results.

[0042] As a concrete example, consider a regional power grid comprising three thermal power units and two wind farms, interconnected with the main grid via four 500kV lines. Six power unit monitoring (PMU) devices are deployed at key grid nodes, with a sampling frequency of 100Hz. The first network model has four hidden layers, each with 128 neurons, a balance coefficient λ of 0.1, and a preset time window of 500ms. When a fault occurs on a tie line, the first network model can predict the abnormal evolution trend of the power angle trajectory 50ms before the fault is cleared, providing a basis for the early initiation of subsequent control measures.

[0043] In an alternative implementation, the first network model in S200 can also employ a hybrid architecture combining a Long Short-Term Memory (LSTM) network with physical constraints. LSTM networks possess excellent temporal feature extraction capabilities, effectively capturing the long-term dependencies and dynamic evolution patterns of power grid phasor data. This hybrid model also embeds generator swing equations and node power balance equations as soft constraints into the LSTM loss function, enabling the network to adhere to the physical laws of the power system while learning the temporal features of historical data.

[0044] In another optional implementation, phasor trajectory prediction in S200 can also employ a technique that fuses graph neural networks (GNNs) with physical information. A power system is essentially a complex topological network structure, with nodes interconnected via transmission lines and engaging in power exchange. By modeling the power grid topology as a graph and using a graph neural network to aggregate and learn node and edge features, the influence of the power grid's spatial topology information on phasor evolution can be fully explored. Simultaneously, physical equation constraints are embedded into the training process of the graph neural network, constructing a prediction model that combines topology awareness and physical consistency. This approach is particularly suitable for phasor trajectory prediction in large-scale interconnected power grids, more accurately reflecting the coupling relationships and disturbance propagation characteristics between nodes in the power grid.

[0045] In this embodiment of the application, S300 performs corresponding evaluation and analysis on the predicted voltage phasor trajectory, power angle trajectory and frequency trajectory, and outputs the corresponding evaluation and analysis results.

[0046] The evaluation and analysis include power angle stability assessment, voltage phase analysis, and subsynchronous oscillation detection. A1. The method for evaluating power angle stability is as follows: calculate the power angle difference between adjacent generator sets based on the predicted power angle trajectory. and the rate of change of work angle difference The power angle difference reflects the relative angular offset between adjacent generator units, while the rate of change of the power angle difference reflects its changing trend and speed. When the absolute value of the power angle difference exceeds the power angle difference threshold or the absolute value of the rate of change of the power angle difference exceeds the rate of change of the power angle difference threshold, the system is deemed to have a risk of power angle instability, and a power angle instability risk signal is output. As a specific example, the power angle difference threshold can be set to 90°, and the rate of change of the power angle difference threshold can be set to 60°, that is, when... or At that time, a risk signal of power angle instability is triggered.

[0047] When the assessment analysis indicates a risk signal of power angle instability, the power angle support mode is activated, and the active power injection command is calculated. In power angle support mode, the active power injection command... The calculation formula is: ; in, For proportional gain, The differential gain is used, and both are tuned according to the system damping requirements and the flywheel energy storage capacity. The power injection phase is opposite to the direction of the power angle difference. By injecting active power, the system damping is enhanced, the power angle oscillation is suppressed, and the power angle instability is prevented.

[0048] A2. The method for voltage phase analysis is as follows: monitor the voltage phase offset at each node based on the predicted voltage phasor trajectory. and rate of change Voltage phase offset reflects the degree of deviation of the node voltage phase angle from the reference value, while voltage phase change rate reflects the speed of phase change. When the voltage phase offset exceeds the voltage phase offset threshold or the voltage phase change rate exceeds the voltage phase change rate threshold, the system is determined to have voltage fluctuation risk, and a voltage fluctuation risk signal is output.

