Power distribution network voltage cooperative control system for high-proportion new energy

Abstract

The invention relates to the technical field of voltage cooperative control, in particular to a power distribution network voltage cooperative control system for high-proportion new energy. According to the system, an effective fluctuation window with enough excitation intensity is screened by synchronously collecting photovoltaic grid-connected point voltage and active power data, and the voltage response sensitivity is estimated in real time based on a first-order difference linear model and an online identification algorithm; on the basis, electric energy unit price, resource availability and model fitting quality are fused, unit voltage regulation reference cost, a resource exhaustion penalty coefficient and an identification residual penalty coefficient are constructed respectively, and voltage regulation comprehensive cost comprehensively reflecting economical efficiency, sustainability and sensing reliability is generated; and finally, dynamic decision making is carried out according to the comprehensive cost, so that the voltage safety is guaranteed, the adjustment efficiency and the operation cost are effectively balanced, and excessive wear of equipment and new energy waste are avoided.

Description

Voltage Coordination Control System for Distribution Networks with High Proportion of New Energy Sources Technical Field

[0001] This invention relates to the field of voltage coordination control technology, and more specifically to a voltage coordination control system for distribution networks with a high proportion of new energy sources. Background Technology

[0002] In distribution networks with high penetration of distributed photovoltaic (PV) power, feeder voltage exceeding limits is of paramount importance. Traditional voltage control methods primarily include local voltage-watt droop control of the PV inverter and on-load tap regulation of the transformer. In the weakly connected region (high impedance) at the end of long-distance feeders, adjusting a small amount of active power by the inverter can effectively improve the voltage. However, in the strongly connected region (low impedance) near the transformer, the voltage is extremely insensitive to active power. If a uniform droop parameter is used, inverters located at strongly connected nodes often need to reduce a large amount of active power to eliminate even small voltage deviations, leading to severe curtailment losses.

[0003] Existing coordinated control strategies typically trigger transformer operation solely based on voltage amplitude, lacking a quantitative assessment of the "cost" of local regulation resources. This leads to premature transformer activation in minor over-limit scenarios that could be resolved through local fine-tuning, increasing mechanical wear; or, in scenarios where local regulation is already extremely uneconomical (such as deep curtailment), the system still forces the inverter to operate, delaying the activation of the transformer, resulting in system-wide energy waste. In summary, existing coordinated control strategies fail to balance the inherent contradiction between "regulation efficiency" and "economic cost," resulting in poor decision-making effectiveness for voltage coordinated control. Summary of the Invention

[0004] To address the technical problem of poor decision-making performance in voltage-coordinated control, this invention provides a voltage-coordinated control system for distribution networks with a high proportion of renewable energy. The specific technical solution is as follows: This invention proposes a voltage-coordinated control system for distribution networks with a high proportion of renewable energy. The system includes: an acquisition module for synchronously acquiring the voltage value of the distributed photovoltaic grid-connected point and the active power output of the inverter at each sampling moment within a sliding time window; a response sensitivity analysis module for filtering out effective fluctuation windows based on the fluctuations in photovoltaic active power within the sliding time window; and within the effective fluctuation window, performing linear analysis based on the first-order difference between the voltage value and active power at adjacent sampling moments, combined with... The line identification algorithm determines the voltage response sensitivity; the penalty analysis module obtains the unit price of electricity within the effective fluctuation window, and determines the unit voltage regulation benchmark cost by combining the unit price of electricity and the voltage response sensitivity; it determines the resource depletion penalty coefficient by combining the active power, rated power, and maximum historical power under the same illumination conditions of the inverter; it performs residual discreteness analysis on the voltage value and active power at each sampling time by combining the voltage response sensitivity, and determines the identification residual penalty coefficient; the voltage control module determines the comprehensive voltage regulation cost by combining the unit voltage regulation benchmark cost, the resource depletion penalty coefficient, and the identification residual penalty coefficient, and performs voltage control at the grid connection point based on the comprehensive voltage regulation cost.

[0005] Furthermore, the step of selecting an effective fluctuation window based on the fluctuation of photovoltaic active power under the sliding time window includes: calculating the range of active power under the sliding time window, and when the range is greater than a preset amplitude, taking the corresponding sliding time window as an effective fluctuation window.

[0006] Furthermore, the online identification algorithm is the RLS algorithm, which performs linear analysis based on the first-order difference between voltage values ​​and active power at adjacent sampling times. The voltage response sensitivity is determined by combining this with the online identification algorithm. This includes: the RLS algorithm performing a complete recursive parameter estimation calculation within the effective fluctuation window to obtain the voltage response sensitivity. This process includes: using zero as the initial value of the parameter estimate, and using a preset matrix value as the initial value of the covariance matrix, wherein the preset matrix value is taken as... The algorithm iterates through each point within the effective fluctuation window, calculates the prediction residual based on the linear changes in the current voltage and active power, and updates the parameter estimates and covariance matrix using a gain vector containing a regularization term. After the iteration is complete, the final parameter estimates are output as the voltage response sensitivity.

[0007] Furthermore, the regularization term is a preset minimum positive number to prevent gain calculation divergence when the change in active power approaches zero; a forgetting factor less than 1 is introduced during the covariance matrix update process to assign decreasing weights to historical data in chronological order; wherein, the preset minimum positive number is a value of The forgetting factor is 0.98.

[0008] Furthermore, the step of combining the unit price of electricity and the voltage response sensitivity to determine the unit voltage regulation benchmark cost includes: taking the absolute value of the voltage response sensitivity and determining the maximum value between the absolute value and the preset minimum constraint value as the sensitivity analysis value; wherein the preset minimum constraint value is greater than 0; using the sensitivity analysis value as the denominator and the unit price of electricity as the numerator, the unit voltage regulation benchmark cost is calculated.

