Adaptive inertia damping cooperative control method and system of energy storage VSG

Abstract

The application discloses a self-adaptive inertia damping cooperative control method and system of energy storage VSG, comprising: collecting the electromagnetic power and angular frequency of a virtual synchronous generator and calculating the deviation, and differentiating after extracting the transient component through high-pass filtering; calculating the comprehensive state index according to the grid-connected state coefficient and the load level, and calculating the deviation weight coefficient and the change rate weight coefficient according to the comprehensive state index and then performing weighted fusion; inputting the historical deviation sequence into a prediction model to output a prediction value and a prediction confidence; adaptively adjusting the learning rate according to the prediction confidence, and iteratively updating the basic parameters of the weight coefficient by an online optimizer; according to the state of charge of the energy storage, remodeling the curve form of the variable gain function in real time, calculating the virtual moment of inertia and the virtual damping coefficient, and substituting into the rotor motion equation to complete the closed-loop control. The application realizes adaptive multi-signal fusion control in grid-connected and island working conditions, has the characteristics of advanced prediction and energy storage friendliness, and effectively suppresses power oscillation and frequency fluctuation.

Description

Technical Field

[0001] This invention relates to the field of smart grid and energy storage grid-connected control technology, specifically to an adaptive inertia damping coordinated control method and system for energy storage VSG. Background Technology

[0002] Virtual synchronous generator (VSG) technology, by simulating the electromechanical transient characteristics of traditional synchronous generators, can provide virtual inertia and damping support for grid-connected inverters, and is one of the core technologies in the current field of energy storage grid-connected control. Existing VSG control strategies are mainly divided into two categories: fixed parameter control and parameter adaptive control.

[0003] Fixed-parameter VSG control employs preset virtual moment of inertia and virtual damping coefficients, and its parameter tuning is typically determined based on the system's steady-state operating conditions or offline optimization methods. This method is simple in structure, easy to implement in engineering, and has been widely used in traditional power systems. Building upon this, researchers have further proposed several parameter adaptive control schemes to improve the dynamic response characteristics of the system during transient processes. One type of scheme designs adaptive adjustment rules for inertia and damping coefficients based on single signals such as frequency deviation and its rate of change; another type introduces a multi-parameter collaborative adjustment mechanism, jointly adjusting the moment of inertia, damping coefficient, and active power droop coefficient.

[0004] However, the aforementioned existing technologies still have the following shortcomings:

[0005] 1. Fixed-parameter VSG control has inherent performance conflicts. The fixed virtual moment of inertia and damping coefficient have strong coupling characteristics on the system performance, making it impossible to take into account both the dynamic response performance and steady-state operation performance of the system. Under disturbances such as sudden load changes and operating condition switching, it is very easy to cause severe frequency fluctuations and continuous oscillation of active power.

[0006] 2. Existing adaptive schemes are mostly based on single-signal frequency driving, resulting in insufficient signal fusion, inadequate control precision, failure to fully utilize key state information such as power, difficulty in accurately identifying disturbance types and intensities, and inability to achieve coordinated optimization control of frequency and power. On the other hand, optimization schemes based on intelligent algorithms or segmented thresholds suffer from poor real-time performance and secondary disturbances caused by parameter jumps.

[0007] 3. Existing adaptive schemes all perform reactive adjustments based on the measured signals at the current moment, lacking the ability to predict the development trend of disturbances in advance. They are slow to respond in the initial stage of disturbances, which restricts the further improvement of control performance.

[0008] 4. Existing adaptive solutions do not fully consider the state of charge (SOC) and state of health (SOH) constraints of energy storage batteries. Excessive or large-scale inertia damping adjustments will accelerate battery aging, shorten the service life of energy storage equipment, and reduce the engineering practicality and economy of the system. Summary of the Invention

[0009] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an adaptive inertia damping cooperative control method and system for energy storage VSG, solving the problems mentioned in the background section.

[0010] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: an adaptive inertia damping cooperative control method and system for energy storage VSG, comprising the following steps: S1: Collect the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate them to obtain the power change rate and frequency change rate. S2, calculate the comprehensive status index based on the grid connection status coefficient and load level. The grid connection status coefficient is 1 for grid connection and 0 for islanding. The load level is the ratio of electromagnetic power to given active power. S3. Calculate the deviation weighting coefficient and the rate of change weighting coefficient based on the comprehensive state index. Then, weight and fuse the electromagnetic power deviation, angular frequency deviation, power change rate and frequency change rate based on the deviation weighting coefficient and the rate of change weighting coefficient to obtain the comprehensive deviation and comprehensive rate of change. S4, input the electromagnetic power deviation sequence and angular frequency deviation sequence in the historical period into the prediction model, and output the future electromagnetic power deviation prediction value, angular frequency deviation prediction value and prediction confidence level; S5, the learning rate of the online optimizer is adaptively adjusted according to the prediction confidence, and the basic parameters of the deviation weight coefficient and the rate of change weight coefficient are iteratively updated by the online optimizer with the system performance index as the objective function. S6 reshapes the curve of the variable gain function in real time according to the state of charge of the energy storage battery. The virtual moment of inertia and virtual damping coefficient are calculated based on the comprehensive deviation and comprehensive rate of change of the reshaped variable gain function. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to complete the closed-loop control.

[0011] In step S4, the prediction model is also used to estimate the expected change in state of charge in the future period based on the predicted value of electromagnetic power deviation; between step S5 and step S6, it also includes: predicting the predicted trajectory of state of charge in the future period based on the expected change in state of charge, and when the predicted trajectory exceeds the safety boundary, actively correcting the given active power based on the deviation between the predicted value of state of charge and the safety boundary.

[0012] In step S5, the learning rate is adaptively adjusted according to the following formula: ; in, For adaptive learning rate, As the baseline learning rate, To adjust the amplitude coefficient, The prediction confidence level is... The current control cycle number; the prediction confidence level The smaller the adaptive learning rate The larger the value, the more aggressively the online optimizer searches for the optimal basic parameters when the prediction reliability is low.