[0049] When the assessment and analysis result indicates a voltage fluctuation risk signal, the voltage phase correction mode is activated, and the reactive power injection command is calculated. Under voltage phase correction mode, the reactive power injection command... The calculation formula is: ; in, For voltage deviation, For voltage deviation gain, This is the voltage change rate gain. By injecting reactive power to adjust the amplitude and phase of node voltages, voltage fluctuations are stabilized, and grid voltage quality is maintained.

[0050] A3. The method for detecting subsynchronous oscillations is as follows: A short-time Fourier transform (STFT) is performed on the predicted power angle and frequency trajectories to identify oscillation components within the 0.1-50Hz frequency band. The STFT divides the signal into short-time segments through windowing, and performs a Fourier transform on each segment to obtain the time-frequency characteristics of the signal. This effectively identifies the frequency, amplitude, and time-varying patterns of the oscillation components. When the amplitude of the oscillation component exceeds the oscillation amplitude threshold and the duration exceeds the oscillation duration threshold, the system is deemed to have a risk of subsynchronous oscillation, and a subsynchronous oscillation risk signal is output.

[0051] When the assessment and analysis results indicate a risk signal of subsynchronous oscillation, the oscillation suppression mode is activated, and the dominant oscillation frequency is extracted. and oscillation phase Calculate the anti-phase power injection command. In oscillation suppression mode, the anti-phase power injection command... The calculation formula is: ; Where A is the power injection amplitude, determined based on the amplitude of the oscillation component and the system damping requirements. By actively injecting power that is out of phase with the oscillation, the oscillation energy is rapidly dissipated, suppressing the development and spread of subsynchronous oscillations.

[0052] As a specific example, when a wind farm experiences subsynchronous oscillations due to power transmission via a series compensation line, the short-time Fourier transform detects an oscillation component with a frequency of 15Hz. If the amplitude exceeds a set threshold and the duration exceeds 100ms, the system immediately outputs a subsynchronous oscillation risk signal and activates the oscillation suppression mode. The flywheel energy storage system extracts the dominant oscillation frequency of 15Hz and its corresponding phase, calculates an anti-phase power injection command, and effectively prevents oscillation divergence by injecting anti-phase power.

[0053] In an optional implementation, the subsynchronous oscillation detection in S300 can also employ wavelet transform instead of short-time Fourier transform for time-frequency analysis. Wavelet transform offers multi-resolution analysis capabilities, providing high frequency resolution in the low-frequency band and high time resolution in the high-frequency band, overcoming the time-frequency resolution contradiction caused by the fixed window function in short-time Fourier transform. By selecting a suitable mother wavelet function (such as the Morlet wavelet or Daubechies wavelet) and performing continuous wavelet transforms on the power angle trajectory and frequency trajectory, the start time and frequency characteristics of the oscillation components can be located more accurately. This approach is particularly suitable for complex operating conditions where the oscillation frequency varies over time or where multiple oscillation modes are superimposed, providing more detailed time-frequency information to support the formulation of oscillation suppression control strategies.

[0054] In this embodiment, S400 calculates the active power injection command in the power angle support mode, the reactive power injection command in the voltage phase correction mode, or the anti-phase power injection command in the oscillation suppression mode based on the evaluation and analysis results, and controls the flywheel body to output power.

[0055] As a concrete example, consider a scenario where a wind farm with a rated capacity of 200MW is connected to the grid via a 200km 500kV transmission line. A 10MW / 10MWh flywheel energy storage system is configured at the wind farm's output. When a sudden increase in wind power output causes the generator at the sending end to lead in power angle, the S300 outputs a power angle instability risk signal, with the detected power angle difference Δδ value being... Rate of change of work angle difference The system immediately switches to power angle support mode and sets... , The active power injection command is calculated. The flywheel energy storage system responds to power commands within 50ms, absorbing 5.1MW of active power. The flywheel speed increases by about 2% from the rated speed, and the system power angle difference recovers to the safe range within 500ms.