[0009] Furthermore, determining the resource depletion penalty coefficient by combining the inverter's active power, rated power, and maximum historical power under the same illumination conditions includes: calculating the difference between the preset maximum historical power and the active power, and obtaining the active power abandonment rate by dividing the difference by the rated power. In the formula, This represents the ratio of the difference to the rated power. This indicates the active power curtailment rate. This represents the function that takes the maximum value. This represents the minimum value function; based on the active power curtailment rate and the preset curtailment resource penalty term, the resource depletion penalty coefficient is determined.

[0010] Furthermore, the residual discrepancy analysis of the voltage value and active power at each sampling moment, combined with voltage response sensitivity, to determine the identification residual penalty coefficient includes: calculating the first-order difference of the voltage value of the sample at each sampling moment within the effective fluctuation window as the voltage difference; calculating the first-order difference of the active power of the sample at each sampling moment within the effective fluctuation window as the power difference; calculating the product of the power difference and the voltage response sensitivity, and using the difference between the voltage difference and the product as the time fitting error of the sample at the sampling moment; averaging the squared values ​​of the time fitting errors of all samples at all sampling moments as the model fitting error variance; and performing a penalty analysis by combining the model fitting error variance, a preset noise sensitivity coefficient, and a preset benchmark model fitting variance constant to determine the identification residual penalty coefficient, wherein the benchmark model fitting variance constant is the allowable noise level of the model fitting error variance under typical stable operating conditions.

[0011] Furthermore, the method for determining the identification residual penalty coefficient includes: the identification residual penalty coefficient is determined based on the ratio of the model fitting error variance to the baseline model fitting variance constant, specifically: taking 1 as the base value, an increment proportional to the ratio is superimposed, and the proportionality coefficient of the increment is a preset noise sensitivity coefficient.

[0012] Furthermore, the determination of the comprehensive voltage regulation cost by combining the unit voltage regulation benchmark cost, the resource depletion penalty coefficient, and the identification residual penalty coefficient includes: performing a weighted average of the resource depletion penalty coefficient and the identification residual penalty coefficient to obtain a total penalty index, wherein the sum of the weights of the resource depletion penalty coefficient and the identification residual penalty coefficient is 1; and taking the product of the total penalty index and the unit voltage regulation benchmark cost as the comprehensive voltage regulation cost.

[0013] Furthermore, the voltage control of the grid connection point based on the comprehensive voltage regulation cost includes: when the comprehensive voltage regulation cost of the grid connection point is less than a preset cost threshold, voltage regulation is performed by reducing the active power of the local inverter; otherwise, a request is sent to the distribution network master station to trigger the adjustment of the tap position of the on-load tap changer.

[0014] This invention offers the following advantages: By constructing a multi-dimensional cost evaluation system comprising a unit voltage regulation benchmark cost, a resource depletion penalty coefficient, and an identification residual penalty coefficient, and generating a comprehensive voltage regulation cost accordingly, this invention effectively solves the decision-making imbalance problem caused by the lack of quantitative evaluation of the "economic efficiency" and "reliability" of local regulation resources in existing collaborative control strategies. Within the effective fluctuation window, the system calculates the unit voltage regulation cost based on the voltage response sensitivity obtained through online identification, combined with real-time electricity unit price, thus accurately reflecting the marginal economic cost of inverter voltage regulation under different grid intensities. Simultaneously, the introduction of a resource depletion penalty coefficient quantifies the current opportunity cost of curtailment, avoiding blind reliance on local regulation in scenarios of deep curtailment. Furthermore, the identification residual penalty coefficient characterizes the model's credibility, preventing misjudgments caused by noise interference or weak excitation. The comprehensive cost formed by the fusion of these three factors serves as the basis for control decisions, enabling the system to perform dynamic voltage control at the grid connection point. Therefore, this invention achieves a dynamic balance between regulation efficiency and economic cost, significantly improving the precision and overall operational efficiency of voltage collaborative control in high-proportion renewable energy distribution networks. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 is a structural diagram of a distribution network voltage collaborative control system for high-proportion new energy sources provided by an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a voltage coordination control system for distribution networks oriented towards high-proportion renewable energy, proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] In distribution networks with high penetration of distributed photovoltaic (PV) power, feeder voltage exceeding limits is of paramount importance. Traditional voltage control methods primarily include local voltage-watt droop control of the PV inverter and on-load tap regulation of the transformer. In the weakly connected region (high impedance) at the end of long-distance feeders, adjusting a small amount of active power by the inverter can effectively improve the voltage. However, in the strongly connected region (low impedance) near the transformer, the voltage is extremely insensitive to active power. If a uniform droop parameter is used, inverters located at strongly connected nodes often need to reduce a large amount of active power to eliminate even small voltage deviations, leading to severe curtailment losses.

[0020] Existing coordinated control strategies typically trigger transformer operation solely based on voltage amplitude, lacking a quantitative assessment of the "cost" of local regulation resources. This leads to premature transformer activation in minor over-limit scenarios that could be resolved through local fine-tuning, increasing mechanical wear; or, in scenarios where local regulation is already extremely uneconomical (such as deep curtailment), the system still forces the inverter to operate, delaying the activation of the transformer, resulting in system-wide energy waste. In summary, existing coordinated control strategies fail to balance the inherent contradiction between "regulation efficiency" and "economic cost," resulting in poor decision-making effectiveness for voltage coordinated control.

[0021] To address the aforementioned technical problems, the following detailed description, in conjunction with the accompanying drawings, illustrates a specific solution for a distribution network voltage collaborative control system oriented towards high-proportion renewable energy, provided by this invention.

