Power regulation method and system for distributed photovoltaic power station cluster

Abstract

The application belongs to the technical field of power system automation, and particularly relates to a power regulation method and system for a distributed photovoltaic power station cluster, which comprises the following steps: collecting real-time operation data of a grid-connected point of the distributed photovoltaic power station to obtain a voltage vector amplitude and a frequency change rate; constructing a power grid state non-stationarity index by using the voltage vector amplitude and an active power prediction error; dynamically adjusting a forgetting factor in an impedance identification process according to the power grid state non-stationarity index to realize online identification of an equivalent impedance modulus of the power grid; and evaluating the dynamic stiffness of the power grid based on the identified impedance and the frequency change rate, and adaptively mapping a damping weight of a model predictive controller according to the evaluation. The application realizes active sensing of the strength of the power grid environment, can automatically adjust the control damping according to the stiffness of the power grid, effectively suppresses wideband oscillation under a weak power grid, and improves the grid connection stability and safety of the distributed photovoltaic cluster under complex power grid working conditions.

Description

Technical Field

[0001] This invention relates to the field of power system automation technology. More specifically, this invention relates to a power regulation method and system for distributed photovoltaic power plant clusters. Background Technology

[0002] Against the backdrop of the ongoing macro-strategic push for dual-carbon goals, the penetration rate of distributed photovoltaic (PV) power station clusters in modern power distribution networks is showing a significant upward trend. However, due to geographical and resource constraints, many distributed PV access points are typically located at the end of the power distribution network or in remote areas. These areas often exhibit significant weak grid characteristics due to their long transmission distances and relatively weak grid structures. Specifically, the core manifestation of this weak grid characteristic is that the equivalent impedance of the grid is relatively large, and it exhibits high time-varying characteristics as the grid's operating conditions change. This makes the physical structure of the grid extremely complex and variable, thereby affecting the safe grid connection and stable operation of distributed PV power stations.

[0003] Faced with the challenges of weak grid environments, traditional photovoltaic grid-connected control schemes, while incorporating grid impedance identification technology and model predictive control strategies, often suffer from parameter rigidity. In the impedance identification stage, most commonly used identification algorithms, such as recursive least squares, employ fixed forgetting factors. Under stable operating conditions, fixed forgetting factors help filter out measurement noise and ensure the steady-state accuracy of impedance identification results. However, when the grid operating environment undergoes sudden changes, or when natural meteorological factors such as cloud cover cause abrupt changes in the output of the photovoltaic cluster, the fixed forgetting factor leads to excessively long algorithm memory. This not only fails to quickly track real-time changes in grid impedance but also results in severe impedance identification lag or large numerical oscillations, leading to extremely distorted grid state information obtained by the power system and creating hidden dangers for subsequent control decisions.

[0004] In the actual execution of model predictive control strategies, existing model predictive controllers also face serious rigidity limitations. Their smoothing weights are usually preset to a fixed value. In traditional strong grid environments, this fixed weight setting can indeed meet the conventional needs of power systems to quickly adjust power. However, when the grid instantly becomes a weak grid due to a sudden fault or internal topology switching, the limitations of fixed parameter control will be fully exposed. At this time, due to the lack of necessary damping adjustment capability of the controller, the excessively small smoothing weight will cause a strong dynamic coupling between the current output by the photovoltaic inverter and the high-impedance weak grid, thereby inducing severe broadband oscillations in the distribution system. In extreme cases, it may even lead to a large-scale grid disconnection accident of the photovoltaic power plant cluster.

[0005] Therefore, overcoming the lack of operational stability caused by parameter rigidity and fixed control strategies, enabling the power system to actively sense the strength of the grid environment, and adaptively adjust the control flexibility according to the grid status, thereby effectively suppressing broadband oscillations under weak grid conditions, is the core technical problem that urgently needs to be solved to ensure the safe grid connection of distributed photovoltaic clusters. Summary of the Invention