[0056] In an optional implementation, the control mode switching in S400 can also employ a fuzzy logic-based intelligent switching strategy. When multiple types of instability risks exist simultaneously in the power grid, traditional single-mode switching may lead to conflicting control objectives or poor control performance. By establishing a fuzzy rule base, risk indicators such as power angle difference, voltage deviation, and oscillation amplitude are used as fuzzy inputs. After fuzzification, fuzzy inference, and defuzzification processes, the weight coefficients and power allocation ratios of each control mode are output. The fuzzy rules are designed based on power system operating experience and simulation verification results; for example, "when the power angle difference is large and the oscillation amplitude is small, the power angle support mode has a high weight." This approach enables multi-mode coordinated control, rationally allocating the active and reactive power capacity of the flywheel energy storage system when multiple risks occur simultaneously, and improving control performance under complex operating conditions.

[0057] In this embodiment, S500 monitors the rotational speed, temperature and actual output power of the flywheel body, compares the actual output power with the predicted voltage phasor trajectory, power angle trajectory and frequency trajectory, calculates the prediction error, and feeds the prediction error back to the first network model for parameter update.

[0058] Specifically, the condition monitoring of the flywheel body is achieved through multiple types of sensors. Speed ​​monitoring uses a high-precision rotary encoder or Hall sensor to acquire the speed signal of the flywheel rotor in real time, with a sampling frequency of not less than 1kHz and a speed measurement accuracy better than 0.1%. Temperature monitoring uses thermocouples or infrared temperature sensors to acquire the temperature of key components such as flywheel bearings, stator windings, and converter power modules to determine the thermal state and operational health of the flywheel energy storage system. The actual output power is calculated from the DC bus voltage and current measurements of the bidirectional converter, or directly acquired through the power measurement device on the AC side.

[0059] Prediction error The calculation method is expressed as follows: ; in, The actual output power corresponds to the grid phasor measurement value. These are the corresponding values ​​of the voltage phasor trajectory, power angle trajectory, and frequency trajectory predicted by the first network model. The prediction error reflects the deviation between the prediction results of the first network model and the actual power grid response, and is a core indicator for evaluating the accuracy of the prediction model and guiding the optimization of model parameters.

[0060] The parameter update uses the stochastic gradient descent algorithm, and the update formula is expressed as: ; in, These are the parameters of the first network model before the update, including the weight matrices and bias vectors of each layer of the neural network; These are the updated parameters of the first network model; The learning rate controls the step size of parameter updates. Its value is generally between 0.001 and 0.01. If the learning rate is too large, it may cause parameter oscillations, while if the learning rate is too small, the convergence speed will be slow. The gradient of the loss function with respect to the prediction error is calculated using the backpropagation algorithm.

[0061] The loss function L is designed to take into account both prediction accuracy and physical constraints, and is expressed as: ; in, The data fitting loss is expressed in the form of mean squared error, i.e. , The number of samples; To account for the residual loss of the physical equations, we ensure that the updated model still satisfies physical constraints such as the generator swing equation and the nodal power balance equation. This is a balancing coefficient used to adjust the weight ratio between data fitting and physical constraints.

[0062] Parameter updates are implemented using online learning. At the end of each control cycle, the system collects actual power grid response data, calculates the prediction error, and uses this error sample to incrementally update the parameters of the first network model. To avoid introducing excessive disturbances in a single update, a mini-batch gradient descent strategy is adopted, using error samples from the most recent control cycles to form a mini-batch dataset for parameter updates. Simultaneously, a trigger condition for parameter updates is set; parameter updates are only initiated when the moving average of the prediction error exceeds a set threshold, preventing frequent updates that could lead to model instability.

[0063] In an alternative implementation, parameter updates in S500 can also employ an adaptive learning rate optimization algorithm instead of the fixed learning rate stochastic gradient descent algorithm. Specifically, the Adam (Adaptive Moment Estimation) optimizer can be used. This optimizer combines the advantages of momentum and RMSprop, and can adaptively adjust the learning rate of each parameter based on the first and second moment estimates of the parameter gradients. The parameter update formula for the Adam optimizer is: ; ; ; in, and These are the first and second moments of the gradient, respectively. and The attenuation coefficient is generally taken as... , , This is the numerical stability constant. This approach allows for faster convergence by using a larger effective learning rate in the early stages of parameter updates, and then reducing the learning rate to improve convergence accuracy as the system approaches its optimum.

