Multi-time scale voltage optimization control method and system for green electricity direct connection and grid connection point

Abstract

This invention discloses a multi-timescale voltage optimization control method and system for green electricity direct-connection points, belonging to the field of distribution network voltage control technology. Addressing the issue of grid connection point voltage exceeding limits caused by fluctuations in photovoltaic power generation and load output in green electricity direct-connection parks, a five-step control system is constructed: multi-dimensional state perception → voltage situation assessment and classification → multi-timescale control strategy matching → GFL / GFM mode adaptive switching → closed-loop feedback. It integrates millisecond-level local adaptive droop control, second-level regional multi-source reactive power coordination, and minute-level global two-stage optimization scheduling to achieve multi-timescale voltage control from rapid suppression to economic optimization. Through online identification of grid impedance parameters using a recursive least squares method with a forgetting factor, combined with a voltage change rate prediction mechanism and a three-level threshold classification mechanism, the green electricity absorption rate is maximized while ensuring voltage safety.

Description

Technical Field

[0001] This invention relates to the field of distribution network voltage control technology, specifically to a multi-timescale voltage optimization control method and system for green electricity direct connection points. Background Technology

[0002] Green electricity direct connection refers to a power supply mode in which new energy sources such as wind power, photovoltaic power, and biomass power generation are not directly connected to the public power grid, but instead supply green electricity directly to power users through dedicated power lines. This realizes the "point-to-point" physical transmission of electricity and the full physical traceability of green electricity.

[0003] Green electricity direct-connection parks are interconnected with the public power grid through a point of common coupling (PCC). These parks contain a high proportion of distributed photovoltaic (PV), energy storage, biomass power generation, and flexible loads. Because green electricity direct-connection lines are typically short, the distribution network's R / X ratio is relatively high. PV output is affected by factors such as cloud cover, resulting in minute-level power fluctuations of 50% to 80% of rated capacity, leading to frequent voltage fluctuations at the grid connection point and even exceeding limits. Traditional reactive power-based voltage regulation methods show significantly reduced sensitivity in distribution networks with high R / X ratios, making it difficult for a single control method to balance response speed and control accuracy.

[0004] Currently, scholars both domestically and internationally have proposed various voltage control methods at different time scales. CN114362196B proposes a multi-time-scale active distribution network voltage control method. On a long time scale, it uses voltage sensitivity analysis to regulate capacitor banks for reactive power compensation. On a short time scale, it regulates the active and reactive power output of distributed power sources to achieve rapid adjustment. However, this method only covers two time scales, failing to cover millisecond-level disturbance suppression and neglecting the grid-based control requirements in weak grid scenarios. CN114784813B proposes a multi-time-scale coordinated reactive power and voltage optimization control method and system. It determines the action plan of discrete reactive power sources on a day-ahead scale, solves for voltage correction limits on an intraday scale, and performs regional reactive power and voltage control on a real-time scale. However, this method only focuses on reactive power voltage regulation and does not coordinate with the active power regulation capabilities of photovoltaics and energy storage, resulting in insufficient resource utilization. CN119853057B proposes a multi-timescale voltage regulation method for distribution networks adapting to topology changes. It identifies topology changes through cluster analysis and dynamically updates the node-branch model, designing long-term and short-term scale strategies for voltage regulation. However, this method primarily addresses adaptability issues in topology-changing scenarios and does not specifically address the high R / X ratio and weak grid characteristics of green energy direct-connection parks. CN119010053B proposes a multi-timescale active distribution network voltage regulation method, constructing a centralized voltage control strategy and a real-time voltage control strategy. It optimizes the adjustment direction of the centralized strategy through feedback adjustment cycles. However, this method lacks a millisecond-level local autonomous control layer and cannot meet the rapid response requirements of second-level changes in photovoltaic output.

[0005] In summary, the existing technologies have the following main shortcomings: (1) Insufficient time scale coverage. Most existing solutions cover two time scales: day-day-intraday-real-time or long-term-short-term. They do not consider millisecond-level local fast suppression layers and cannot cope with voltage surges and drops caused by severe fluctuations in photovoltaic power in green power direct-connection parks; (2) Failure to distinguish between grid-following (GFL) and grid-connected (GFM) operation modes. Existing solutions all assume that the grid strength is sufficient and the PCC voltage is supported by the large grid. They do not consider the problem of grid-following control failure in scenarios such as weak grids or PCC disconnection in green power direct-connection parks and lack the voltage support capability of grid-connected systems; (3) Lack of Current predictive mechanisms all rely solely on voltage amplitude deviation as the primary control criterion, neglecting to consider voltage change rate. Including advance scheduling, it can only respond passively after the voltage exceeds the limit; (4) The quota management of discrete equipment is lacking. In the current optimization, the daily operation quota of discrete variables such as on-load tap changer and capacitor bank lacks intelligent allocation strategy, which easily leads to problems such as quota exhaustion in the morning and no adjustment means when the limit is exceeded in the afternoon. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-timescale voltage optimization control method and system for green electricity direct connection points, in order to solve the problem of how to achieve multi-timescale continuous voltage control from millisecond-level rapid disturbance suppression to minute-level economic optimization in green electricity direct connection park distribution networks, while adaptively switching between grid-connected and grid-structured operation modes, so as to maximize the green electricity absorption rate and system operation economy while ensuring the voltage safety of the connection point.

[0007] To address the aforementioned technical problems, this invention provides a multi-time-scale voltage optimization control method for green electricity direct-connection grid points, comprising the following steps: Step S1: Collect the three-phase voltage and current of the PCC through the PMU, and calculate the actual power, voltage deviation and voltage change rate based on the voltage and current. Simultaneously collect the operating data of each controllable device and the time-of-use electricity price. Collect the tap position signal of the transformer tap changer in real time through the on-load tap-changing transformer configured at the grid connection point, and calculate the short-circuit ratio at the grid connection point based on the voltage and current. Step S2: Based on the voltage deviation, perform voltage status assessment and classification according to three threshold levels, and notify the control layer in advance of the assessment and classification results using the voltage change rate prediction mechanism; Step S3: Match multi-timescale control strategies based on the assessment and grading results; Step S4: Implement adaptive switching between GFL / GFM modes based on the short-circuit ratio composite criterion at the grid connection point; Step S5: Monitor the deviation between the actual power and the predicted power through the energy management system, and determine whether the deviation exceeds a threshold; wherein, the predicted power is output by the prediction model, and the predicted power includes photovoltaic output and load power; If yes, return to step S3 and correct the parameters online; otherwise, return to step S5.