[0006] To address the technical problems of inaccurate impedance identification and insufficient operational stability caused by fixed control strategy parameters in distributed photovoltaic clusters under weak grid environments, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a power regulation method for a distributed photovoltaic power station cluster, comprising: real-time acquisition of operating data of the grid-connected points of the distributed photovoltaic power station; performing coordinate transformation and feature extraction on the operating data to obtain voltage vector amplitude and frequency change rate; constructing a grid state non-stationarity index using the voltage vector amplitude and active power prediction error; dynamically adjusting the forgetting factor in the impedance identification process according to the grid state non-stationarity index to achieve online identification of the grid equivalent impedance magnitude; evaluating the grid dynamic stiffness coefficient based on the grid equivalent impedance magnitude and the frequency change rate; adaptively mapping the smoothing weights of the model predictive controller according to the grid dynamic stiffness coefficient; updating the smoothing weights to the cost function of the model predictive controller; solving for the optimal voltage vector through rolling optimization and driving the inverter to perform power regulation.

[0008] This invention constructs a grid state non-stationarity index to achieve a dynamic and refined characterization of the grid's operating state. This index integrates voltage fluctuation statistics and power prediction errors, enabling it to sensitively distinguish between steady-state and transient conditions. Based on this, it dynamically adjusts the forgetting factor of the impedance identification algorithm, allowing the power system to filter out noise during stable periods and quickly track abrupt changes. This significantly improves the accuracy and dynamic performance of online equivalent impedance identification. Simultaneously, by comprehensively evaluating the grid's dynamic stiffness coefficient using the real-time equivalent impedance magnitude and frequency change rate, the invention accurately obtains the grid's strength and weakness. Furthermore, a nonlinear function is used to smoothly map the grid's dynamic stiffness coefficient to the damping weights of the model predictive controller. Through this mechanism, when grid stiffness decreases, the damping weights automatically increase to enhance control damping; when grid stiffness increases, the damping weights decrease accordingly. This achieves adaptive adjustment of control parameters, effectively suppressing the risk of broadband oscillations under weak grid conditions, thereby improving the stability and security of photovoltaic cluster grid connection.

[0009] Preferably, the operating data includes the instantaneous values ​​of the three-phase voltage and the three-phase current at the grid connection point. The operating data is subjected to coordinate transformation and feature extraction. The instantaneous values ​​of the three-phase voltage and the three-phase current in the stationary coordinate system are transformed into components in the synchronous rotating coordinate system. The voltage vector amplitude at the current moment is calculated based on the square root of the sum of the squares of the two-axis voltage components.

[0010] Preferably, the method for obtaining the frequency change rate is as follows: using the frequency data output by the phase-locked loop, the absolute value of the frequency change rate is calculated using the sliding window difference method.

[0011] Preferably, the construction of the grid state non-stationarity index includes: obtaining the standard deviation of the voltage vector amplitude within a sliding time window; calculating the ratio of the absolute value of the active power prediction error at the current moment to the rated installed capacity of the photovoltaic power station, adding 1 to the ratio and calculating the natural logarithm, and adding the obtained natural logarithm to 1 to obtain the logarithmic adjustment term; multiplying the standard deviation of the voltage vector amplitude by the logarithmic adjustment term to obtain the grid state non-stationarity index.

[0012] Preferably, the forgetting factor in the impedance identification process is dynamically adjusted according to the power grid state non-stationarity index, including: multiplying the power grid state non-stationarity index by a preset attenuation adjustment coefficient and taking the negative value as the exponential term of the natural index to calculate the attenuation coefficient; calculating the difference between the upper limit and the lower limit of the forgetting factor, multiplying the difference by the attenuation coefficient to obtain the dynamic increment; and adding the lower limit of the forgetting factor to the dynamic increment to obtain the adjusted forgetting factor.

[0013] The power grid state non-stationarity index constructed in this invention analyzes the statistical characteristics of voltage fluctuations and normalizes power mutations using a logarithmic function, thereby achieving accurate identification of the power system's operating state. It can effectively distinguish between steady-state background noise and transient disturbances. Based on this identification result, the power system can provide an objective basis for dynamically adjusting the forgetting factor in the impedance identification algorithm, enabling it to focus on noise suppression during stable periods and enhance tracking capabilities during mutation periods. This ensures that the identification algorithm maintains fast and stable convergence performance under different disturbance scenarios, thus significantly improving the accuracy and stability of online identification of the power grid's equivalent impedance.