[0013] In step S5, the objective function is composed of a weighted sum of a frequency deviation integral term and a power deviation integral term. The frequency deviation integral term characterizes the overall deviation of the system frequency, and the power deviation integral term characterizes the overall deviation of the active power. The weight coefficient of the frequency deviation integral term increases as the maximum value of the predicted angular frequency deviation increases, and the weight coefficient of the power deviation integral term increases as the maximum value of the predicted electromagnetic power deviation increases, so that the greater the prediction disturbance, the higher the optimization priority of the corresponding channel in the objective function. The maximum value of the predicted angular frequency deviation and the maximum value of the predicted electromagnetic power deviation are determined by the predicted values ​​in the future time domain output by the prediction model.

[0014] In step S3, the comprehensive deviation is calculated as follows: the rate of change weighting coefficient multiplied by the absolute value of the power rate of change, plus a factor less than the rate of change weighting coefficient multiplied by the absolute value of the frequency rate of change; the comprehensive rate of change is calculated as follows: the deviation weighting coefficient multiplied by the absolute value of the transient component of the electromagnetic power deviation, plus a factor less than the deviation weighting coefficient multiplied by the absolute value of the angular frequency deviation; the comprehensive deviation represents the comprehensive rate at which the system state deviates from the steady state, and the comprehensive rate of change represents the comprehensive magnitude at which the system state deviates from the steady state, both of which are used for the subsequent adaptive adjustment of the virtual damping coefficient and the virtual moment of inertia.

[0015] In step S4, the prediction model is an autoregressive model, and its predicted value is composed of a linear weighted sum of historical state vectors. The model coefficients are identified online using the recursive least squares method. The prediction confidence is calculated based on the variance of the historical prediction residuals, and the prediction confidence range is from 0 to 1. The variance of the recent prediction residuals is obtained by performing variance statistics on the prediction residual sequences of the most recent control cycles, and the benchmark variance is the benchmark value of the prediction residual variance under the steady-state operation conditions of the system.

[0016] In step S6, the variable gain function is a variable gain function with shape parameters: ; in, The input variable for the variable gain function is... The shape parameters; The shape parameters change dynamically with the state of charge according to the following formula: ; in, This refers to the state of charge of the energy storage battery. Based on the reference shape parameters, To adjust the amplitude coefficient, It is a monotonically increasing and bounded function of the state of charge; the... Satisfy: When the state of charge is at the preset healthy midpoint Set to 0, when the state of charge reaches the preset safety limit. Set to 1, when the state of charge reaches the preset safety lower limit. Set to -1; the shape parameter It increases monotonically as the degree to which the state of charge deviates from the healthy midpoint increases.

[0017] In step S6, the sign of the product of the angular frequency deviation and the rate of change of frequency determines whether parameter adjustment is required: when the product of the angular frequency deviation and the rate of change of frequency is greater than zero, the virtual moment of inertia and the virtual damping coefficient are in the dynamic adjustment range, and the virtual moment of inertia and the virtual damping coefficient are updated according to the calculated values; when the product of the angular frequency deviation and the rate of change of frequency is less than zero, the virtual moment of inertia and the virtual damping coefficient are in the recovery and maintenance range, and the virtual moment of inertia and the virtual damping coefficient are kept at their respective reference values ​​to accelerate the convergence of the system to steady state and avoid over-adjustment of parameters; the sign of the product is used to characterize the consistency of the direction of the angular frequency deviation and the rate of change of frequency.

[0018] The deviation weighting coefficient and the rate of change weighting coefficient are calculated according to the following formulas: ; in, The deviation weighting coefficient is... The change rate weighting coefficient, The comprehensive status index, , The preset adjustment range, , These are the basic parameters of deviation and the basic parameters of rate of change, respectively. This is the current control cycle number; The deviation basic parameters and the basic parameters of the rate of change The online optimizer is iteratively updated periodically using the following gradient descent method: ; in, For adaptive learning rate, Let the objective function of the online optimizer be... This is the sequence number of the next control cycle; and These are the partial derivatives of the objective function with respect to the basic deviation parameter and the basic rate of change parameter, respectively, used to guide the basic deviation parameter and the basic rate of change parameter to be updated in the direction that reduces the objective function; the online optimizer performs micro-iterations on the basic deviation parameter and the basic rate of change parameter in each control cycle based on the real-time response performance of the system, thereby realizing online self-tuning of the basic weight parameters.

[0019] An adaptive inertia damping cooperative system for energy storage VSG, applied to the aforementioned control methods, includes: The signal acquisition and preprocessing module is used to acquire the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate them to obtain the power change rate and frequency change rate. The status assessment and weight calculation module is used to calculate a comprehensive status index based on the grid connection status coefficient and load level, calculate a deviation weight coefficient and a change rate weight coefficient based on the comprehensive status index, and perform weighted fusion of the electromagnetic power deviation, the angular frequency deviation, the power change rate and the frequency change rate based on the deviation weight coefficient and the change rate weight coefficient to obtain a comprehensive deviation and a comprehensive change rate. The prediction module is used to input the electromagnetic power deviation sequence and angular frequency deviation sequence from the historical period into the prediction model, and output the predicted values ​​of future electromagnetic power deviation, angular frequency deviation, and prediction confidence. The prediction optimization collaboration module is used to adaptively adjust the learning rate according to the prediction confidence, and iteratively update the basic parameters of the deviation weight coefficient and the rate of change weight coefficient with the system performance index as the objective function. The energy storage state management module is used to monitor the state of charge and health of the energy storage battery and actively correct the given active power based on the predicted trajectory of the state of charge. The variable gain calculation and execution module is used to reshape the curve shape of the variable gain function in real time according to the state of charge. The reshaped variable gain function is used to calculate the virtual moment of inertia and virtual damping coefficient based on the comprehensive deviation and the comprehensive rate of change. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to generate a pulse width modulation signal to drive the inverter.