[0022] Please refer to Figure 1, which shows a structural diagram of a distribution network voltage collaborative control system for high-proportion renewable energy provided by an embodiment of the present invention. The system includes: an acquisition module 101, a response sensitivity analysis module 102, a penalty analysis module 103, and a voltage control module 104.

[0023] The acquisition module 101 is used to synchronously acquire the voltage value of the distributed photovoltaic grid-connected point and the active power output of the inverter at each sampling moment of the sliding time window.

[0024] In this embodiment of the invention, a space with a length of [length missing] can be allocated in the memory of the inverter controller. (The embodiments of the present invention take) The first-in-first-out (FIFO) circular data buffer serves as a sliding time window, and the controller uses the sampling frequency of the real-time control cycle (e.g., The voltage value v(t) (in volts) at the grid connection point of the distributed photovoltaic system and the active power p(t) (in kilowatts) of the inverter output are sampled, and the two data are combined and written into the buffer, where t is the time index of the real-time control domain.

[0025] The response sensitivity analysis module 102 is used to filter out the effective fluctuation window based on the fluctuation of photovoltaic active power under the sliding time window; within the effective fluctuation window, linear analysis is performed based on the first-order difference between the voltage value and active power at adjacent sampling times, and the voltage response sensitivity is determined by combining the online identification algorithm.

[0026] By synchronously acquiring voltage and active power data of distributed photovoltaic grid-connected points within a sliding time window, the system already has the foundation for real-time perception of local operating status. On this basis, it further filters out effective fluctuation windows that meet preset amplitudes based on the fluctuation range of photovoltaic active power, aiming to identify power disturbance events with sufficient excitation intensity caused by changes in natural illumination.

[0027] Because the electrical strength (short-circuit capacity) of distribution network feeders exhibits time-varying characteristics with changes in network topology and operating mode, and the voltage at the grid connection point is constantly disturbed by random switching of background loads, it is impossible to effectively distinguish whether voltage changes are caused by local power injection or by fluctuations in the background network by directly comparing steady-state operating data.

[0028] Furthermore, in some embodiments of the present invention, the effective fluctuation window is selected based on the fluctuation of photovoltaic active power under the sliding time window, including: calculating the range of active power under the sliding time window, and when the range is greater than a preset amplitude, the corresponding sliding time window is taken as the effective fluctuation window.

[0029] The preset amplitude can be a threshold value for active power fluctuations. In this embodiment of the invention, it can be set to the rated power of the inverter. This serves as a preset amplitude. Therefore, it allows for adaptive threshold setting based on different inverters, avoiding the poor robustness of a single threshold.

[0030] Understandably, only when power changes are sufficiently significant can the voltage response signal possess adequate signal-to-noise ratio and identifiability, thus providing a high-quality data foundation for subsequent online estimation of grid voltage response sensitivity. Forcing identification under minor fluctuations or stable operating conditions is highly susceptible to measurement noise or external interference, leading to distorted parameter estimations. Therefore, this effective fluctuation window filtering mechanism not only avoids ineffective calculations but also significantly improves the reliability and robustness of voltage sensitivity identification, laying the groundwork for accurate and economical voltage coordination control in high-proportion renewable energy scenarios.

[0031] After successfully selecting an effective fluctuation window with sufficient excitation strength, the system has obtained a set of voltage-power dynamic response data with high signal-to-noise ratio caused by natural power disturbances. At this point, an online identification algorithm is introduced to linearly model the first-order difference component of adjacent sampling times within the window, aiming to estimate the voltage response sensitivity characterizing the local power grid strength in real time.

[0032] The online identification algorithm is the Recursive Least Squares (RLS) algorithm.

[0033] Furthermore, in some embodiments of the present invention, linear analysis is performed based on the first-order difference between voltage values ​​and active power at adjacent sampling times, and the voltage response sensitivity is determined by combining an online identification algorithm. This includes: the RLS algorithm performs a complete recursive parameter estimation calculation within the effective fluctuation window to obtain the voltage response sensitivity. This process includes: using zero as the initial value of the parameter estimate, and using a preset matrix value as the initial value of the covariance matrix, wherein the preset matrix value is taken as... The algorithm iterates through each point within the effective fluctuation window, calculates the prediction residual based on the linear changes in the current voltage and active power, and updates the parameter estimates and covariance matrix using a gain vector containing a regularization term. After the iteration is complete, the final parameter estimates are output as the voltage response sensitivity.

[0034] Furthermore, in some embodiments of the present invention, the regularization term is a preset minimum positive number to prevent gain calculation divergence when the change in active power approaches zero; a forgetting factor less than 1 is introduced during the covariance matrix update process to assign decreasing weights to historical data in chronological order; wherein, the preset minimum positive number is a value of The forgetting factor is 0.98.

[0035] It is understandable that, for the effective fluctuation window locked by the k-th event, the sample index within the effective fluctuation window is defined as... ( The controller first constructs the difference sequence of adjacent samples. as well as .

[0036] Based on the local linearization model of the distribution network, the observation equations are established as follows: . This represents the fitting residuals when performing linear fitting. Voltage response sensitivity (unit: ).

[0037] Analysis based on the RLS algorithm: setting parameters and estimating initial values. Preset matrix values As the initial value of the covariance matrix, it is a 1×1 matrix and can be regarded as a scalar for subsequent calculations. The embodiment of the present invention establishes a single-input single-output (SISO) model of voltage change versus active power change, therefore the covariance matrix is ​​preset to degenerate into a scalar value and directly participate in algebraic operations.