[0008] As a preferred embodiment, a multi-timescale voltage optimization control method for green electricity direct-connection grid points, wherein the voltage status assessment and classification based on the voltage deviation according to three-level thresholds specifically includes: Determine whether the voltage deviation is less than or equal to 0.07. ; If yes, it is classified as Level I, and only the local control layer operates; otherwise, it is determined whether the voltage deviation is greater than 0.07. And less than or equal to 0.2 ; If yes, it is classified as Level II, and the regional coordination layer is triggered sequentially, and the global optimization layer is notified to increase the voltage weight. If no, it is classified as Level III, and all control regional coordination layers are activated and the global optimization layer is urgently re-optimized.

[0009] As a preferred embodiment, a multi-timescale voltage optimization control method for green electricity direct grid connection points, which specifically includes notifying the control layer in advance of the evaluation and classification results using the voltage change rate prediction mechanism, includes: Determine whether the voltage change rate is greater than zero; If not, notify the regional coordination layer to pre-schedule energy storage discharge to replenish the undervoltage; if yes, notify the regional coordination layer to pre-schedule energy storage charging to suppress overvoltage, and determine whether the voltage change rate is greater than 0.03. If yes, then even if the voltage deviation is still within Level I, the regional coordination layer must be notified to pre-schedule energy storage charging for overvoltage suppression. If no, then return to determine whether the voltage deviation is greater than 0.07. And less than or equal to 0.2 The steps are as follows. As a preferred embodiment, a multi-timescale voltage optimization control method for green electricity direct-connection grid points, wherein matching the multi-timescale control strategy based on the evaluation and grading results specifically includes: When the condition is determined to be Level I, local adaptive droop control is executed; When the classification is Level II, FF-RLS online impedance identification and multi-source reactive power coordination are triggered. When the level is determined to be Level III, the entire control area coordination layer is activated and the global optimization layer is triggered for emergency re-optimization. The global optimization layer adopts a two-stage optimization using a day-ahead mixed integer quadratic programming scheduling model and an intraday nonlinear programming rolling correction model.

[0010] As a preferred embodiment, a multi-timescale voltage optimization control method for green electricity direct grid connection points, wherein the global optimization layer employs a two-stage optimization process using a day-ahead mixed integer quadratic programming scheduling model and an intraday nonlinear programming rolling correction model, specifically including: The day-ahead mixed integer quadratic programming scheduling model optimizes the baseline curves of discrete and continuous variables for the next 24 hours with a granularity of 15 minutes. The discrete variables include the Tap position signal and the capacitor switching status. The intraday nonlinear programming rolling correction model optimizes the continuous variables for the next 4 hours with a granularity of 5 minutes. It achieves closed-loop control through model prediction and rolling time domain, advances the window every 5 minutes, and executes only the first time period instruction.

[0011] As a preferred embodiment, a multi-time-scale voltage optimization control method for green electricity direct-connection grid points, wherein the total reactive power reference value of the inverter is determined according to the following formula in the local adaptive droop control:

[0012] in, The adaptive coefficient is the steady-state reactive power baseline value issued by the global optimization layer. , The reactive power margin factor, Voltage rate of change factor, base droop coefficient , The rated capacity of the inverter. U ref =1p.u, This represents the maximum controllable voltage deviation range of the local control layer. For voltage deviation, J u This is the virtual voltage inertia coefficient. This represents the rate of change of voltage.

[0013] As a preferred embodiment, a multi-time-scale voltage optimization control method for a green electricity direct-connection grid-connected point, wherein the adaptive switching of GFL / GFM mode based on the short-circuit ratio composite criterion of the grid-connected point specifically includes: The composite criterion for the short-circuit ratio at the grid connection point is:

[0014] in, This is the system equivalent voltage at PCC. P rated For the rated capacity of grid-connected equipment, The equivalent impedance from the PCC to the system's equivalent power supply is obtained through the FF-RLS online impedance identification. For GFM switching threshold, The GFL recovery threshold; when Or when PCC disconnects, it triggers GFL→GFM switching. At that time, GFM→GFL recovery is triggered. The value is 3.0. The value is 4.5, and the hysteresis width is ; When in the hysteresis interval At that time, maintain the original mode.

[0015] As a preferred embodiment, a multi-time-scale voltage optimization control method for green electricity direct-connection points includes the following online correction parameters: The global optimization layer is triggered to update the daily plan, and the line impedance parameters of the FF-RLS online impedance identification and the prediction model coefficients are corrected online.

[0016] As a preferred embodiment, a multi-timescale voltage optimization control method for green electricity direct grid connection points is provided, wherein the controllable equipment includes photovoltaic, energy storage, load, and reactive power compensation equipment.

[0017] To address the aforementioned technical problems, this invention also provides a multi-time-scale voltage optimization control system for green electricity direct-connection grid points, comprising: The data acquisition module is used to acquire the three-phase voltage and current of the PCC, the operating data of each controllable device, the time-of-use electricity price, and the tap position signal of the transformer tap changer. The calculation module is used to calculate the actual power, voltage deviation, voltage change rate, and grid connection point short-circuit ratio based on the voltage and current collected by the data acquisition module. The assessment and grading module is used to assess and grade the voltage situation according to the voltage deviation calculated by the calculation module based on three threshold levels, and to notify the control layer in advance of the assessment and grading results using the voltage change rate prediction mechanism. The result matching module is used to match multi-timescale control strategies based on the results obtained from the evaluation and grading module. An energy management system is used to monitor the deviation between the actual power and the predicted power, and to determine whether the deviation exceeds a threshold. The predicted power is output by a prediction model and includes photovoltaic output and load power. If the deviation is exceeded, the result matching module and the online correction module are triggered; otherwise, monitoring continues. The online correction module is used to correct the line impedance parameters and prediction model coefficients of FF-RLS online impedance identification.