[0020] (III) Beneficial Effects (1) This invention integrates grid-connected state coefficient and load level by combining comprehensive state index, and calculates deviation weight coefficient and rate of change weight coefficient in real time based on the index. It realizes adaptive weighted fusion of power channel and frequency channel under different operating conditions of grid connection and islanding. It overcomes the defects of traditional fixed parameter VSG control that cannot take into account dynamic response and steady-state performance, as well as insufficient signal fusion of single signal drive scheme. It effectively suppresses power oscillation and frequency fluctuation caused by load change and operating condition switching.

[0021] (2) This invention introduces a prediction model to predict the future electromagnetic power deviation and angular frequency deviation online and outputs the prediction confidence. The learning rate of the online optimizer is adaptively adjusted according to the prediction confidence, so that the basic parameters of the weight coefficient can be self-tuned in real time with the change of the dynamic characteristics of the system. This overcomes the defect of the existing adaptive scheme that only performs post-response adjustment based on the measured signal at the current moment. It realizes the advance prediction and active control of the disturbance development trend, and further improves the dynamic response quality of frequency and power.

[0022] (3) This invention predicts the expected change in state of charge in the future period through a predictive model. When the predicted trajectory exceeds the safety boundary, the given active power is actively corrected. At the same time, the curve shape of the variable gain function is reshaped in real time according to the current state of charge, so that the control law changes dynamically with the energy storage state. This overcomes the defect of existing adaptive schemes that do not consider the constraints of energy storage SOC and SOH, avoids over-adjustment that accelerates battery aging or triggers protective shutdown, extends the service life of energy storage equipment, and improves the engineering practicality and economy of the system.

[0023] (4) The present invention determines the adjustment range of virtual moment of inertia and virtual damping coefficient by the sign of the product of angular frequency deviation and frequency change rate. Combined with the smoothing limit of variable gain function, it avoids the problem of secondary disturbance caused by parameter jump. The system dynamic adjustment process is smooth, the convergence speed is fast, and the steady-state accuracy is high. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the control method of the present invention; Figure 2 This is a system module structure block diagram of the present invention; Figure 3 This is a schematic diagram of the energy storage VSG topology studied in this invention; Figure 4 This is a block diagram of the VSG control structure of the present invention; Figure 5 This is a schematic diagram of the typical response curves of the VSG output active power versus angular frequency under a power command step disturbance. Figure 6 This is a comparison chart of the active power output of VSG before and after the improvement under the condition of sudden change in load power; Figure 7 This is a comparison chart of VSG output frequencies before and after the improvement under the condition of sudden change in load power; Figure 8 This is a comparison chart of the active power output of VSG before and after the improvement under the condition of sudden change in rated reference power; Figure 9 This is a comparison chart of the VSG output frequencies before and after the improvement under the condition of sudden change in rated reference power; Figure 10 This is a comparison chart of the active power output of VSG before and after the improvement under the grid-connected-island mode switching condition; Figure 11 This is a comparison chart of VSG output frequencies before and after the improvement under the grid-connected-island mode switching condition. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figures 1 to 11 As shown, the embodiments of the present invention provide the following technical solutions: The adaptive inertia damping cooperative control method for energy storage VSG includes the following steps, which are detailed in the following reference. Figure 1 : S1, collect the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate to obtain the power change rate and frequency change rate; the energy storage VSG topology studied in this invention is as follows: Figure 3 As shown.

[0027] Specifically, the output electromagnetic power of the virtual synchronous generator (VSG) is collected. and output angular frequency Calculate the electromagnetic power deviation respectively and angular frequency deviation : (1); (2); In equations (1) and (2), For a given active power, Given a rated angular velocity.

[0028] Electromagnetic power deviation Transient components are extracted using high-pass filtering. The power change rate is then obtained by differentiation. ; angular frequency deviation Extracting transient components using high-pass filtering The rate of change of frequency is then obtained by differentiation. .

[0029] The cutoff frequency of the high-pass filter can be set according to the system's rated frequency and the desired transient response speed, typically between 1 and 5 Hz. For example, taking a 50 Hz power frequency system as an example, when the cutoff frequency is 2 Hz, slowly varying components below 2 Hz are effectively attenuated, while transient components above 2 Hz are retained, ensuring that the response is only to sudden disturbances and avoiding unnecessary parameter adjustments caused by slow changes in steady-state load. After extracting the transient components through high-pass filtering, differentiation can be performed to obtain information about the signal's trend; that is, the sign of the derivative value is used to characterize whether the deviation continues to expand or has begun to decline, providing a basis for judgment in the subsequent staged interval determination in Table 1. The aforementioned high-pass filtering is a well-known technique in the field of signal processing. Its function is to allow high-frequency components to pass while attenuating low-frequency components. In this invention, it is used to extract transient components from electromagnetic power deviation and angular frequency deviation, filtering out steady-state or slowly varying components.

[0030] S2, calculate the comprehensive status index based on the grid connection status coefficient and load level. The grid connection status coefficient is 1 for grid connection and 0 for islanding. The load level is the ratio of electromagnetic power to given active power. Based on grid connection status coefficient and load level Calculate the comprehensive status index : (3); In equation (3), This is the grid connection status coefficient, which is 1 when the system is connected to the grid and 0 when the system is isolated. For load level, It is the ratio of electromagnetic power to a given active power.