[0014] Preferably, the step of evaluating the dynamic stiffness coefficient of the power grid based on the equivalent impedance modulus of the power grid and the frequency change rate includes: adding the real-time identified equivalent impedance modulus of the power grid to a preset non-zero protection constant to obtain a corrected impedance value; calculating the ratio of the effective value of the rated voltage at the grid connection point to the corrected impedance value to obtain a reference stiffness term; multiplying a preset frequency sensitivity coefficient by the frequency change rate and taking the negative value as the exponential term of the natural exponent to calculate the frequency attenuation coefficient; and multiplying the reference stiffness term by the frequency attenuation coefficient to obtain the dynamic stiffness coefficient of the power grid at the current moment.

[0015] This invention introduces a correction for the dynamic stiffness coefficient of the power grid by introducing an exponential decay term in the relational formula. It not only considers the impedance strength under Ohm's law, but also takes into account the frequency stability of the power grid. This makes the assessment of the dynamic stiffness coefficient of the power grid more physically convincing and can accurately identify scenarios in which the power generation and consumption in the power grid are unbalanced, resulting in rapid or large changes in the frequency of the power grid.

[0016] Preferably, the step of adjusting the smoothing weights of the model predictive controller based on the adaptive mapping of the power grid dynamic stiffness coefficient includes: calculating the ratio of the power grid dynamic stiffness coefficient to the reference stiffness benchmark value at the current moment, and calculating the hyperbolic tangent of the ratio; subtracting the hyperbolic tangent from 1 to obtain the weight adjustment coefficient; calculating the difference between the maximum and minimum values ​​of a preset smoothing weight adjustment range, multiplying the difference by the weight adjustment coefficient to obtain the weight increment; and adding the minimum value of the preset smoothing weight adjustment range to the weight increment to obtain the smoothing weights updated in the cost function of the model predictive controller.

[0017] This invention utilizes the nonlinear range of the hyperbolic tangent function to achieve a smooth transition of the weight during the switching process between strong and weak power grids. This ensures that the power system does not experience abrupt changes in control parameters when the power grid environment changes, enabling the power system to adapt to complex and ever-changing operating conditions while maintaining the stability of the control process and avoiding new instability caused by parameter mutations.

[0018] Preferably, the mapping relationship of the smoothing weight of the predictor based on the adaptive mapping model of the power grid dynamic stiffness coefficient is as follows: when the power grid dynamic stiffness coefficient increases, the smoothing weight is reduced through nonlinear mapping; when the power grid dynamic stiffness coefficient decreases, the smoothing weight is increased through nonlinear mapping.

[0019] Preferably, the power regulation of the drive inverter includes: sending the optimal voltage vector obtained by rolling optimization to the pulse width modulation module to drive the power switching transistors of the photovoltaic inverter to operate.

[0020] Secondly, the present invention provides a power regulation system for a distributed photovoltaic power station cluster, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the power regulation method for the distributed photovoltaic power station cluster described above is implemented.

[0021] By adopting the above technical solution, the power regulation method of the distributed photovoltaic power station cluster is generated into a computer program and stored in a memory for loading and execution by a processor. Terminal equipment is then made based on the memory and processor for easy use. By deploying it in an existing photovoltaic monitoring system or energy management system, remote and dynamic adjustment of the control parameters of each inverter in the cluster can be achieved, which greatly reduces the cost of manual debugging and operational safety hazards.

[0022] The beneficial effects of this invention are as follows: This invention achieves online closed-loop tuning of control parameters through data feature perception. The power system can dynamically sense the strength of the grid at the grid connection point and automatically and smoothly adjust the output characteristics of the controller accordingly. This adaptive mechanism can break through the limitations of traditional fixed parameter control and fundamentally improve the interaction stability between the inverter and the weak grid.