[0038] Then, to from to Perform iterative analysis: First step, calculate the prediction error. The second step is to calculate the gain vector. To prevent due to Minimal values ​​can lead to numerical instability or zero denominators. Introducing a minimal regularization parameter into the denominator can address this issue. (This embodiment takes) ): In the formula, This represents the forgetting factor, with a value of 0.98. This represents a preset minimum positive number, with a value of [value missing]. The third step is to update the parameter estimates. : Fourth step, update the covariance matrix. : .

[0039] The above four steps form a logical loop, and when the loop ends ( When this is done, the final parameter estimates are obtained. The final parameter estimates As voltage response sensitivity .

[0040] It should be noted that the forgetting factor The introduction of this factor allows the algorithm to assign exponentially decaying weights to historical observation data in chronological order during parameter estimation; that is, the weight of the most recent data is 1, the weight of the data from the previous time step is 1, and the weight of the data from the previous time step is 1. The moment before that was And so on. This weighting mechanism enables the identification results to effectively track the slow time-varying voltage response characteristics of the distribution network caused by load fluctuations or topology changes, while maintaining sufficient estimation stability.

[0041] Although the parameter identification process adopted in this embodiment of the invention is based on the basic framework of Recursive Least Squares (RLS), it has been specifically improved in key steps for application scenarios with a high proportion of new energy distribution networks, which is significantly different from the standard RLS algorithm. First, the algorithm only performs one complete traversal iteration within the effective fluctuation window (n=1 to n=2). Instead of the continuous online recursion of traditional RLS, it avoids invalid updates during periods without excitation or dominated by noise, thus improving the physical rationality of the identification results.

[0042] Secondly, a minimum regularization parameter is explicitly introduced in the gain vector calculation ( This effectively prevents ill-conditioned denominators or numerical divergence when Δp(n) approaches zero due to the gradual increase in photovoltaic power output, thus enhancing the robustness of the algorithm. The preset matrix value is set to a larger value ( The initial parameters are given high uncertainty, giving the algorithm strong learning ability at the beginning of the window; at the same time, the forgetting factor... The value is fixed at 0.98, and the covariance update formula is embedded. This allows for exponential decay weighting of historical data within a window in chronological order, making recent disturbances have a greater impact on the final estimate and thus better tracking the slow time-varying characteristics of the power grid.

[0043] The synergistic effect of these measures enables a stable output of voltage response sensitivity with clear physical meaning even under finite, non-stationary, and noisy natural power disturbances. This provides a reliable basis for subsequent economic voltage control and overcomes the technical defects of traditional RLS when directly applied to weakly excited distribution network scenarios, such as easy instability, slow convergence, or large estimation deviation.

[0044] It should be further noted that in some other embodiments of the present invention, after obtaining... Subsequently, to prevent erroneous updates to model parameters in the absence of sufficient excitation (i.e., minimal power changes), the system can perform a post-hoc verification before outputting results. The controller calculates the variance of the first-order difference values ​​of all active power within the effective fluctuation window. And set a preset minimum excitation variance threshold. Among them, the minimum incentive variance threshold For example, Its dimensions are the same as same.

[0045] exist If this occurs, it indicates that although the captured event meets the range condition, the process is too gradual to stimulate a dynamic response in the power grid, and the identification result is unreliable. In this case, the system discards the calculation result and keeps the original parameters unchanged; if... The system determines that the identification is valid. The system will then use the estimated parameter values. The output is the voltage response sensitivity for this event. .

[0046] The penalty analysis module 103 is used to obtain the unit price of electricity within the effective fluctuation window, and determine the unit voltage regulation benchmark cost by combining the unit price of electricity and voltage response sensitivity; determine the resource depletion penalty coefficient by combining the active power, rated power and maximum historical power under the same illumination conditions of the inverter; and determine the identification residual penalty coefficient by combining the voltage value and active power at each sampling time with voltage response sensitivity.

[0047] The technical focus of related technologies is usually on how to force voltage to remain within the standard range through physical regulation, but it often overlooks the significant differences in the economic cost and equipment loss risk of achieving a unit voltage improvement under different grid strengths and equipment operating conditions. Indiscriminate regulation can lead to under-regulation at weakly connected nodes or over-regulation at strongly connected nodes.

[0048] In distribution networks with a high proportion of renewable energy, relying solely on physical models or fixed thresholds for voltage regulation makes it difficult to balance economic efficiency and control effectiveness. On the one hand, the voltage changes resulting from the same power regulation vary greatly under different grid intensities, making regulation costs incomparable. On the other hand, blindly reducing photovoltaic output may cause unnecessary curtailment losses, and if the identification results are affected by noise, they may mislead control decisions.

[0049] To this end, this step constructs three key cost factors simultaneously based on real-time operating data within the effective fluctuation window: calculating the benchmark cost per unit voltage regulation based on the unit price of electricity and voltage response sensitivity to quantify the direct economic loss of voltage regulation behavior; assessing the degree of resource depletion by combining current active power output and historical maximum available power to form a resource depletion penalty coefficient that reflects the opportunity cost of curtailment; and generating an identification residual penalty coefficient that characterizes the credibility of the model by analyzing the dispersion of the voltage-power response residual.

[0050] In this embodiment of the invention, discrete power grid physical characteristics (response characteristics), real-time resource depletion status (curtailment rate), and environmental observation uncertainties (fitting error) are mapped to a unified economic value dimension, which facilitates subsequent analysis.

[0051] Furthermore, in some embodiments of the present invention, the unit voltage regulation benchmark cost is determined by combining the unit price of electricity and the voltage response sensitivity, including: taking the absolute value of the voltage response sensitivity and determining the maximum value of the absolute value and the preset minimum constraint value as the sensitivity analysis value; wherein the preset minimum constraint value is greater than 0; using the sensitivity analysis value as the denominator and the unit price of electricity as the numerator, the unit voltage regulation benchmark cost is calculated.