[0031] Comprehensive status indicators The physical meaning is as follows: When operating in grid-connected mode, the system frequency is supported by the main power grid, and control should focus on power point tracking accuracy; when operating in islanded mode, the system frequency lacks support, and control should focus on frequency stability. Load level The introduction of this is used to distinguish between light and heavy load conditions, that is, the tolerance for power deviation is lower under heavy load, and the tolerance for frequency deviation is lower under light load. The value ranges from 0 to 1. A larger value indicates that the system is currently closer to a grid-connected and heavily loaded operating condition, and the weight allocation should be more biased towards the power channel; a smaller value indicates that the system is currently closer to an islanded and lightly loaded operating condition, and the weight allocation should be more biased towards the frequency channel.

[0032] S3. Calculate the deviation weighting coefficient and the rate of change weighting coefficient based on the comprehensive state index. Then, weight and fuse the electromagnetic power deviation, angular frequency deviation, power change rate and frequency change rate based on the deviation weighting coefficient and the rate of change weighting coefficient to obtain the comprehensive deviation and comprehensive rate of change. According to the comprehensive status index Calculate the deviation weighting coefficient and rate of change weighting coefficient : (4); In equation (4), The deviation weighting coefficient is... The change rate weighting coefficient, The comprehensive status index, , The preset adjustment range, , These are the basic parameters of deviation and the basic parameters of rate of change, respectively. The current control cycle number; basic deviation parameters. and rate of change basic parameters It is updated iteratively cycle by cycle by the online optimizer (see S5 for details), rather than by a fixed constant.

[0033] In S3, the comprehensive deviation is calculated as follows: the change rate weighting coefficient multiplied by the absolute value of the power change rate, plus a minus change rate weighting coefficient multiplied by the absolute value of the frequency change rate; the comprehensive change rate is calculated as follows: the deviation weighting coefficient multiplied by the absolute value of the transient component of the electromagnetic power deviation, plus a minus deviation weighting coefficient multiplied by the absolute value of the angular frequency deviation. That is, the overall deviation and comprehensive rate of change Calculate according to the following formulas: (5); In equation (5), The comprehensive deviation characterizes the rate at which the system state deviates from steady state, and is used for subsequent virtual damping coefficients. Adaptive adjustment, i.e. The adjustment amount is proportional to the overall rate, suppressing rapid changes in power / frequency; The comprehensive rate of change characterizes the overall magnitude of the system state's deviation from steady state, and is used for subsequent virtual rotational inertia. Adaptive adjustment, i.e. The adjustment amount is proportional to the overall amplitude, providing inertial support to the system that matches the degree of deviation. Overall deviation A larger value indicates a faster change in system state, corresponding to a larger virtual damping coefficient. To suppress rapid changes; comprehensive A larger value indicates that the system deviates further from steady state, corresponding to a larger virtual moment of inertia. To provide stronger inertia support; weighting coefficient and The online optimizer in S5 updates in real time, ensuring that the weighted fusion always adapts to the current optimal control requirements of the system.

[0034] S4, input the electromagnetic power deviation sequence and angular frequency deviation sequence in the historical period into the prediction model, and output the future electromagnetic power deviation prediction value, angular frequency deviation prediction value and prediction confidence level; Electromagnetic power deviation sequence within historical periods and angular frequency deviation sequence Input the prediction model, output the future Electromagnetic power deviation prediction value of step Angular frequency deviation prediction value and prediction confidence .

[0035] The prediction model is an autoregressive (AR) model, whose predicted values ​​are composed of a linearly weighted sum of historical state vectors. Model coefficients are identified online using the recursive least squares method, and the forgetting factor can range from 0.95 to 0.99 to ensure the model can quickly track changes in the system's dynamic characteristics. The aforementioned AR model is a commonly used tool in time series analysis. Its basic assumption is that the current value can be predicted by a linear combination of historical values, a well-known technique in signal processing. For example, the AR model order is typically 4 to 8. Furthermore, the recursive least squares method is an online parameter identification method that can update model parameters periodically during system operation, a well-known algorithm in adaptive control.

[0036] The prediction confidence level The variance was calculated based on the historical prediction residuals: (6); In equation (6), The variance of the recent forecast residuals is obtained by performing variance statistics on the forecast residual sequences of the most recent control periods; The baseline variance is determined by the statistical value of the predicted residual variance under steady-state operating conditions. Prediction confidence level. The value range is from 0 to 1, that is... The closer the confidence level is to 1, the more reliable the prediction; the closer it is to 0, the less reliable the prediction. When the prediction confidence level is close to 1, it indicates that the current prediction result is highly reliable and the dynamic characteristics of the system are relatively stable. When the prediction confidence level decreases, it indicates that the dynamic characteristics of the system have changed significantly or that there are unmodeled disturbances. In this case, the prediction result needs to be treated with more caution.

[0037] In step S4, the prediction model is further used to estimate the expected change in state of charge over future periods based on the predicted electromagnetic power deviation value: (7); In equation (7), This represents the expected change in the state of charge. For the rated capacity of energy storage, for Electromagnetic power at any given time For a given active power, The prediction time domain is determined based on the system's dynamic response time constant. This estimated value is used for subsequent SOC proactive health management.

[0038] S5, the learning rate of the online optimizer is adaptively adjusted according to the prediction confidence, and the basic parameters of the deviation weight coefficient and the rate of change weight coefficient are iteratively updated by the online optimizer with the system performance index as the objective function. Based on prediction confidence The learning rate of the online optimizer is adaptively adjusted according to the following formula: (8); In equation (8), For adaptive learning rate, As the baseline learning rate, To adjust the amplitude coefficient, The prediction confidence level is... The current control cycle number; the prediction confidence level The smaller the adaptive learning rate The larger the base learning rate, the more aggressively the online optimizer searches for the optimal basic parameters when prediction reliability is low. and adjustment range It can be set according to the system convergence speed requirements.

[0039] The online optimizer uses system performance metrics as the objective function and iteratively updates the basic deviation parameters. and rate of change basic parameters The objective function is composed of a weighted sum of a frequency deviation integral term and a power deviation integral term. The frequency deviation integral term characterizes the overall deviation of the system frequency, and the power deviation integral term characterizes the overall deviation of the active power.