[0023] Furthermore, it not only improves the accuracy and stability of the impedance identification algorithm under dynamic disturbances, but also predicts the smoothing weight of the controller through the adaptive mapping model of the grid dynamic stiffness coefficient, realizing deep matching between the control system and the complex grid environment, and providing core technical support for peak shaving, frequency regulation and safe grid connection of high-penetration distributed photovoltaic. Attached Figure Description

[0024] Figure 1 This is a flowchart of the power regulation method for a distributed photovoltaic power station cluster in an embodiment of the present invention; Figure 2 This is a comparative schematic diagram of the online identification accuracy of power grid impedance in embodiments of the present invention; Figure 3 This is a comparative schematic diagram of the voltage stability at the grid connection point under weak grid disturbances in an embodiment of the present invention; Figure 4 This is a schematic diagram of the adaptive smoothing weight dynamic adjustment process in an embodiment of the present invention. 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, not all, of the embodiments of the present invention. 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] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] This invention discloses a power regulation method for distributed photovoltaic power station clusters, referring to... Figure 1 This includes steps S1-S4: S1. Data collection and feature extraction of grid connection points.

[0028] Real-time acquisition of operational data from grid-connected points of distributed photovoltaic power stations; coordinate transformation and feature extraction of operational data to obtain voltage vector amplitude and frequency change rate.

[0029] In an optional embodiment, the control center of the distributed photovoltaic power station continuously acquires the three-phase voltage and three-phase current signals on the AC side through high-speed measuring equipment installed at the grid connection point. To facilitate subsequent DC calculations, the power system performs a coordinate transformation, projecting the instantaneous values ​​of the three-phase voltage and current in the abc coordinate system onto a synchronously rotating coordinate system, resulting in mutually orthogonal... Axis voltage components and Axis voltage components Subsequently, according to shaft and The voltage vector magnitude at the current moment is calculated by taking the square root of the sum of the squares of the axis voltage components. The calculation formula is as follows: .

[0030] For example, suppose that at a certain sampling time, the coordinate transformation is used to obtain... It is 311V. The value is 0V. Substituting this into the above formula, we obtain the result. The voltage vector amplitude at the current moment is 311V; if a single-phase fluctuation occurs in the power grid, leading to... If a deviation of 10V is produced, the calculation will yield... Meanwhile, by analyzing the grid frequency curve output by the phase-locked loop, the frequency fluctuation value per unit time is calculated to obtain the frequency change rate.

[0031] In this way, the power system completes the real-time acquisition, coordinate transformation and feature extraction of electrical signals at the grid connection point. This real-time and accurate calculation of voltage amplitude and frequency change rate provides an accurate and reliable data foundation for subsequent online identification of grid impedance, grid stiffness assessment and adaptive control decision-making.

[0032] S2. Non-stationarity assessment and online identification of equivalent impedance.

[0033] A grid state non-stationarity index is constructed using voltage vector magnitude and active power prediction error. The forgetting factor in the impedance identification process is dynamically adjusted based on the grid state non-stationarity index to achieve online identification of the grid equivalent impedance magnitude.

[0034] In an optional embodiment, after acquiring the voltage sequence, the power system calculates the voltage fluctuation by using the length of the sliding time window. Simultaneously, it uses the measured power value from the previous moment as a prediction reference, calculates the difference between this and the current actual power, and constructs a grid state non-stationarity index using the voltage vector magnitude and the absolute value of the active power prediction error. ; in, For the current moment The power grid non-stationarity index, The length of the sliding time window. For the first Voltage vector magnitude at time t. The arithmetic mean of the voltage vector magnitudes within the sliding window. For the current moment The absolute value of the active power prediction error. This refers to the rated installed capacity of the photovoltaic power station.

[0035] When the power grid is disturbed by load fluctuations or sudden changes in illumination, the fluctuation amplitude of voltage and power will be significantly aggravated, thereby increasing the power grid non-stationarity index accordingly. This power grid non-stationarity index can comprehensively characterize the severity of voltage fluctuations and power changes.

[0036] The calculated By substituting the preset forgetting factor adjustment function, the forgetting factor can be dynamically set. When the power grid is running smoothly, the forgetting factor approaches 1 to effectively filter out measurement noise. Specifically, a mapping function based on exponential decay is used: .in, Forgetting factor, The upper limit of the forgetting factor is close to 1. This is the lower limit of the forgetting factor. The preset attenuation adjustment coefficient is used to make the forgetting factor of the power system approach 1 in a steady state to filter out noise, and to reduce the forgetting factor in a transient state to quickly track impedance changes. For example, the following is set: It is 0.99. It is 0.9. The value is 0.5 when the power grid is stable. The value approaches 0, at which point the forgetting factor is 0.99, which approaches 1, effectively filtering out measurement noise; when voltage or power experiences transient changes, The exponential term decreases rapidly as the value increases significantly, thereby reducing the forgetting factor and enabling rapid tracking of actual impedance changes. By using this dynamically adjusted forgetting factor for online identification through methods such as recursive least squares, a more accurate equivalent impedance modulus of the power grid can be obtained.