[0064] In summary, this invention, by predicting the phasor trajectory of the power grid, can anticipate risks and proactively implement control measures before instability occurs, overcoming the shortcomings of traditional reactive responses and transforming passivity into proactivity, thus improving the timeliness and effectiveness of control. It integrates three control modes—phase support, voltage phase correction, and oscillation suppression—to address different types of instability risks with tailored strategies, constructing a unified framework for handling various power grid stability problems and avoiding the limitations of single control modes under complex operating conditions. Through a state monitoring and feedback module, the actual execution results are compared with the predicted values, continuously optimizing the prediction model parameters, enabling the system to have adaptive learning capabilities, constantly improving as the power grid operating conditions change, and maintaining prediction accuracy.

[0065] Example 3 illustrates a schematic scheme for an active phase modulation control method for flywheel energy storage. It should be noted that the technical solution of this flywheel energy storage active phase modulation control system belongs to the same concept as the technical solution of the aforementioned active phase modulation control method for flywheel energy storage. Details not described in detail in this embodiment can be found in the description of the technical solution of the aforementioned active phase modulation control method for flywheel energy storage.

[0066] This embodiment also provides a flywheel energy storage active phase modulation control system, including: The synchronous phasor measurement unit is used to acquire multi-source power data, preprocess it, and output the preprocessed phasor data. The phasor trajectory prediction unit is used to predict the voltage phasor trajectory, power angle trajectory, and frequency trajectory based on the preprocessed phasor data. The instability risk assessment unit is used to perform assessment and analysis based on voltage phasor trajectory, power angle trajectory and frequency trajectory, and output the assessment and analysis results. The phase modulation control law calculation unit is used to calculate the active power injection command, reactive power injection command, or anti-phase power injection command based on the evaluation and analysis results, and control the flywheel body to output power. The state monitoring and feedback unit is used to calculate the prediction error and feed it back to the phasor trajectory prediction unit for parameter updates of the first network model.

[0067] This embodiment also provides an electronic device suitable for flywheel energy storage active phase modulation control, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the flywheel energy storage active phase modulation control method proposed in the above embodiment.

[0068] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the flywheel energy storage active phase modulation control method proposed in the above embodiments.

[0069] The storage medium proposed in this embodiment and the method for implementing active phase modulation control of flywheel energy storage proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0070] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A flywheel energy storage active phase modulation control method, characterized in that, Includes the following steps: Collect multi-source power data from the power grid, preprocess the collected multi-source power data, and obtain preprocessed phasor data; The preprocessed phasor data is input into the first network model to predict the voltage phasor trajectory, power angle trajectory and frequency trajectory within a preset time window in the future. The predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory are used to perform corresponding evaluation and analysis, and the corresponding evaluation and analysis results are output. Based on the evaluation and analysis results, the active power injection command in the power angle support mode, the reactive power injection command in the voltage phase correction mode, or the anti-phase power injection command in the oscillation suppression mode are calculated respectively, and the flywheel body is controlled to output power. The rotational speed, temperature, and actual output power of the flywheel are monitored. The actual output power is compared with the predicted voltage phasor trajectory, power angle trajectory, and frequency trajectory. The prediction error is calculated and fed back to the first network model for parameter updates.

2. The flywheel energy storage active phase modulation control method as described in claim 1, characterized in that, The multi-source power data includes grid voltage phasor, current phasor, and power phasor data; The preprocessing includes filtering and noise reduction, missing data interpolation, and outlier removal; The filtering and denoising process employs a Kalman filtering algorithm to eliminate measurement noise. The missing data interpolation uses cubic spline interpolation to fill in missing data caused by communication delays or data packet loss. The outlier removal is performed using the 3σ criterion of sliding window statistics to identify and remove outlier data points.