[0040] The weighting coefficient of the frequency deviation integral term increases as the maximum value of the predicted angular frequency deviation increases, and the weighting coefficient of the power deviation integral term increases as the maximum value of the predicted electromagnetic power deviation increases, so that the larger the prediction disturbance, the higher the optimization priority of the corresponding channel in the objective function; the maximum value of the predicted angular frequency deviation and the maximum value of the predicted electromagnetic power deviation are determined by the predicted values ​​in the future time domain output by the prediction model.

[0041] The deviation basic parameters and the basic parameters of the rate of change The online optimizer is iteratively updated periodically using the following gradient descent method: (9); in, For adaptive learning rate, Let the objective function of the online optimizer be... This is the sequence number of the next control cycle; and These are the partial derivatives of the objective function with respect to the basic deviation parameter and the basic rate of change parameter, respectively, used to guide the basic deviation parameter and the basic rate of change parameter to be updated in the direction that reduces the objective function; the online optimizer performs micro-iterations on the basic deviation parameter and the basic rate of change parameter in each control cycle based on the real-time response performance of the system, thereby realizing online self-tuning of the basic weight parameters.

[0042] Between steps S5 and S6, the method further includes: predicting the trajectory of the state of charge in the future period based on the expected change in the state of charge; and when the predicted trajectory exceeds the safety boundary, actively correcting the given active power based on the deviation between the predicted state of charge and the safety boundary.

[0043] The expected change in state of charge obtained from equation (7) Predicting the trajectory of the state of charge in future time periods : (10); When the predicted trajectory exceeds the safety boundary, the given active power is actively corrected based on the deviation between the predicted state of charge and the safety boundary. (11); In equations (10) and (11), This is the corrected active power reference value. For active correction coefficients, For the corresponding safety boundary value ( or ) This is the upper limit of the safe state of charge. This is the lower limit of the safe state of charge. The upper and lower safety limits can be set according to the charging and discharging characteristics of the energy storage battery. This modification allows the energy storage to reduce charging and discharging power in advance, avoiding... Reaching the boundary triggers a protective shutdown. The correction force varies accordingly. The degree of approximation to the boundary increases linearly, that is, when The correction force is smaller when the distance is far from the safety boundary. The correction force increases to its maximum value when approaching the boundary.

[0044] S6 reshapes the curve of the variable gain function in real time according to the state of charge of the energy storage battery. The virtual moment of inertia and virtual damping coefficient are calculated based on the comprehensive deviation and comprehensive rate of change of the reshaped variable gain function. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to complete the closed-loop control.

[0045] Based on the state of charge of the energy storage battery The curve shape of the variable gain function is reshaped in real time. In step S6, the variable gain function is a variable gain function with a shape parameter: (12); In equation (12), The input variable for the variable gain function is... The shape parameters; The shape parameters change dynamically with the state of charge according to the following formula: (13); In equation (13), This refers to the state of charge of the energy storage battery. Based on the reference shape parameters, To adjust the amplitude coefficient, It is a monotonically increasing and bounded function of the state of charge; the... Satisfy: When the state of charge is at the preset healthy midpoint Set to 0, when the state of charge reaches the preset safety limit. Set to 1, when the state of charge reaches the preset safety lower limit. Set to -1; the shape parameter It increases monotonically as the degree to which the state of charge deviates from the healthy midpoint increases.

[0046] when When in a healthy zone far from the safety boundary, The aforementioned variable gain function exhibits a rapid rise followed by gradual saturation response, indicating a responsive system. When the state of charge approaches the safe boundary, the shape parameters... Increasing the gain function reshapes the response characteristics into a slow-rising and rapidly saturating state, resulting in a smoother and more flexible system response that effectively suppresses over-adjustment.

[0047] Calculate the adjustment coefficient using the reshaped variable gain function. and : (14); In equation (14), , This is the preset benchmark adjustment coefficient.

[0048] In step S6, the sign of the product of the angular frequency deviation and the rate of change of frequency determines whether parameter adjustment is required: when the product of the angular frequency deviation and the rate of change of frequency is greater than zero, the virtual moment of inertia and the virtual damping coefficient are in the dynamic adjustment range, and the virtual moment of inertia and the virtual damping coefficient are updated according to the calculated values; when the product of the angular frequency deviation and the rate of change of frequency is less than zero, the virtual moment of inertia and the virtual damping coefficient are in the recovery and maintenance range, and the virtual moment of inertia and the virtual damping coefficient are kept at their respective reference values ​​to accelerate the convergence of the system to steady state and avoid over-adjustment of parameters; the sign of the product is used to characterize the consistency of the direction of the angular frequency deviation and the rate of change of frequency.

[0049] Virtual moment of inertia and virtual damping coefficient Determine by the following formula: (15); (16); At the same time, if The calculated value is outside the preset range. If so, then take the corresponding boundary value; if Exceeding the preset range If so, then take the corresponding boundary value.

[0050] when At this time, the virtual moment of inertia and virtual damping coefficient are in the recovery and holding range, and the virtual moment of inertia and virtual damping coefficient are maintained at their respective reference values. , The parameters remain unchanged to accelerate the system's convergence to steady state and avoid over-tuning of parameters. (Product sign) The sign of the angular frequency deviation is used to characterize the consistency between the direction of the frequency change rate and the angular frequency deviation: when the signs are the same (product is greater than zero), the deviation will continue to increase and active adjustment is needed to suppress the deviation; when the signs are opposite (product is less than zero), the deviation has entered the convergence track and the parameters can be kept unchanged.