[0037] For example, setting the length of the sliding time window to 100 sampling points, the calculated average voltage within the sliding time window is 311V, with a standard deviation of 2. The active power prediction error caused by the current photovoltaic output fluctuation is 20kW, and the rated installed capacity of the photovoltaic power station is 1000kW. Substituting these values ​​into the formula: Calculations show that the grid non-stationarity index is 2.04. Based on this grid non-stationarity index, the power system lowers the forgetting factor of the impedance identification algorithm from 1 to 0.95, so that the impedance identification algorithm will refer more to the new data at the current moment, thereby quickly identifying the change in the equivalent impedance magnitude of the grid.

[0038] Thus, by accurately assessing the stability of power grid operation and dynamically adjusting the forgetting factor, the impedance identification algorithm acquires the core capability of adaptive operation. In stable conditions, it filters out noise with a high forgetting factor to ensure steady-state accuracy; in the event of abrupt changes, it rapidly reduces the forgetting factor to achieve rapid tracking of impedance changes, providing a real-time and reliable foundation for power grid state perception for subsequent stiffness assessment and adaptive control.

[0039] S3, Power grid dynamic stiffness assessment and smoothing weight mapping.

[0040] The dynamic stiffness coefficient of the power grid is evaluated based on the equivalent impedance modulus and frequency change rate of the power grid, and the smoothing weight of the controller is predicted based on the adaptive mapping model of the dynamic stiffness coefficient of the power grid.

[0041] In an optional embodiment, the identified equivalent impedance magnitude of the power grid is... and frequency change rate Combined, the dynamic stiffness coefficient of the power grid at the current moment is calculated using the following relationship: ; in, The current dynamic stiffness coefficient of the power grid. The effective value of the rated voltage at the grid connection point. To prevent the protection constant from having a denominator of zero, its dimension is set to ohms to maintain consistency with the grid impedance. The frequency sensitivity coefficient is set to the reciprocal of time and frequency (s / Hz) to ensure dimensionlessness within the exponential term. This dynamic stiffness coefficient of the power grid comprehensively reflects the impedance strength and frequency stability of the power grid.

[0042] Subsequently, based on the range of the dynamic stiffness coefficient of the power grid, the smoothing weight of the model predictor controller is calculated using a nonlinear mapping function. The calculation formula is as follows: ; in, This is the minimum value of the preset range for adjusting the slack weight. This is the maximum value of the preset range for adjusting the slack weight. This serves as a reference stiffness benchmark value.

[0043] For example, setting the effective value of the rated voltage at the grid connection point to 10000V, the currently identified equivalent impedance modulus of the grid to 4Ω, the frequency change rate to 0.1Hz / s, the protection constant to 0.01, and the frequency sensitivity coefficient to 0.5, substituting these values ​​into the above formula yields the following result: ; If the remote line disconnects at this time, the equivalent impedance of the power grid increases to 8Ω, and the rate of frequency change increases to 2Hz / s due to the power shortfall. Recalculation yields: ; At this point, the hyperbolic tangent function in the smoothing weight mapping formula enters the nonlinear rising region, automatically increasing the smoothing weight from the minimum value of 20 to 85. This process breaks through the rigidity limitation of traditional fixed parameter control and realizes closed-loop adaptive tuning of control parameters to the grid state. When the grid stiffness decreases and the oscillation risk increases, the power system increases the smoothing weight, which is equivalent to increasing the damping ratio of the control loop in real time, thereby suppressing the wideband coupling oscillation that may occur between the inverter and the weak grid from the source.