[0052] Due to differences in electrical strength at different nodes in a distribution network, the voltage improvement effect resulting from the same amount of active power reduction can vary drastically. At nodes with high electrical strength (i.e., small response eigenvalues), reducing even small voltage deviations often requires cutting a significant amount of active power, making this "high-cost" regulation economically unreasonable. Therefore, this step utilizes the inverse relationship between real-time electricity prices and apparent response eigenvalues ​​to define the basic economic cost required to improve a unit voltage under ideal linear conditions.

[0053] This means directly calculating the ratio of the unit price of electricity to the sensitivity analysis value, which serves as the benchmark cost per unit voltage regulation. In areas with strong power grids (such as near substations), the sensitivity of voltage to changes in active power is... The voltage is extremely small, meaning that even a significant reduction in photovoltaic output would only cause a slight voltage drop, resulting in extremely low voltage regulation efficiency and huge economic losses. Without constraints, the control system might misjudge this scenario as "low-cost voltage regulation," leading to unnecessary curtailment of solar power. Therefore, this invention takes the absolute value of the identified voltage response sensitivity to ensure that the cost calculation depends only on the response intensity and is not affected by the sign, and further introduces a preset minimum constraint value F (such as...). V / kW), through As a denominator, it effectively prevents... When the voltage approaches zero, the unit voltage regulation cost tends to infinity, potentially causing numerical overflow or control logic collapse. The calculated unit voltage regulation baseline cost, expressed in "yuan / (V·h)," intuitively quantifies the hourly basic economic loss corresponding to each 1-volt voltage reduction when only considering the constraints of the power grid's physical architecture. The ultimate goal is to achieve the overall coordinated control objective of "maintaining voltage safety at the lowest economic cost."

[0054] By combining the inverter's active power, rated power, and maximum historical power under the same illumination conditions, the resource depletion penalty coefficient is determined. This includes: calculating the difference between the preset maximum historical power and the active power; and then using the ratio of this difference to the rated power to obtain the active power abandonment rate. In the formula, This represents the ratio of the difference to the rated power. This indicates the active power curtailment rate. This represents the function that takes the maximum value. This represents the minimum value function; based on the active power curtailment rate and the preset curtailment resource penalty term, the resource depletion penalty coefficient is determined.

[0055] The preset maximum historical power can be obtained by searching the historical database. This maximum historical power can be similar to the scene at the corresponding sampling time, such as similar lighting conditions and other similar scene features, to improve the rationality and enable subsequent analysis. The specific data retrieval and analysis process is existing technology and is not within the protection scope of this invention, so it will not be described in detail.

[0056] In some embodiments of the present invention, the resource depletion penalty coefficient is determined based on the active power curtailment rate and a preset curtailment resource penalty term, and the corresponding formula is as follows: In the formula, This represents the preset light-wasting resource penalty term, which is dimensionless data, and the value of the light-wasting resource penalty term is... This means that when the inverter is in a state of full solar power curtailment ( When this happens, the adjustment cost will be amplified to five times the base value, thus logically forcing the system to tend to retain the remaining power generation capacity.

[0057] Of course, in other embodiments of the present invention, other calculation or analysis methods can also be used, such that the larger the active power curtailment rate and the preset curtailment resource penalty item value, the larger the resource depletion penalty coefficient value.

[0058] Further, in some embodiments of the present invention, combined with voltage response sensitivity, residual discreteness analysis is performed on the voltage value and active power at each sampling time to determine the identification residual penalty coefficient, including: calculating the first-order difference value of the voltage value of the sample at each sampling time within the effective fluctuation window as the voltage difference; calculating the first-order difference value of the active power of the sample at each sampling time within the effective fluctuation window as the power difference; calculating the product of the power difference and the voltage response sensitivity, and using the difference between the voltage difference and the product as the time fitting error of the sample at the sampling time; averaging the squared values ​​of the time fitting errors of all samples at all sampling times as the model fitting error variance; and combining the model fitting error variance, a preset noise sensitivity coefficient, and a preset benchmark model fitting variance constant to perform penalty analysis and determine the identification residual penalty coefficient, wherein the benchmark model fitting variance constant is the allowable noise level of the model fitting error variance under typical stable operating conditions, in units of 1000 kJ / m². (Voltage squared), in its physical sense, is "the permissible level of background voltage noise fluctuation under typical operating conditions".

[0059] In distribution networks with a high proportion of distributed photovoltaic (PV) grid integration, online identification of voltage response sensitivity is susceptible to multiple disturbances, such as random load fluctuations, multi-source coordinated output, communication noise, or nonlinear equipment dynamics. These disturbances lead to a non-negligible deviation between the established local linear model and actual observation data. Directly using such low-reliability identification results for subsequent control decisions may result in incorrect voltage regulation, excessive curtailment of solar power, or even voltage oscillations. To address this, this invention introduces an identification reliability assessment mechanism based on residual discreteness analysis.

[0060] Specifically, within the effective fluctuation window, the first-order difference between voltage and active power is first calculated to obtain the dynamic change; then, the voltage response is predicted using the identified voltage response sensitivity, and the difference between this and the measured voltage difference is calculated to obtain the fitting error at each moment. Finally, the mean of the squares of all errors is calculated to form the model fitting error variance, and the corresponding calculation formula is as follows: In the formula, This represents the variance of the model fitting error for the k-th effective fluctuation window. This represents the voltage difference of the nth sample within the effective fluctuation window. This represents the power difference of the nth sample within the effective fluctuation window. This represents the voltage response sensitivity of the k-th effective fluctuation window. This represents the time fitting error corresponding to the sampling time of the nth sample.