[0051] Finally, the virtual moment of inertia calculated in real time is... and virtual damping coefficient Substituting the VSG rotor motion equations into the equations, closed-loop control is completed, generating a PWM signal to drive the inverter. The VSG rotor motion equations are common knowledge in the field of VSG control, specifically describing the virtual moment of inertia. With damping coefficient The regulation mechanisms for system frequency and active power will not be described in detail here. The complete control structure is as follows: Figure 4 As shown.

[0052] Figure 5 Typical response curves of VSG output active power and angular frequency under a power command step disturbance are presented. The single complete oscillation process of the virtual synchronous generator output angular frequency can be divided into four characteristic stages, and the steady-state offset and instantaneous rate of change of the angular frequency exhibit differentiated dynamic evolution patterns in different stages. Based on this, this invention addresses the virtual inertia... With damping coefficient The design incorporates a phased adaptive control mechanism, enabling the dynamic parameters of both devices to accurately respond to and adapt to the entire process of transient angular frequency adjustment.

[0053] Virtual Inertia With damping coefficient The collaborative adaptive adjustment rules are shown in Table 1.

[0054] Table 1:

[0055] In Table 1, For electromagnetic power deviation, The rate of change of electromagnetic power deviation. For angular frequency deviation, This represents the rate of change of angular frequency. Interval ① indicates a stage where the angular frequency deviates upwards and is still accelerating; therefore, it needs to be increased simultaneously. and To suppress the rate of frequency rise and enhance damping; interval ② represents the stage where the angular frequency deviates upward but begins to decelerate and fall back, requiring a reduction in... To speed up response time Slightly larger to maintain the damping effect; interval ③ indicates the stage where the angular frequency deviates downward and is still accelerating downward, requiring simultaneous increase. and To suppress frequency drops; interval ④ represents the stage where the angular frequency deviates downward but begins to decelerate and rebound, requiring reduction. To speed up the recovery process Slightly larger to maintain the damping effect.

[0056] The adaptive inertia damping cooperative system of energy storage VSG is applied to the control method mentioned above. For details, please refer to [link / reference]. Figure 2 ,include: The signal acquisition and preprocessing module is used to acquire the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate them to obtain the power change rate and frequency change rate. The status assessment and weight calculation module is used to calculate a comprehensive status index based on the grid connection status coefficient and load level, calculate a deviation weight coefficient and a change rate weight coefficient based on the comprehensive status index, and perform weighted fusion of the electromagnetic power deviation, the angular frequency deviation, the power change rate and the frequency change rate based on the deviation weight coefficient and the change rate weight coefficient to obtain a comprehensive deviation and a comprehensive change rate. The prediction module is used to input the electromagnetic power deviation sequence and angular frequency deviation sequence from the historical period into the prediction model, and output the predicted values ​​of future electromagnetic power deviation, angular frequency deviation, and prediction confidence. The prediction optimization collaboration module is used to adaptively adjust the learning rate according to the prediction confidence, and iteratively update the basic parameters of the deviation weight coefficient and the rate of change weight coefficient with the system performance index as the objective function. The energy storage state management module is used to monitor the state of charge and health of the energy storage battery and actively correct the given active power based on the predicted trajectory of the state of charge. The variable gain calculation and execution module is used to reshape the curve shape of the variable gain function in real time according to the state of charge. The reshaped variable gain function is used to calculate the virtual moment of inertia and virtual damping coefficient based on the comprehensive deviation and the comprehensive rate of change. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to generate a pulse width modulation signal to drive the inverter.

[0057] The following are examples of the workflow of each core module in this embodiment under typical application scenarios.

[0058] After system startup, the signal acquisition and preprocessing module acquires the output electromagnetic power and angular frequency of the VSG in real time at a set sampling period, calculates the electromagnetic power deviation and angular frequency deviation respectively, extracts transient components after high-pass filtering, and then differentiates to obtain the power change rate and frequency change rate. Simultaneously, the state assessment and weight calculation module calculates a comprehensive state index in real time based on the current grid connection status (grid-connected or islanded) and load level. This index changes smoothly with operating conditions, ensuring a smooth transition between grid-connected and islanded modes.

[0059] The prediction module takes the electromagnetic power deviation sequence and angular frequency deviation sequence within the historical window as input, and uses an online-identified autoregressive model to output the predicted deviation value and prediction confidence for the short-term future time domain. When the prediction confidence is high, it indicates that the current prediction result is reliable, and the online optimizer iteratively updates the basic deviation parameters and rate of change parameters with a normal step size. When the prediction confidence decreases, it indicates that the dynamic characteristics of the system have changed significantly or that there is an unmodeled disturbance. The online optimizer automatically increases the learning rate and searches for the optimal basic parameters at a more aggressive pace, allowing the weight coefficients to quickly adapt to the new system state. The objective function of the online optimizer comprehensively weighs the integral performance of frequency deviation and power deviation, where the weights of the frequency channel and the power channel are automatically allocated according to the intensity of the predicted disturbance; that is, the channel with a larger predicted disturbance receives a higher optimization priority.

[0060] After updating the weighting coefficients, the system enters the SOC active health management phase. The prediction module estimates the expected change in the state of charge (SOC) over future periods based on the electromagnetic power deviation prediction value. When it is predicted that the SOC will exceed the safety boundary, the energy storage state management module proactively corrects the given active power, enabling the energy storage to adjust its charge and discharge depth in advance, effectively preventing the SOC from triggering a protective shutdown. The correction intensity increases as the SOC approaches the safety boundary, ensuring maximum utilization of energy storage capacity while guaranteeing stable system operation.

[0061] The variable gain calculation module reshapes the variable gain function curve in real time based on the current state of charge. When the state of charge is in the healthy range, the variable gain function exhibits a rapid rise followed by gradual saturation, resulting in a responsive system. When the state of charge approaches the safe boundary, the shape parameter increases, and the variable gain function is reshaped into a slow rise followed by rapid saturation, resulting in a smoother and more flexible system response that effectively suppresses over-adjustment. Finally, the variable gain calculation and execution module calculates the virtual moment of inertia and virtual damping coefficient based on the comprehensive deviation and comprehensive rate of change, substitutes them into the VSG rotor motion equation, generates a PWM signal to drive the inverter, and completes the entire control closed loop.