[0044] In this way, by smoothly mapping the physical stiffness of the power grid from a nonlinear function to the damping weights of the control layer, the power system can automatically match the optimal damping characteristics under the switching between strong and weak power grids and various disturbances, which significantly improves the dynamic stability, anti-interference ability and overall operational resilience of photovoltaic clusters under complex power grid conditions.

[0045] S4. Rolling optimization solution and inverter power regulation.

[0046] The smoothing weights are updated in the cost function of the model predictive controller, and the optimal voltage vector is solved through rolling optimization to drive the inverter to perform power regulation.

[0047] In an optional embodiment, the power system sends the latest calculated slack weights to the execution layer in real time. The model predictive controller performs mathematical solutions using a cost function that includes the new slack weights within each control step. The new cost function will more strictly limit voltage deviation, thereby making the optimal voltage vector more robust during execution. The voltage command is converted into a high-frequency switching signal through a pulse width modulation module to control the operation of the inverter arm and complete the power injection into the grid.

[0048] Thus, through the complete process of real-time parameter sensing, online identification, weight-adaptive mapping and closed-loop control, the power system can significantly improve the dynamic stability and anti-interference capability of distributed photovoltaic clusters under weak grids and complex operating conditions, effectively ensuring grid security.

[0049] This scheme deeply maps the "short-circuit ratio" and the physical laws of inertia support in a weak power grid environment. Firstly, the dynamic stiffness coefficient of the power grid... benchmark items In physical terms, it characterizes the equivalent short-circuit current at the grid connection point under the current impedance; the exponential decay term introduced on this basis is a nonlinear penalty for the lack of system inertia when the grid frequency drops; secondly, it smooths out the weights. The hyperbolic tangent function was used as the boundary of the control theory mapping when Much larger than the benchmark value When the hyperbolic tangent function approaches 1, the penalty term returns to zero, and the system maintains normal output; however, under extremely weak power grid conditions with high impedance and large frequency deviation, The sharp reduction leads to a decrease in the value of the hyperbolic tangent function, which smooths out the nonlinear ramping of the weights and enables a smooth, disturbance-free switching between high efficiency in strong networks and stable performance in weak networks from a mathematical perspective.

[0050] Reference Figure 2 Existing technologies produce huge overshoot at impedance step points, while the present invention, by adjusting the forgetting factor in a timely manner, produces a smooth waveform and can quickly lock in new values.

[0051] Reference Figure 3 When entering extremely weak network conditions, existing technologies induce large voltage fluctuations, while the present invention, with its adaptively increased smoothing weight, enables the voltage to quickly return to stability.

[0052] Reference Figure 4 When the physical stiffness of the power grid decreases, the stabilization weight immediately and spontaneously increases to support the power system.

[0053] This invention also discloses a power regulation system for a distributed photovoltaic power station cluster, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a power regulation method for a distributed photovoltaic power station cluster according to the present invention.

[0054] In an alternative embodiment, by deploying it in an existing photovoltaic monitoring system or energy management system, remote and dynamic tuning of the control parameters of each inverter in the cluster can be achieved, which greatly reduces the cost of manual debugging and operational safety hazards.

[0055] The power regulation system of the aforementioned distributed photovoltaic power station cluster also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

Claims

1. A power regulation method for distributed photovoltaic power station clusters, characterized in that, include: Real-time acquisition of operation data from grid-connected points of distributed photovoltaic power stations; coordinate transformation and feature extraction of the operation data to obtain voltage vector amplitude and frequency change rate; A grid state non-stationarity index is constructed using the voltage vector magnitude and active power prediction error. The forgetting factor in the impedance identification process is dynamically adjusted based on the grid state non-stationarity index to achieve online identification of the grid equivalent impedance magnitude. The dynamic stiffness coefficient of the power grid is evaluated based on the equivalent impedance modulus of the power grid and the frequency change rate, and the smoothing weight of the controller is predicted based on the adaptive mapping model of the dynamic stiffness coefficient of the power grid. The smoothing weights are updated in the cost function of the model predictive controller, and the optimal voltage vector is solved by rolling optimization to drive the inverter to perform power regulation.