[0018] Compared with existing technologies, the multi-timescale voltage optimization control method and system for green electricity direct grid connection points provided by this invention has at least the following beneficial effects: Addressing the issue of grid connection point voltage exceeding limits caused by fluctuations in photovoltaic power output and load in green electricity direct connection parks, a five-step control system is constructed: multi-dimensional state perception → voltage situation assessment and classification → multi-timescale control strategy matching → GFL / GFM mode adaptive switching → closed-loop feedback. It integrates millisecond-level local adaptive droop control, second-level regional multi-source reactive power coordination, and minute-level global two-stage optimization scheduling to achieve multi-timescale voltage control from rapid suppression to economic optimization. Through online identification of grid impedance parameters using a recursive least squares method with a forgetting factor, combined with a voltage change rate prediction mechanism and a three-level threshold classification mechanism, the green electricity absorption rate is maximized while ensuring voltage safety. Attached Figure Description

[0019] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0020] Figure 1 This is a flowchart of a multi-time-scale voltage optimization control method for green electricity direct-connection grid points provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a three-level voltage hierarchical logic provided in an embodiment of the present invention; Figure 3 This is a flowchart of an FF-RLS online impedance identification method provided in an embodiment of the present invention; Figure 4 This is a GFL / GFM mode switching criterion diagram provided in an embodiment of the present invention; Figure 5 A comparison of the grid connection point voltage control effect provided by an embodiment of the present invention; Figure 6 A schematic diagram illustrating the voltage classification ratio comparison (Level I / II / III) provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of a multi-time-scale voltage optimization control system for a green electricity direct-connection grid point provided in an embodiment of the present invention; In the diagram: 1. Data acquisition module; 2. Calculation module; 3. Evaluation and classification module; 4. Result matching module; 5. Energy management system; 6. Online correction module. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings.

[0022] The core of this invention is to provide a multi-time-scale voltage optimization control method and system for green electricity direct connection grid connection points, which solves the problems of voltage exceeding limits and multi-time-scale collaborative control at green electricity direct connection grid connection points.

[0023] Figure 1 This is a flowchart of a multi-time-scale voltage optimization control method for green electricity direct-connection grid points provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a three-level voltage hierarchical logic provided in an embodiment of the present invention; Figure 3 This is a flowchart of an FF-RLS online impedance identification method provided in an embodiment of the present invention; Figure 4 This is a GFL / GFM mode switching criterion diagram provided in an embodiment of the present invention; Figure 5 A comparison of the grid connection point voltage control effect provided by an embodiment of the present invention; Figure 6 A schematic diagram illustrating the voltage classification ratio comparison (Level I / II / III) provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of a multi-time-scale voltage optimization control system for a green electricity direct-connection grid point provided in an embodiment of the present invention; see also Figures 1 to 7 As shown.

[0024] The control method provided by this invention is applicable to green electricity direct connection projects and industrial park distribution networks that include a high proportion of distributed photovoltaic, energy storage, biomass power generation and flexible loads.

[0025] The present invention provides a multi-timescale voltage optimization control method for green electricity direct grid connection points, which can achieve high-precision, fast-response, and low-overshoot control of grid connection point voltage; multi-timescale coordination of source-storage-load-grid; support for grid-type control and islanded operation; and take into account voltage quality, green electricity consumption, and economic operation.

[0026] This invention adopts a three-tiered hierarchical control architecture: "local-regional-global". Local control layer (millisecond level): The local control layer includes an inverter / energy storage converter (Power Conversion System, PCS). The local controller independently executes droop control and reactive power distribution based on local voltage / current sampling. The outer layer control (voltage RMS estimation, droop calculation, global optimization command reference tracking) has a cycle of 10ms, corresponding to one fundamental half-cycle of a 50Hz system. Voltage sag / surge detection refers to the IEC61000-4-30A level half-cycle RMS refresh rate. The inner layer current loop and PWM modulation operate at the switching frequency (several kHz to tens of kHz). The entire process does not rely on external communication links.

[0027] Regional coordination layer (second-level): The regional coordination layer is responsible for multi-source reactive power coordination and allocation, load classification scheduling, and emergency protection. Within the station, IEC61850 GOOSE (end-to-end ≤3ms) and EtherCAT real-time Ethernet (single-frame whole-station transmission <1ms) are used to achieve high-speed data interaction between multiple nodes. The control master cycle is measured in seconds (1-5s). The local edge computing unit (CCU / MGCC) aggregates phasor data from the synchronous phasor measurement unit (PMU), battery management system (BMS), and distributed control system (DCS), completing online power-voltage sensitivity identification and load classification scheduling, and distributing multi-source reactive power allocation benchmarks to the local layer. This layer possesses station-side autonomy capabilities and does not rely on the remote link of the energy management system (EMS).

[0028] Global optimization layer (minute level): The global optimization layer, through EMS via 4G / 5G or fiber optic leased lines, aggregates operational data from the entire region and formulates a 24-hour plan for the following day (time resolution of 15 minutes) and an intraday rolling correction plan (time resolution of 5 minutes). The optimization objectives are to minimize grid connection point voltage deviation and achieve optimal overall economic operation (including green energy consumption). The optimization results of the global optimization layer serve as the benchmark values ​​for the local control layer. Under normal circumstances ( The local control layer strictly follows the instructions of the global optimization layer; when the voltage exceeds the limit ( When a device constraint is triggered, the local layer autonomously deviates from the baseline, performs safety control, and uploads the measured status; the global layer initiates event-triggered re-optimization (rolling correction of subsequent windows) with the current measured value as the initial value. Each layer operates the closed loop independently.