[0061] The variance of the model fitting error intuitively characterizes the reliability analysis of the linear model for the fluctuations in the k-th effective fluctuation window. The smaller the value, the more reliable the model is, indicating that the voltage change is mainly caused by photovoltaic power disturbances. The larger the value, the more significant the unmodeled dynamics or strong disturbances, and the identification results are questionable.

[0062] To ensure that subsequent control modules can acquire valid grid characteristic parameters at any time, in some embodiments of this invention, a parameter holding, initialization, and backup mechanism can be established. The controller sets up a set of non-volatile global shared registers to store voltage response sensitivity. and model fitting error variance Once the identification calculation is completed and passes verification, the new [process] will be immediately [implemented]. and Write to this register. The value in the register remains unchanged (Zero-Order Hold) until the next valid event update.

[0063] Specifically, to address the "parameter aging" or "logic deadlock" issues caused by prolonged periods of stable control, this system incorporates a model validity timer. If the system fails to detect any valid power fluctuations within a preset timeframe (1 hour in this embodiment), the controller determines that it is currently in an ultra-stable operating condition. At this point, the controller executes a safety net: forcibly resetting the model fitting error variance in the register to a minimum value. (or minimum value). This is based on the reasonable inference that "extremely stable operating conditions mean minimal environmental noise interference", to prevent the uncertainty penalty from being mistakenly increased due to long-term lack of excitation, and to keep the voltage response sensitivity in the register unchanged from the previous effective value.

[0064] For situations where no events are detected during the initial power-on startup phase, the controller presets the voltage response sensitivity in the register to a pre-defined default reference value for a strong power grid during the initialization phase. (This embodiment takes) ), and preset the model fitting error variance to . This default value eliminates the parameter vacuum during the cold start phase, ensuring the continuity of the calculation process.

[0065] By combining the model fitting error variance, the preset noise sensitivity coefficient, and the preset baseline model fitting variance constant, a penalty analysis is performed to determine the identification residual penalty coefficient. The method for determining the identification residual penalty coefficient includes: the identification residual penalty coefficient is determined based on the ratio of the model fitting error variance to the baseline model fitting variance constant, specifically: using 1 as the base value, an increment proportional to the ratio is superimposed, and the proportionality coefficient of this increment is the preset noise sensitivity coefficient.

[0066] Wherein, the reference model fitting variance constant is the allowable noise level of the model fitting error variance under typical steady operating conditions. Optionally, the reference model fitting variance constant is taken as 0.05 in this embodiment, and the noise sensitivity coefficient is taken as 1.0 in this embodiment.

[0067] In the formula, This represents the identification residual penalty coefficient for the k-th effective fluctuation window. The noise sensitivity coefficient is a dimensionless data point. This represents the variance of the model fitting error for the k-th effective fluctuation window. This represents the variance constant of the baseline model fit, and... They have the same dimensions.

[0068] The identification residual penalty coefficient increases with the severity of model mismatch, which is essentially a penalty for "using unreliable parameters ( The system imposes an economic penalty on the decision-making risk associated with "larger numerical values". Through this mechanism, the system can automatically suppress the influence of low-quality identification results in the overall cost, ensuring that the voltage regulation strategy is always based on high-confidence grid perception, thereby improving the robustness, safety and economy of overall collaborative control, and effectively supporting the autonomous voltage stability operation in high-penetration new energy scenarios.

[0069] This step, through the collaborative construction of multi-dimensional cost factors, maps the physical characteristics of the power grid, market electricity price signals, and model uncertainties into comparable economic indicators, providing a unified decision-making basis for the subsequent realization of refined and adaptive voltage coordinated control.

[0070] The voltage control module 104 is used to determine the comprehensive voltage regulation cost by combining the unit voltage regulation reference cost, the resource depletion penalty coefficient and the identification residual penalty coefficient, and to perform voltage control at the grid connection point based on the comprehensive voltage regulation cost.

[0071] In distribution networks with a high proportion of distributed photovoltaic (PV) grid integration, a single-dimensional control criterion is insufficient to balance voltage safety, economic operation, and model reliability: relying solely on voltage limit exceedances can easily lead to excessive curtailment of solar power or frequent equipment operation; while ignoring the identification of uncertainties or resource status may result in costly or even harmful regulatory decisions based on erroneous perceptions.

[0072] To this end, this invention organically integrates the unit voltage regulation benchmark cost, resource depletion penalty coefficient, and identification residual penalty coefficient to construct a unified comprehensive voltage regulation cost index. This comprehensive cost is not a simple summation, but rather a multi-factor collaborative quantification of "the overall economic and operational risks arising from local voltage regulation under the current grid conditions, electricity price levels, renewable energy availability, and model credibility." Based on this comprehensive cost, grid connection point voltage control can minimize unnecessary curtailment losses while ensuring voltage safety and avoid malfunctions caused by low-quality identification, thereby improving the economy, robustness, and autonomous decision-making capability of the entire distributed voltage control system.

[0073] Furthermore, in some embodiments of the present invention, the comprehensive voltage regulation cost is determined by combining the unit voltage regulation benchmark cost, the resource depletion penalty coefficient, and the identification residual penalty coefficient, including: performing a weighted average of the resource depletion penalty coefficient and the identification residual penalty coefficient to obtain a total penalty index, wherein the sum of the weights of the resource depletion penalty coefficient and the identification residual penalty coefficient is 1; and using the product of the total penalty index and the unit voltage regulation benchmark cost as the comprehensive voltage regulation cost.