[0062] The above modules are executed sequentially in each control cycle to form a complete control link, enabling continuous, smooth, and adaptive adjustment of the inertia damping parameters of the energy storage VSG.

[0063] Simulation data verification: To verify the effectiveness of the control strategy proposed in this invention, a single VSG grid-connected simulation model was built in PSCAD / EMTDC. Three typical disturbance conditions were designed: load power change, rated reference power change, and grid-connected-island switching. The disturbance was triggered at t=1.0s, and the dynamic response characteristics of the traditional fixed parameter strategy and the scheme of this invention were compared. For details, please refer to the appendix. Figure 6 To the attached Figure 11 , attached Figure 6 To the attached Figure 11 In the attached diagram, f represents the VSG output frequency. This indicates the electromagnetic power output by the VSG.

[0064] (1) Load power sudden change condition (load suddenly increases from 6kW to 8kW), the simulation results are as follows Figure 6 , Figure 7 As shown: Under the traditional control strategy, the lowest active power output of the VSG is 5631W, with an under-adjustment of 369W, accompanied by continuous oscillation. Under the proposed solution, the lowest dynamic power output is increased to 5952W, with an under-adjustment of only 48W, and the oscillation amplitude and duration are significantly suppressed. Under the traditional strategy, the VSG output frequency drops to a maximum of 49.943Hz, with a frequency deviation of 0.057Hz. Under the proposed solution, the lowest frequency output is increased to 49.973Hz, with a maximum frequency deviation of only 0.027Hz, a reduction of 52.6% compared to the traditional strategy, and overshoot during the recovery process is eliminated.

[0065] (2) Operating condition with sudden change in rated reference power ( (From 6kW to 8kW), simulation results are as follows Figure 8 , Figure 9 As shown: Under the traditional strategy, the peak active power output of the VSG reaches 8313W, with an overshoot of 313W; under the scheme of this invention, the peak active power is suppressed to 8049W, with an overshoot of only 49W. Under the traditional strategy, the peak frequency reaches 50.056Hz; under the scheme of this invention, the peak frequency is only 50.028Hz, which is 50% lower than the traditional strategy, while eliminating under-adjustment and continuous oscillation during the frequency regulation process.

[0066] (3) Grid-connected-island mode switching condition, simulation results are as follows: Figure 10 , Figure 11 As shown: Under the traditional strategy, the peak frequency climbs to 50.0756Hz, the recovery process is slow, and the steady-state fluctuations are obvious. Under the scheme of this invention, the peak frequency only rises to 50.0347Hz, which is significantly lower than the traditional strategy. It also eliminates overshoot and continuous oscillation in the frequency adjustment process, resulting in faster convergence speed and higher steady-state accuracy.

[0067] Simulation results for three typical operating conditions show that the proposed solution effectively suppresses power oscillations and frequency fluctuations caused by source-load disturbances and operating condition switching without weakening the fast tracking and smooth convergence performance of VSG active power, thus verifying the effectiveness and good adaptability of the solution.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0070] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. An adaptive inertia-damped coordinated control method for energy storage VSG, characterized in that, Includes the following steps: S1: Collect the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate them to obtain the power change rate and frequency change rate. S2, calculate the comprehensive status index based on the grid connection status coefficient and load level. The grid connection status coefficient is 1 for grid connection and 0 for islanding. The load level is the ratio of electromagnetic power to given active power. S3. Calculate the deviation weighting coefficient and the rate of change weighting coefficient based on the comprehensive state index. Then, weight and fuse the electromagnetic power deviation, angular frequency deviation, power change rate and frequency change rate based on the deviation weighting coefficient and the rate of change weighting coefficient to obtain the comprehensive deviation and comprehensive rate of change. S4, input the electromagnetic power deviation sequence and angular frequency deviation sequence in the historical period into the prediction model, and output the future electromagnetic power deviation prediction value, angular frequency deviation prediction value and prediction confidence level; S5, the learning rate of the online optimizer is adaptively adjusted according to the prediction confidence, and the basic parameters of the deviation weight coefficient and the rate of change weight coefficient are iteratively updated by the online optimizer with the system performance index as the objective function. S6 reshapes the curve of the variable gain function in real time according to the state of charge of the energy storage battery. The virtual moment of inertia and virtual damping coefficient are calculated based on the comprehensive deviation and comprehensive rate of change of the reshaped variable gain function. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to complete the closed-loop control.

2. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S4, the prediction model is also used to estimate the expected change in state of charge in the future period based on the electromagnetic power deviation prediction value. Between steps S5 and S6, the method further includes: predicting the trajectory of the state of charge in the future period based on the expected change in the state of charge; and when the predicted trajectory exceeds the safety boundary, actively correcting the given active power based on the deviation between the predicted state of charge and the safety boundary.

3. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S5, the learning rate is adaptively adjusted according to the following formula: ； in, For adaptive learning rate, As the baseline learning rate, To adjust the amplitude coefficient, The prediction confidence level is... The current control cycle number; the prediction confidence level The smaller the adaptive learning rate The larger the value, the more aggressively the online optimizer searches for the optimal basic parameters when the prediction reliability is low.

4. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S5, the objective function is composed of a weighted sum of a frequency deviation integral term and a power deviation integral term. The frequency deviation integral term characterizes the overall deviation of the system frequency, and the power deviation integral term characterizes the overall deviation of the active power. The weight coefficient of the frequency deviation integral term increases as the maximum value of the predicted angular frequency deviation increases, and the weight coefficient of the power deviation integral term increases as the maximum value of the predicted electromagnetic power deviation increases, so that the greater the prediction disturbance, the higher the optimization priority of the corresponding channel in the objective function. The maximum value of the predicted angular frequency deviation and the maximum value of the predicted electromagnetic power deviation are determined by the predicted values ​​in the future time domain output by the prediction model.

5. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S3, the comprehensive deviation is calculated as follows: the rate of change weighting coefficient multiplied by the absolute value of the power rate of change, plus a factor less than the rate of change weighting coefficient multiplied by the absolute value of the frequency rate of change; the comprehensive rate of change is calculated as follows: the deviation weighting coefficient multiplied by the absolute value of the transient component of the electromagnetic power deviation, plus a factor less than the deviation weighting coefficient multiplied by the absolute value of the angular frequency deviation; the comprehensive deviation represents the comprehensive rate at which the system state deviates from the steady state, and the comprehensive rate of change represents the comprehensive magnitude at which the system state deviates from the steady state, both of which are used for the subsequent adaptive adjustment of the virtual damping coefficient and the virtual moment of inertia.

6. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S4, the prediction model is an autoregressive model, and its predicted value is composed of a linear weighted sum of historical state vectors. The model coefficients are identified online using the recursive least squares method. The prediction confidence is calculated based on the variance of the historical prediction residuals, and the prediction confidence range is from 0 to 1. The variance of the recent prediction residuals is obtained by performing variance statistics on the prediction residual sequences of the most recent control cycles, and the benchmark variance is the benchmark value of the prediction residual variance under the steady-state operation conditions of the system.

7. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S6, the variable gain function is a variable gain function with shape parameters: ； in, The input variable for the variable gain function is... The shape parameters; The shape parameters change dynamically with the state of charge according to the following formula: ； in, This refers to the state of charge of the energy storage battery. Based on the reference shape parameters, To adjust the amplitude coefficient, It is a monotonically increasing and bounded function of the state of charge; the... Satisfy: When the state of charge is at the preset healthy midpoint Set to 0, when the state of charge reaches the preset safety limit. Set to 1, when the state of charge reaches the preset safety lower limit. Set to -1; the shape parameter It increases monotonically as the degree to which the state of charge deviates from the healthy midpoint increases.

8. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 1, characterized in that, In step S6, the sign of the product of the angular frequency deviation and the rate of change of frequency determines whether parameter adjustment is required: when the product of the angular frequency deviation and the rate of change of frequency is greater than zero, the virtual moment of inertia and the virtual damping coefficient are in the dynamic adjustment range, and the virtual moment of inertia and the virtual damping coefficient are updated according to the calculated values; when the product of the angular frequency deviation and the rate of change of frequency is less than zero, the virtual moment of inertia and the virtual damping coefficient are in the recovery and maintenance range, and the virtual moment of inertia and the virtual damping coefficient are kept at their respective reference values ​​to accelerate the convergence of the system to steady state and avoid over-adjustment of parameters; the sign of the product is used to characterize the consistency of the direction of the angular frequency deviation and the rate of change of frequency.

9. The adaptive inertia damping cooperative control method for energy storage VSG according to claim 2, characterized in that, The deviation weighting coefficient and the rate of change weighting coefficient are calculated according to the following formulas: ； in, The deviation weighting coefficient is... The change rate weighting coefficient, The comprehensive status index, , The preset adjustment range, , These are the basic parameters of deviation and the basic parameters of rate of change, respectively. This is the current control cycle number; The deviation basic parameters and the basic parameters of the rate of change The online optimizer is iteratively updated periodically using the following gradient descent method: ； in, For adaptive learning rate, Let the objective function of the online optimizer be... This is the sequence number of the next control cycle; and These are the partial derivatives of the objective function with respect to the basic deviation parameter and the basic rate of change parameter, respectively, used to guide the basic deviation parameter and the basic rate of change parameter to be updated in the direction that reduces the objective function; the online optimizer performs micro-iterations on the basic deviation parameter and the basic rate of change parameter in each control cycle according to the real-time response performance of the system, thereby realizing online self-tuning of the basic weight parameter.

10. An adaptive inertia-damped cooperative system for energy storage VSG, applied to the control method described in claim 1, characterized in that, include: The signal acquisition and preprocessing module is used to acquire the electromagnetic power and angular frequency of the virtual synchronous generator, calculate the electromagnetic power deviation and angular frequency deviation, extract the transient components of the electromagnetic power deviation and angular frequency deviation through high-pass filtering, and then differentiate them to obtain the power change rate and frequency change rate. The status assessment and weight calculation module is used to calculate a comprehensive status index based on the grid connection status coefficient and load level, calculate a deviation weight coefficient and a change rate weight coefficient based on the comprehensive status index, and perform weighted fusion of the electromagnetic power deviation, the angular frequency deviation, the power change rate and the frequency change rate based on the deviation weight coefficient and the change rate weight coefficient to obtain a comprehensive deviation and a comprehensive change rate. The prediction module is used to input the electromagnetic power deviation sequence and angular frequency deviation sequence from the historical period into the prediction model, and output the predicted values ​​of future electromagnetic power deviation, angular frequency deviation, and prediction confidence. The prediction optimization collaboration module is used to adaptively adjust the learning rate according to the prediction confidence, and iteratively update the basic parameters of the deviation weight coefficient and the rate of change weight coefficient with the system performance index as the objective function. The energy storage state management module is used to monitor the state of charge and health of the energy storage battery and actively correct the given active power based on the predicted trajectory of the state of charge. The variable gain calculation and execution module is used to reshape the curve shape of the variable gain function in real time according to the state of charge. The reshaped variable gain function is used to calculate the virtual moment of inertia and virtual damping coefficient based on the comprehensive deviation and the comprehensive rate of change. The virtual moment of inertia and virtual damping coefficient are substituted into the rotor motion equation of the virtual synchronous generator to generate a pulse width modulation signal to drive the inverter.

Images