[0029] Based on the above three-layer architecture, the voltage deviation of the PCC point is directly connected to the green electricity. (in the formula) This is the per-unit value of the PCC common coupling point voltage, in pu. Driven by the rated voltage per unit value, a multi-level time-scale control system is constructed by integrating local fast control and global optimization scheduling.

[0030] A method for multi-time-scale voltage optimization control of green electricity direct-connection grid points includes the following steps: Step S1: Collect the three-phase voltage and current of the PCC through the PMU, and calculate the actual power, voltage deviation and voltage change rate based on the voltage and current. Simultaneously collect the operating data of each controllable device and the time-of-use electricity price. Collect the tap position signal of the transformer tap changer in real time through the on-load tap changer configured at the grid connection point, and calculate the short circuit ratio (SCR) at the grid connection point based on the voltage and current.

[0031] Specifically, the three-phase voltage is synchronously acquired at a 20ms cycle by a PMU deployed in the PCC. and current Calculate voltage deviation and voltage change rate Simultaneously collect voltage / current / frequency / power factor at the green electricity direct-connection grid point, as well as the current operating data of each controllable device, such as the active and reactive power of the photovoltaic power generation unit. Energy storage state of charge (SOC) and charging / discharging power Controllable load power Reactive power compensation device reactive power output Meteorological forecast data (mainly irradiance) ) and time-of-use electricity price signals The grid connection point is equipped with an on-load tap changer transformer (OLTC) to collect the current tap position signal (DI / communication) in real time. If the system is equipped with synchronous units such as biomass generators, their active power output, reactive power output, and available fuel output are collected via the power plant's DCS through IEC 61850 / Modbus. Calculate the short-circuit ratio at the grid connection point, which is used to determine the switching between Grid-Following (GFL) and Grid-Forming (GFM) grid types.

[0032] Step S2: Based on the voltage deviation, perform voltage status assessment and classification according to three threshold levels, and notify the control layer in advance of the assessment and classification results using the voltage change rate prediction mechanism.

[0033] calculate and Symbols indicate direction: This indicates overvoltage (voltage is too high). Undervoltage (voltage is too low); The symbols indicate the trend: positive indicates that the voltage is rising, and negative indicates that the voltage is falling.

[0034] This scheme primarily applies to direct green power connections to the 10kV / 20kV grid. Other voltage level direct green power connection projects can refer to this scheme for implementation. According to Clause 4.1 of GB / T 12325-2008 Power Quality Supply Voltage Deviation, the voltage deviation limit for 20kV and below supply voltage is ±7%, meaning the normal voltage range at the grid connection point is 0.93~1.07 pu. Based on this, the three-level voltage classification thresholds are set.

[0035] Classification is based on a three-tiered threshold progression: Level I (Normal Fluctuation Zone): Only the local control layer runs; Level II (Warning Zone): The regional coordination layer is triggered sequentially, and the global optimization layer is notified to increase the voltage weight (emphasis on voltage suppression during overvoltage and emphasis on voltage boosting during undervoltage). Level III (Emergency Crossing Zone): The regional coordination layer is fully activated and the global optimization layer undergoes emergency re-optimization. Overvoltage-directed photovoltaic (PV) active power reduction is enabled, meaning that the grid connection voltage is lowered by reducing PV active power output. The regional coordination layer can urgently reduce PV output, while the global optimization layer only... (Energy storage can no longer absorb excess photovoltaic power) and it is predicted that photovoltaic output will be reduced to achieve economic optimization when the time limit is exceeded. If the State of Charge (SOC) still has a margin, the benefits of energy storage charging to absorb photovoltaic power and discharging at the right time are better than direct reduction. In the case of undervoltage, priority should be given to discharging energy storage and reducing the power consumption of non-essential loads, such as fine-tuning air conditioning temperature, reducing water pump frequency, and halving some lighting, to avoid direct cut-off affecting production operations.

[0036] Prediction: If Significantly increased (e.g.) According to the SEMI F47 semiconductor equipment immunity specification, semiconductor manufacturing equipment can withstand a voltage change rate of approximately 0.3 pu / s. Taking a safety factor of 1 / 10, 0.03 pu / s, this is taken as... dU / dt Early warning thresholds ensure sufficient prediction time before voltage sag / droop events occur, even if Even if it's still within Level I, the corresponding control layer needs to be notified in advance based on the direction, i.e. (Voltage surge) Notify the regional coordination layer to pre-schedule energy storage charging for overvoltage suppression. (Voltage drop) Notify the regional coordination layer to pre-schedule energy storage discharge to supplement the undervoltage.

[0037] Multi-reactive-source coordination priority: According to the scheduling sequence, slow-speed equipment starts first, i.e., OLTC on-load tap changer (5~10s, ≤30 operations per day, prioritized to gain operating time) → Biomass synchronous generator automatic voltage regulator (AVR; 0.5~1s, handling slow reactive base load) → Static Var Generator (SVG; <20ms) → Energy storage PCS reactive power (<10ms, fast fallback) → Photovoltaic inverter reactive power. After the biomass unit handles the slow reactive base load, the remaining rapid fluctuation deviation is compensated by the energy storage PCS with a 200ms response (PCS power electronic device switching frequency in the kHz range, reactive power step response <10ms).

[0038] Step S3: Match multi-timescale control strategies based on the evaluation and grading results (millisecond level → second level → minute level).

[0039] When the condition is determined to be Level I, local adaptive droop control is executed; When the classification is Level II, FF-RLS online impedance identification and multi-source reactive power coordination are triggered. When the level is determined to be Level III, the entire control area coordination layer is activated and the global optimization layer is triggered for emergency re-optimization. The global optimization layer adopts a two-stage optimization using a day-ahead mixed-integer quadratic programming (MIQP) scheduling model and an intraday nonlinear programming (NLP) rolling correction model.

[0040] Local control layer (millisecond level): Adaptive droop control The traditional formula for fixed-coefficient droop is: (1) in This is the inverter reactive power reference value (MVar). This is the steady-state reactive power output (MVar). For a fixed droop factor (MVar / pu).