[0074] It should be noted that this module aims to construct a nonlinear comprehensive voltage regulation cost by combining the unit voltage regulation benchmark cost, resource depletion penalty coefficient, and identification residual penalty coefficient. This comprehensive voltage regulation cost serves as the sole quantitative basis for system decision-making and can automatically identify ineffective regulation situations with low input-output ratios.

[0075] In this embodiment of the invention, the weight of the resource depletion penalty coefficient can be 0.6, and the weight of the identification residual penalty coefficient can be 0.4. Then, a weighted average is performed to obtain the total penalty index. The product of the total penalty index and the unit voltage regulation baseline cost is used as the comprehensive voltage regulation cost. When the risk of curtailment is high or the identification result is unreliable, the comprehensive voltage regulation cost is significantly amplified, thereby suppressing inefficient or high-risk regulation behavior; conversely, it allows for a more aggressive local response.

[0076] Furthermore, in some embodiments of the present invention, voltage control at the grid connection point is performed based on the comprehensive voltage regulation cost, including: when the comprehensive voltage regulation cost at the grid connection point is less than a preset cost threshold, voltage regulation is performed by reducing the active power of the local inverter; otherwise, a request is sent to the distribution network master station to trigger the adjustment of the tap position of the on-load tap changer.

[0077] The preset cost threshold is a threshold value for the comprehensive cost of voltage regulation. In this embodiment of the invention, the preset cost threshold can be specifically calculated based on the transformer regulation cost. Specifically, it can be calculated by multiplying the current voltage over-limit deviation (volts) by the estimated over-limit duration (hours) to obtain the influencing index. Then, the average cost of a single action of the transformer tap changer mechanism (the ratio of the total value of equipment value and overhaul cost to the rated number of regulation times under the total lifespan) is calculated to the ratio of the influencing index to obtain the preset cost threshold.

[0078] The estimated duration of the voltage exceedance (in hours) can be specifically set, for example, to be 1 hour, based on historical experience. The current voltage exceedance deviation is specifically the absolute value of the difference between the voltage at the grid connection point at the corresponding sampling time and the preset upper voltage threshold, i.e. , This indicates the current grid connection point voltage. This indicates the target reference voltage (e.g., 400V, which may be adjusted depending on the actual scenario).

[0079] It should be noted that in distribution networks with high photovoltaic penetration, grid connection voltage exceedances can typically be mitigated through two methods: one is local inverters reducing active power output (fast and flexible but resulting in curtailment losses), and the other is remotely regulating on-load tap changers (OLTCs, which have a wide impact range, no power generation losses, but slow response and limited lifespan). Indiscriminately prioritizing either method will lead to poor system reliability. For example, in areas with strong grids (where voltage is insensitive to power changes), forced peak shaving requires significant curtailment to fine-tune the voltage, resulting in extremely poor economic efficiency; while in areas with weak grids, relying on OLTCs leads to delayed response and may cause persistent voltage exceedances.

[0080] To this end, this invention introduces the comprehensive cost of voltage regulation as a decision-making basis: when the cost is lower than a preset cost threshold, it indicates that the "unit voltage improvement cost" of voltage regulation by local inverters is low enough (i.e., high grid sensitivity, low electricity price, low curtailment of solar power, and reliable model). In this case, the active power reduction of the inverter is prioritized to achieve fast, accurate, and low-cost local regulation. When the comprehensive cost of voltage regulation is higher than the preset cost threshold, it indicates that local voltage regulation is no longer economical or reliable (e.g., weak incentives, high electricity price, risk of deep curtailment of solar power, or identification distortion). The system then actively sends a coordination request to the distribution network master station to trigger OLTC level adjustment and restore the voltage level globally, avoiding unnecessary waste of renewable energy and equipment losses.

[0081] Specifically, voltage regulation is achieved through active power reduction of the local inverter, including: the controller adjusting the voltage based on the current voltage exceedance and the identified voltage response sensitivity. (Unit: V / kW), to calculate the amount of active power that needs to be reduced. : In the formula, This indicates the current grid connection point voltage. This indicates the target reference voltage (e.g., 400V, which may be adjusted depending on the actual scenario).

[0082] This invention addresses the decision-making imbalance caused by the lack of quantitative assessment of the "economic efficiency" and "reliability" of local regulation resources in existing collaborative control strategies by constructing a multi-dimensional cost evaluation system that includes a unit voltage regulation benchmark cost, a resource depletion penalty coefficient, and an identification residual penalty coefficient. Based on this system, a comprehensive voltage regulation cost is generated, effectively resolving the issue. Within the effective fluctuation window, the system calculates the unit voltage regulation cost based on the voltage response sensitivity obtained through online identification, combined with real-time electricity price, thus accurately reflecting the marginal economic cost of inverter voltage regulation under different grid intensities. Simultaneously, a resource depletion penalty coefficient is introduced to quantify the current opportunity cost of curtailment, avoiding blind reliance on local regulation in scenarios of deep curtailment. Furthermore, the identification residual penalty coefficient characterizes the model's reliability, preventing misjudgments due to noise interference or weak excitation. The comprehensive cost formed by these three factors serves as the basis for control decisions, enabling the system to prioritize inverter fine-tuning when there is slight over-limit and local voltage regulation is efficient and economical, avoiding unnecessary transformer operations to reduce mechanical wear. Conversely, when local regulation becomes uneconomical or unreliable, the system promptly triggers coordinated responses from on-load tap-changing transformers to prevent energy waste. Thus, this invention achieves a dynamic balance between regulation efficiency and economic cost, significantly improving the precision level and overall operational efficiency of voltage coordination control in high-proportion new energy distribution networks.