3. The flywheel energy storage active phase modulation control method as described in claim 2, characterized in that, The evaluation and analysis include power angle stability assessment, voltage phase analysis, and subsynchronous oscillation detection. The method for evaluating the power angle stability is as follows: Calculate the power angle difference and the rate of change of power angle difference between adjacent generator sets based on the predicted power angle trajectory; When the absolute value of the power angle difference is greater than the power angle difference threshold or the absolute value of the power angle difference change rate is greater than the power angle difference change rate threshold, a power angle instability risk signal is output. The method for voltage phase analysis is as follows: Monitor the offset and rate of change of the voltage phase at each node based on the predicted voltage phasor trajectory; When the voltage phase offset exceeds the voltage phase offset threshold or the voltage phase change rate exceeds the voltage phase change rate threshold, a voltage fluctuation risk signal is output. The method for detecting the subsynchronous oscillation is as follows: Short-time Fourier transform is performed on the predicted power angle trajectory and frequency trajectory to identify oscillation components; When the amplitude of the oscillation component exceeds the oscillation amplitude threshold and the duration exceeds the oscillation duration threshold, a subsynchronous oscillation risk signal is output.

4. The flywheel energy storage active phase modulation control method as described in claim 3, characterized in that, When the assessment and analysis results indicate a risk signal of power angle instability, the power angle support mode is activated, and the active power injection command is calculated. When the assessment and analysis result is a voltage fluctuation risk signal, the voltage phase correction mode is activated, and the reactive power injection command is calculated. When the assessment and analysis result indicates a risk signal of subsynchronous oscillation, the oscillation suppression mode is activated, the dominant oscillation frequency and oscillation phase are extracted, and the anti-phase power injection command is calculated.

5. The flywheel energy storage active phase modulation control method as described in claim 4, characterized in that, The loss function of the first network model Represented as: ； in, For data fitting loss, For the physical equation residuals, This is the balance coefficient; The physical equation residuals include the generator swing equation residuals and the nodal power balance equation residuals; The generator swing equation is expressed as: ； ； in, For the angle of attack, Angular velocity, For synchronous angular velocity, The inertial time constant, For mechanical power, Electromagnetic power, The damping coefficient; The node power balance equation is: and .

6. The flywheel energy storage active phase modulation control method as described in claim 5, characterized in that, The prediction error The calculation method is expressed as follows: ； in, The actual output power corresponds to the grid phasor measurement value. These are the corresponding values ​​of the voltage phasor trajectory, power angle trajectory, and frequency trajectory predicted by the first network model.

7. The flywheel energy storage active phase modulation control method as described in claim 6, characterized in that, The parameter update employs a stochastic gradient descent algorithm, and the update formula is expressed as follows: ； in, These are the parameters of the first network model before the update. Here are the updated parameters of the first network model, and η is the learning rate. This is the gradient of the loss function with respect to the prediction error.

8. A flywheel energy storage active phase modulation control system, using the method described in any one of claims 1-7, characterized in that, include: The synchronous phasor measurement unit is used to acquire multi-source power data, preprocess it, and output the preprocessed phasor data. The phasor trajectory prediction unit is used to predict the voltage phasor trajectory, power angle trajectory, and frequency trajectory based on the preprocessed phasor data. The instability risk assessment unit is used to perform assessment and analysis based on voltage phasor trajectory, power angle trajectory and frequency trajectory, and output the assessment and analysis results. The phase modulation control law calculation unit is used to calculate the active power injection command, reactive power injection command, or anti-phase power injection command based on the evaluation and analysis results, and control the flywheel body to output power. The state monitoring and feedback unit is used to calculate the prediction error and feed it back to the phasor trajectory prediction unit for parameter updates of the first network model.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the flywheel energy storage active phase modulation control method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the flywheel energy storage active phase modulation control method according to any one of claims 1 to 7.

Images