[0041] This invention improves upon adaptive coefficients. : (2) in Based on the droop factor (MVar / pu). U ref =1p.u., The rated capacity of the inverter (MVA). The maximum controllable voltage deviation range of the local control layer is consistent with the threshold of Level II in step S2; reactive power margin factor. (Dimensionless) is: (3) in This represents the current reactive power margin (MVar). This is the upper limit of reactive power capacity (MVar). The engineering tuning parameters are designed based on Article 5.1.2 of the "Technical Regulations for Photovoltaic Power Stations Connected to Power Systems" (GB / T 19964-2024), which requires photovoltaic power stations to have reactive power regulation capabilities. Specific values ​​are determined through simulation of the inverter's reactive power capacity margin and voltage regulation sensitivity. Physically, a larger reactive power margin means a more aggressive droop, while a smaller margin means a more conservative approach to avoid overload. Voltage change rate factor. for: (4) in, The rated rate of change, This is the voltage change rate response gain coefficient (dimensionless). The voltage measurement is filtered using a first-order infinite impulse response (IIR) low-pass filter (the filtering implementation follows the voltage measurement filtering requirements in Clause 5.4 of IEEE 1547-2018 Interconnection and Interoperability Standard for Distributed Energy Resources and Related Power Systems). δ The specific value is determined by simulation based on a trade-off between voltage change rate response sensitivity and noise immunity: too small a value results in a sluggish response, while too large a value makes it susceptible to noise interference and oscillations. (The voltage change rate after cutoff frequency is also considered.) time constant It effectively suppresses the interference of power electronic switch noise (3~20kHz) on the differential term and avoids oscillation caused by high-frequency jumps in the droop coefficient. The physical meaning is that the faster the voltage changes, The larger the droop factor, the faster it responds to voltage changes and suppresses fluctuations. (Adaptive droop factor...) , Factoring primarily involves algebraic operations; a first-order IIR low-pass filter requires only 3 floating-point operations, resulting in a total multiply-accumulate operation volume within a 10ms cycle. Modern digital signal processors (DSPs, such as the TMS320F28335 with a 150MHz clock speed) can complete the task within 10ms. The floating-point operations provide ample computational margin.

[0042] Traditional "virtual inertia" is mostly used in active-frequency (PF) applications. PfThis invention introduces virtual voltage inertia to control and smooth out the rate of frequency change. (MVar·s / pu, typical value 0.05~0.2), and creatively applied to reactive power-voltage ( Q- V Control, the total reactive power reference value after superposition: (5) in The steady-state reactive power baseline value (MVar) is issued by the global optimization layer. If there is no global optimization layer instruction, it defaults to the current steady-state reactive power output. J u The virtual voltage inertia coefficient (MVar·s / pu, positive value), when dU / dt >0 means that this term is negative when the voltage rises, prompting the inverter to absorb reactive power to suppress the voltage rise; the opposite is also true. This is a voltage rate-of-change-based advance compensation term used to provide instantaneous reactive power support during rapid voltage changes, simulating the excitation inertia of a synchronous machine. It is applied proportionally to the reactive power margin among multiple reactive power source devices (including inverters, energy storage PCS, and static var generators SVG). Distribute: (6) If compensated by the local control layer The local control layer continues to operate independently; if It automatically triggers the intervention of the regional collaboration layer or the global optimization layer.

[0043] Regional coordination layer (second-level event triggering): Load scheduling and emergency protection The regional coordination layer sorts adjustable loads by response time from fastest to slowest: fast-switching loads (<200ms) → medium-speed adjustable loads (<5s, adjustment frequency ≤5 times / h) → slow-speed interruptible loads (<1min, requires ≥15min to recover after adjustment). During undervoltage, loads are dispatched in this order, with feedback on actual voltage changes after each adjustment to update remaining power demand and avoid over-adjustment.

[0044] If the system is equipped with biomass synchronous generator sets and operates connected to the grid, it serves as a basic reactive power support source for the regional collaborative layer. AVR (Automatic Dynamic Range) is used to change the excitation current to achieve smooth bidirectional regulation between lagging and leading phases; the regulation range is limited by… PQCapacity curve and rated apparent power / thermal stability limit. AVR response is approximately 0.5~1s, using Smith's prediction compensation, with commands taken as the moving average of the past 1s. Short-term reactive power support is provided by the excitation system, and short-term frequency support is provided by the rotor inertia. After the biomass unit assumes the slow reactive base load, the remaining rapid fluctuation deviation is compensated by the energy storage PCS with a response in the 200ms range, where the switching frequency of the PCS power electronic devices is in the kHz range, and the reactive power step response is <10ms.

[0045] First, calculate the power-voltage sensitivity. The external power grid of the PCC is equivalent to a Thevenin equivalent circuit – an ideal voltage source. (The system equivalent voltage at PCC is approximately equal to the grid rated voltage, which differs from the measured operating voltage.) U PCC (t) Series equivalent impedance Under the DistFlow model of the distribution network, the voltage change and the power change approximately satisfy a linear relationship. Therefore, we can conclude that: (7) (8) in and These are the equivalent resistance and equivalent reactance (in Ω) of the PCC to the system's equivalent power supply, respectively. The real and imaginary parts (not just line parameters); in the sensitivity calculation and energy storage active power compensation formula of this paper. R, X Substitute the per-unit value, i.e., divide by the reference impedance. Z base = U base 2 / S base， per unit value ≈1p.u.

[0046] To obtain the external power grid characteristics of the PCC in real time, the Forgetting Factor Recursive Least Squares (FF-RLS) method is used to identify the Thevenin equivalent impedance online. The measurement equation is based on the small-signal linearization approximation. (9) in, k For FF-RLS recursive sampling sequence number ( k =1,2,…, corresponding to the first… k (each PMU sampling time) P ( k ), Q ( kThe active and reactive power at the PCC are calculated from the voltage and current synchronously acquired by the PMU. , These represent the changes in active and reactive power at adjacent sampling times at PCC, respectively. This represents the change in PCC voltage between adjacent sampling times.