2. The power regulation method for a distributed photovoltaic power station cluster according to claim 1, characterized in that, The operational data includes the instantaneous values ​​of the three-phase voltage and the three-phase current at the grid connection point. The operational data is subjected to coordinate transformation and feature extraction. The instantaneous values ​​of the three-phase voltage and the three-phase current in the stationary coordinate system are transformed into components in the synchronous rotating coordinate system. The voltage vector amplitude at the current moment is calculated based on the square root of the sum of the squares of the two-axis voltage components.

3. The power regulation method for a distributed photovoltaic power station cluster according to claim 1, characterized in that, The method for obtaining the frequency change rate is as follows: using the frequency data output by the phase-locked loop, the absolute value of the frequency change rate is calculated using the sliding window difference method.

4. The power regulation method for a distributed photovoltaic power station cluster according to claim 1, characterized in that, The construction of the power grid state non-stationarity index includes: Obtain the standard deviation of the voltage vector magnitude within the sliding time window; Calculate the ratio of the absolute value of the active power prediction error at the current moment to the rated installed capacity of the photovoltaic power station, add 1 to the ratio and calculate the natural logarithm, and add the obtained natural logarithm to 1 to obtain the logarithmic adjustment term; The standard deviation of the voltage vector magnitude is multiplied by the logarithmic adjustment term to obtain the power grid state non-stationarity index.

5. The power regulation method for a distributed photovoltaic power station cluster according to claim 4, characterized in that, The forgetting factor in the impedance identification process is dynamically adjusted based on the power grid state non-stationarity index, including: The attenuation coefficient is calculated by multiplying the power grid non-stability index by a preset attenuation adjustment coefficient and taking the negative value, which is then used as the exponent term of the natural index. Calculate the difference between the upper limit and the lower limit of the forgetting factor, and multiply the difference by the decay coefficient to obtain the dynamic increment; The lower limit of the forgetting factor is added to the dynamic increment to obtain the adjusted forgetting factor.

6. The power regulation method for a distributed photovoltaic power station cluster according to claim 1, characterized in that, The evaluation of the dynamic stiffness coefficient of the power grid based on the equivalent impedance modulus of the power grid and the rate of frequency change includes: The equivalent impedance modulus of the power grid obtained in real time is added to the preset non-zero protection constant to obtain the corrected impedance value; the ratio of the effective value of the rated voltage at the grid connection point to the corrected impedance value is calculated to obtain the reference stiffness term. The frequency attenuation coefficient is calculated by multiplying the preset frequency sensitivity coefficient by the frequency change rate and taking the negative value, which is used as the exponential term of the natural index. Multiplying the reference stiffness term by the frequency attenuation coefficient yields the current dynamic stiffness coefficient of the power grid.

7. The power regulation method for a distributed photovoltaic power station cluster according to claim 6, characterized in that, The step of predicting the smoothing weights of the controller based on the adaptive mapping model of the power grid dynamic stiffness coefficient includes: Calculate the ratio of the current power grid dynamic stiffness coefficient to the reference stiffness benchmark value, and calculate the hyperbolic tangent of the ratio; Subtract the hyperbolic tangent value from 1 to obtain the weight adjustment coefficient; calculate the difference between the maximum and minimum values ​​of the preset smoothing weight adjustment range, and multiply the difference by the weight adjustment coefficient to obtain the weight increment; The minimum value of the preset smoothing weight adjustment range is added to the weight increment to obtain the smoothing weight updated in the cost function of the model prediction controller.

8. The power regulation method for a distributed photovoltaic power station cluster according to claim 7, characterized in that, The mapping relationship of the damping weight of the predictive controller according to the adaptive mapping model of the power grid dynamic stiffness coefficient is as follows: when the power grid dynamic stiffness coefficient increases, the damping weight is reduced through nonlinear mapping; when the power grid dynamic stiffness coefficient decreases, the damping weight is increased through nonlinear mapping.

9. The power regulation method for a distributed photovoltaic power station cluster according to claim 1, characterized in that, The drive inverter performs power regulation, including sending the optimal voltage vector obtained by rolling optimization to the pulse width modulation module to drive the power switching transistors of the photovoltaic inverter to operate.

10. A power regulation system for a distributed photovoltaic power station cluster, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the power regulation method for a distributed photovoltaic power station cluster according to any one of claims 1-9.

Images