[0083] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A voltage coordination control system for distribution networks with a high proportion of new energy sources, characterized in that, The system includes: an acquisition module for synchronously acquiring the voltage value of the distributed photovoltaic grid-connected point and the active power output of the inverter at each sampling moment within a sliding time window; a response sensitivity analysis module for filtering out effective fluctuation windows based on the fluctuations in photovoltaic active power within the sliding time window; within the effective fluctuation window, performing linear analysis based on the first-order difference between the voltage value and active power at adjacent sampling moments, and determining the voltage response sensitivity using an online identification algorithm; a penalty analysis module for acquiring the unit price of electricity within the effective fluctuation window, determining the unit voltage regulation benchmark cost based on the unit price of electricity and the voltage response sensitivity; determining the resource depletion penalty coefficient based on the active power, rated power, and maximum historical power under the same illumination conditions of the inverter; and determining the identification residual penalty coefficient based on the residual dispersion analysis of the voltage value and active power at each sampling moment using the voltage response sensitivity; and a voltage control module for determining the comprehensive voltage regulation cost based on the unit voltage regulation benchmark cost, the resource depletion penalty coefficient, and the identification residual penalty coefficient, and performing voltage control of the grid-connected point based on the comprehensive voltage regulation cost.

2. The voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The step of selecting an effective fluctuation window based on the fluctuation of photovoltaic active power under the sliding time window includes: calculating the range of active power under the sliding time window, and when the range is greater than a preset amplitude, taking the corresponding sliding time window as an effective fluctuation window.

3. The voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The online identification algorithm is the RLS algorithm, which performs linear analysis based on the first-order difference between voltage values ​​and active power at adjacent sampling times. The online identification algorithm is then used to determine the voltage response sensitivity. This includes: the RLS algorithm performing a complete recursive parameter estimation calculation within the effective fluctuation window to obtain the voltage response sensitivity. This process includes: using zero as the initial value of the parameter estimate and using a preset matrix value as the initial value of the covariance matrix, wherein the preset matrix value is... The algorithm iterates through each point within the effective fluctuation window, calculates the prediction residual based on the linear changes in the current voltage and active power, and updates the parameter estimates and covariance matrix using a gain vector containing a regularization term. After the iteration is complete, the final parameter estimates are output as the voltage response sensitivity.

4. The voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 3, characterized in that, The regularization term is a preset minimum positive number to prevent gain calculation divergence when the change in active power approaches zero; a forgetting factor less than 1 is introduced during the covariance matrix update process to assign decreasing weights to historical data in chronological order; wherein, the preset minimum positive number is a value of The forgetting factor is 0.

98.

5. A voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The step of combining the unit price of electricity and the voltage response sensitivity to determine the unit voltage regulation benchmark cost includes: taking the absolute value of the voltage response sensitivity and determining the maximum value between the absolute value and the preset minimum constraint value as the sensitivity analysis value; wherein the preset minimum constraint value is greater than 0; using the sensitivity analysis value as the denominator and the unit price of electricity as the numerator, the unit voltage regulation benchmark cost is calculated.

6. The voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The determination of the resource depletion penalty coefficient, which combines the inverter's active power, rated power, and maximum historical power under the same illumination conditions, includes: calculating the difference between the preset maximum historical power and the active power; and then using the ratio of this difference to the rated power to obtain the active power abandonment rate. In the formula, This represents the ratio of the difference to the rated power. This indicates the active power curtailment rate. This represents the function that takes the maximum value. This represents the minimum value function; based on the active power curtailment rate and the preset curtailment resource penalty term, the resource depletion penalty coefficient is determined.

7. A voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The method involves combining voltage response sensitivity with residual discreteness analysis of voltage values ​​and active power at each sampling time to determine the identification residual penalty coefficient. This includes: calculating the first-order difference of voltage values ​​for samples at each sampling time within the effective fluctuation window, as the voltage difference; calculating the first-order difference of active power for samples at each sampling time within the effective fluctuation window, as the power difference; calculating the product of the power difference and the voltage response sensitivity, and using the difference between the voltage difference and the product as the time fitting error of the sample at the sampling time; averaging the squared values ​​of the time fitting errors of all samples at all sampling times, as the model fitting error variance; and combining the model fitting error variance, a preset noise sensitivity coefficient, and a preset benchmark model fitting variance constant to perform penalty analysis and determine the identification residual penalty coefficient. The benchmark model fitting variance constant represents the allowable noise level of the model fitting error variance under typical stable operating conditions.

8. A voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 7, characterized in that, The method for determining the identification residual penalty coefficient includes: the identification residual penalty coefficient is determined based on the ratio of the model fitting error variance to the baseline model fitting variance constant, specifically: taking 1 as the base value, an increment proportional to the ratio is superimposed, and the proportionality coefficient of the increment is a preset noise sensitivity coefficient.

9. A voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The method of determining the comprehensive voltage regulation cost by combining the unit voltage regulation benchmark cost, the resource depletion penalty coefficient, and the identification residual penalty coefficient includes: taking a weighted average of the resource depletion penalty coefficient and the identification residual penalty coefficient to obtain a total penalty index, wherein the sum of the weights of the resource depletion penalty coefficient and the identification residual penalty coefficient is 1; and taking the product of the total penalty index and the unit voltage regulation benchmark cost as the comprehensive voltage regulation cost.

10. A voltage coordination control system for distribution networks with a high proportion of new energy sources as described in claim 1, characterized in that, The voltage control of the grid connection point based on the comprehensive voltage regulation cost includes: when the comprehensive voltage regulation cost of the grid connection point is less than a preset cost threshold, voltage regulation is performed by reducing the active power of the local inverter; otherwise, a request is sent to the distribution network master station to trigger the adjustment of the tap position of the on-load tap changer.

Images