[0047] Parameter vector to be identified Measurement vector The FF-RLS recursive formula is as follows: Gain vector: (10) Parameter update: (11) Covariance update: (12) in This is the forgetting factor, typically ranging from 0.95 to 0.99, balancing tracking speed and noise resistance. →1 Tracking is slow but noise resistance is strong, when →0.95 provides a fast response, suitable for scenarios with frequent changes in operating conditions. FF-RLS is a passive identification method, utilizing natural fluctuations in load and photovoltaic power, without the need for actively injecting harmonic disturbance signals. An excitation detection and parameter freezing mechanism is added: when... Fluctuation < 0.005 pu and Freeze FF-RLS parameter updates when the fluctuation is less than 1% of the rated reactive power capacity. , (Maintain current values) to prevent the covariance matrix from diverging due to a lack of sustained stimulus. The identified... R, X Used for the above sensitivity calculations and S4 SCR calculate: (13).

[0048] To eliminate residual bias at the local control level, reactive power compensation is prioritized: (14) Symbol convention: Overvoltage At that time, it absorbs reactive power to smooth out the voltage and avoids positive feedback; k q (Same as S3.1 local reactive power-voltage droop control) Identified by FF-RLS Adaptive tuning; where if / MVar (The line is nearly purely resistive, and the reactive power voltage regulation sensitivity is extremely low), jump to the active power compensation branch. In the formula, all quantities are per-unit values, Δ P ESS (t) multiplied by the base capacity Sbase (i.e., obtain MW); if the reactive power is insufficient even after reaching the PCS capacity limit, supplement with active power. ,like Then it is limited to the rated power of the PCS. Rated capacity (MVA) of the energy storage converter.

[0049] Each energy storage unit is evenly distributed according to its State of Charge (SOC), meaning that during discharge, it is distributed according to... Proportional allocation, charging according to Proportional allocation; all instructions must meet PCS capacity constraints. (15).

[0050] Global optimization layer (5min-15min cycle): Two-stage multi-objective optimization using a day-ahead mixed integer quadratic programming scheduling model and an intraday nonlinear programming rolling correction model.

[0051] The global optimization layer, deployed within the EMS, is the core of global optimization decision-making. It operates in two phases: a day-ahead mixed-integer quadratic programming scheduling model optimizes the baseline curves of discrete and continuous variables for the next 24 hours at a 15-minute granularity, including transformer tap positions, capacitor switching status, photovoltaic output plans, and energy storage charging / discharging plans; an intraday nonlinear programming rolling correction model optimizes the continuous variables for the next 4 hours at a 5-minute granularity, adjusting the output setpoints of each controllable resource and the action corrections of discrete equipment, achieving closed-loop control through model predictive control in the rolling time domain. The window advances every 5 minutes, executing only the first 5-minute instruction.

[0052] Decision vector The mixed-integer quadratic programming scheduling model plans the next 24 hours in 15-minute × 96 time periods, focusing on comprehensive economic operation and locking discrete variables, such as Tap position and Cap switching status, as long as voltage safety constraints are met. The intraday nonlinear programming rolling correction model optimizes the next 4 hours in 5-minute × 48 time periods, using the previous day's plan as the tracking benchmark, with weights dynamically switching in real time and automatically increasing the voltage deviation target when the limit is exceeded, optimizing only continuous variables.

[0053] (16) in, This indicates that the voltage deviation at the grid connection point is minimized. The vector of decision variables represents the optimal overall economic operation (including green energy consumption). Includes decision variables for all time periods within the daily rolling optimization time domain:

[0054] in , The integer discrete variables are optimized for the day-ahead phase, and locked to the day-ahead planned values ​​for the intraday phase. The decision vector degenerates into a subset containing only continuous variables, with the remainder being continuous variables.

[0055] Active power output (MW) of each photovoltaic inverter Reactive power output (MVar) of each inverter Active power of energy storage charging and discharging (MW, positive for discharging, negative for charging) Energy storage reactive power output (MVar) Reactive power output (MVar) of a static var generator (SVG). Flexible load active power regulation (MW, allocated to each load by the regional coordination layer) Power purchased from the grid (MW) On-load tap position of transformer (integer discrete variable) Number of parallel capacitor banks switched on (integer discrete variable) : No. The active power output (MW) of a biomass unit is constrained by the unit's rated capacity and the amount of fuel available. : No. The reactive power (MVar) of a biomass generator unit is constrained by the capacity of the synchronous generator. Optimization Objective Sub-objective 1: Minimize voltage deviation at the point of direct grid connection for green electricity (technical objective) (17) In the formula: t For scheduling period index ( t =1,2,…, N ); To optimize the number of window periods, intraday rolling is 5-minute granularity × 48 time periods = 4-hour window, and day-ahead scheduling is 15-minute granularity × 96 time periods = 24 hours; M The penalty coefficient for exceeding the limit (dimensionless, typically 10). 2 ~10 3 ); (Same as S2 level II threshold); . For the early warning indication function, when Exceed (Soft constraint threshold, with a safety margin of 0.07 pu below the hard constraint) is set to 1, otherwise to 0. Squared term. Its function is to apply an increasing penalty to persistent small deviations; the larger the deviation, the faster the penalty increases, preventing the optimizer from being indifferent to small voltage shifts. Indicator function term. Function: Much larger than the square term, once Exceeding the soft constraint threshold This triggers a penalty, reducing the frequency of the optimizer exceeding limits at any given time, indirectly preventing voltage degradation from reaching Level III. This soft constraint threshold... (Minimum ±0.07 pu below the hard constraint C4) Gradient protection is formed with the hard constraint: The soft constraint-driven optimizer actively controls the voltage within 0.05 pu to leave a safety margin, while the 0.07 pu hard constraint is the safety baseline with higher priority than the soft constraint. Through the guidance of soft constraints, the voltage is kept within the safety margin for a long time, avoiding frequent touching of the hard constraint, extending equipment life and improving power quality.

[0056] Sub-objective two: Optimal overall economic operation (including green electricity consumption): (18) In the formula: C base Based on the baseline cost (yuan), make J 2. Dimensionless and comparable cost items J 1 Weighted summation. The meanings of each term are: The price is based on time-of-use electricity pricing (RMB / MWh). Power purchased from the grid (MW); The depreciation factor for energy storage charging and discharging (yuan / MWh, typical value 50~150, i.e. 0.05~0.15 yuan / kWh, taking into account the battery cycle life loss); The inverter reactive power loss factor (yuan / MVarh, typical value 10~30). The price per unit of biomass fuel (RMB / MWh) The fuel consumption thermal power (MW) for a given period; The economic penalty for photovoltaic power generation is reduced (RMB / MWh, usually taken as 300~500 RMB from the photovoltaic feed-in tariff). This refers to the photovoltaic power reduction (MW), reflecting the green electricity consumption target. Transformer tap changer and capacitor operating costs are also included. Adopt quota-aware dynamic penalty: ,in The remaining number of allowed operations for the day (30 taps per day, 5 capacitors per day). The equivalent cost of the baseline action (yuan / time), where this term in the summation formula is actually... , The number of tap / capacitor actions in time period t (an integer of 0 / 1, with the product dimension in yuan, consistent with other cost items). The fewer the remaining actions, the heavier the penalty. When approaching the quota, it naturally suppresses the impulse to act, so that the limited daily quota is intelligently allocated among 96 15-minute optimization points, avoiding the situation where the quota is exhausted in the morning and there is no means of adjustment when the voltage exceeds the limit in the afternoon.

[0057] Constraints Load constraint description: The power of slow interruptible loads (response time <1min) is regarded as an unadjustable variable in the optimization model and only participates in the C1 power balance constraint (interrupted sequentially by the regional coordination layer in case of emergency undervoltage). After adjustment, it can be restored after ≥15min; the power of fast switchable loads (<200ms) and medium-speed adjustable loads (<5s, ≤5 times / h) are optimization decision variables.

[0058] C1. Power Flow Constraints: Use DistFlow or a linearized power flow model.

[0059] (19) C2·Inverter capacity constraint: ( S i For the first i (Inverter rated capacity, MVA).

[0060] C3·Energy Storage Constraint: During Discharge , P ESS >0; During charging: , P ESS <0; the state of charge constraint is Energy storage capacity constraint is Energy storage ramping constraints are Furthermore, the DC-side voltage is implicitly constrained. Δ represents the energy storage ramp-up rate (MW / h). t The duration is calculated in hours (h). The daily rolling duration is 5 minutes, and the daily scheduling duration is 15 minutes. E max Rated energy storage capacity (MWh) η ch Energy storage charging efficiency (dimensionless). SOC min and SOC max These represent the lower and upper limits of the energy storage SOC operation, respectively.

[0061] C4·Voltage Safety: (±7%, consistent with GB / T12325-2008 10kV voltage deviation limit and S2 Class I threshold), grid connection point power factor (From advanced to lagging, according to Section 5.4 of GB / T 29319-2024 Technical Regulations for Photovoltaic Power Generation Systems Access to Distribution Networks).

[0062] C5 Transformer tap changer: There are 17 voltage levels in total, with ±8 levels. Each level adjusts the voltage by approximately 1.25%, or 0.0125 pu. The full scale is approximately ±10%, meaning the maximum adjustment range is 0.2 pu. The total number of operations per day... The mechanical action time is approximately 5-10 seconds per gear, and the lock is engaged for 60 seconds after each action.

[0063] C6·Capacitor: Total number of switching operations per day .

[0064] C7·Power Reduction from Photovoltaics: Only when and At times, photovoltaic active power reduction is permitted.

[0065] C8 Gradient: .

[0066] C9 Biomass Unit: , Total fuel consumption shall not exceed the upper limit, of which With merit and effort satisfaction , The typical value for biomass power generation efficiency is 0.25~0.35, and .

[0067] Two-stage solution Day-ahead scheduling (15min x 96 time periods): Using photovoltaic and load forecasts as inputs, a mixed-integer quadratic programming model is constructed, with decision variables including discrete variables. , With continuous variables The solution is obtained using Gurobi or CPLEX, with a single solution time of <30s. The optimization results are used as a baseline curve for intraday rolling calculations.

[0068] Intraday Rolling (5min x 48 time periods, optimizing for the next 4 hours): Using the day-ahead plan as the tracking benchmark, input ultra-short-term forecasts (PV model – driven by Sandia physics + Long Short-Term Memory network data, with a combined Mean Absolute Percentage Error (MAPE) <8%; Autoregressive Integrated Moving Average (ARIMA) model, MAPE <3%), construct a nonlinear programming rolling correction model (containing only continuous variables), and use a sequential quadratic programming algorithm combined with an interior-point optimizer solver, with a single solution time <10s. Only the first time period (5min) instruction is executed; instructions for the remaining time periods are used for the next rolling window.

[0069] Intraday weighting adaptive (only during intraday rolling phase) During the intraday phase, the objective function weight combination is dynamically adjusted based on the real-time voltage grading results of S2. The switching process uses a first-order IIR low-pass filter (time constant τ=30s) for smooth transition to avoid control oscillations. Level I normal operating condition ( ): , (Balanced mode, balancing voltage quality and economical operation) Level II warning condition ( ): , (Emphasis on voltage quality) Level III Over-limit Operating Conditions ( ): , (Prioritize eliminating violations of limits; photovoltaic active power reduction can be activated, aiming to "minimize economic losses under safe conditions.") Step S4: Implement adaptive switching between GFL / GFM modes based on the short-circuit ratio composite criterion at the grid connection point; Based on short-circuit ratio The core criterion is the ratio of the system short-circuit capacity at the PCC to the rated capacity of the grid-connected equipment, where... P rated The rated capacity (MW) of the grid-connected equipment is used; simultaneously, voltage-active power sensitivity is used as an auxiliary criterion to avoid misjudgment in the critical region by relying solely on the SCR indicator. Definition For GFM switching threshold, The GFL recovery threshold is [value], and the hysteresis width is [value]. - To prevent frequent switching of the critical region, typical value , The switching target is the energy storage PCS. The green electricity is directly connected to the park and operates in GFL mode most of the time. GFM is only triggered briefly in weak grid or island conditions, and switches back to GFL after the grid is restored.

[0070] GFL mode ( It is in the hysteresis interval (Keep the original mode internally) All grid-connected equipment operates in grid-following mode. Photovoltaic inverters track the grid voltage and frequency via a phase-locked loop (PLL), outputting active power and participating in reactive power regulation according to the local control layer strategy in step S3. Energy storage PCS executes the charging and discharging plan issued by the global optimization layer, while also participating in adaptive droop reactive power regulation in step S3. Biomass synchronous generators provide base load reactive power support through AVR excitation regulation. The regional coordination layer and the global optimization layer operate normally and collaboratively in this mode.

[0071] GFM mode ( (Or PCC disconnected, power grid weak or in islanded state): Trigger the GFL→GFM switchover. Select the energy storage PCS with the highest SOC as the main grid unit and start the Virtual Synchronous Generator (VSG) control.

[0072] Active-frequency loop: (20) in The active power command (kW) for the main network PCS. The active power (kW) of the dispatch plan. The active power-frequency droop factor (kW / Hz) is used. This is the virtual inertia (kW·s / Hz). Reference frequency (Hz) The measured frequency (Hz) at the PCS output.

[0073] Reactive power-voltage loop: (twenty one) in (Reactive power reference value, same as defined in S3.1) The planned reactive power (MVar) issued by the dispatcher. The reactive power-voltage droop factor (MVar / pu) is used. The measured voltage (pu) at the PCS outlet.

[0074] The remaining energy storage PCS automatically switches to a symmetrical droop mode, achieving automatic active / reactive power sharing among multiple VSGs through droop characteristics (droop coefficient inversely proportional to the rated capacity of each unit). Simultaneously, virtual impedance droop control is introduced to suppress circulating currents among multiple VSGs. The photovoltaic inverter operates in two scenarios: if equipped with local energy storage, the energy storage PCS and photovoltaic inverter operate collaboratively, with the energy storage PCS maintaining the VSG network structure, and the photovoltaic inverter outputting according to MPPT or dispatch commands, not participating in frequency regulation (or acting only as a PQ node); if no energy storage is provided or sunlight is insufficient, it remains shut down until GFM→GFL recovery and reconnects to the grid. Biomass synchronous units, in islanded mode, are limited by fuel supply and can operate at reduced power or shut down, not participating in the VSG network logic.

[0075] The three-layer control logic adaptation under GFM mode is as follows: ① Local control layer: Each PCS operates in a peer-to-peer mode with a droop coefficient inversely proportional to the rated capacity. The reference value is switched to local frequency and voltage (provided by the main grid unit). The virtual inertia term in adaptive droop control is replaced by the virtual rotational inertia provided by VSG to maintain system frequency stability; ② Regional coordination layer: Power-voltage sensitivity identification is maintained online. Load scheduling only retains the emergency load reduction function and no longer performs SOC equalization allocation (it is replaced by automatic equalization by peer-to-peer droop); ③ Global optimization layer: The optimization target is switched to "island autonomous operation". Priority is given to ensuring the reliability of local load power supply and the maintenance of energy storage SOC. Economic optimization is secondary. The power exchange with the main grid is forcibly set to zero, that is, the PCC switch is disconnected or the power transmission channel is blocked.

[0076] During GFM→GFL recovery, PLL pre-synchronization is used (the PCC switch is closed after a phase difference of <0.01 rad and a duration of >100 ms). During the switching process, an acknowledgment delay of ≥1 s and hysteresis logic are set (hysteresis bandwidth and SCR are both dimensionless; SCR threshold difference ≥0.5) to avoid power oscillation caused by frequent switching.

[0077] Step S5: Monitor the deviation between actual power and predicted power through the energy management system, i.e., the deviation between actual collected power and predicted power, and determine whether the deviation exceeds the threshold; wherein, the predicted power is output by the prediction model, and the predicted power includes photovoltaic output and load power; If yes, return to step S3 and correct the parameters online; otherwise, return to step S5.

[0078] Restoring hysteresis: Downgrade to restore hysteresis band settings, Level II to Level I requires... And lasts for 30 seconds; Level III to Level II requires And it lasts for 10 seconds.

[0079] EMS continuously monitors the deviation between photovoltaic output and load power, that is, the deviation between the actual power collected and the photovoltaic output and load power (predicted power). The absolute value of the deviation is then averaged over three consecutive rolling cycles (15 minutes). (The overline indicates that the average is taken over three consecutive rolling cycles; the absolute value has been de-ambiguous.) (Typical threshold) This triggers a global optimization layer re-optimization update of the current day's plan, while simultaneously using recent measured data to online correct the line impedance parameters identified by FF-RLS. The parameters of the prediction model (PV model weight coefficients, load model ARIMA coefficients) are updated according to the recursive formula described above. The PV model weight coefficients are retrained every 24 rolling cycles (6 hours) using the measured data of the most recent 7 days as samples. The load model ARIMA coefficients are updated online using recursive least squares. The weights of the multiple objectives are dynamically adjusted according to the operating conditions, where sub-objective one ( Minimize voltage deviation, sub-objective two ( The overall economic operation is optimal (including green energy consumption). Under normal operating conditions, sub-target two economic operation is the main focus. Under over-limit operating conditions, the voltage deviation weight of sub-target one is increased, and the weight combination is switched according to the S2 classification results (Level I). Level II Level III The data is fed back to the S3 policy matching layer, where the prediction model parameters are corrected and the weights are optimized before the next round of control continues.

[0080] To enable those skilled in the art to better understand this solution, the following detailed explanation is provided using a specific application scenario as an example: Taking a green electricity direct-connection industrial park as an application scenario, the park is equipped with: a total installed capacity of 20MWp distributed photovoltaic power, a 5MW / 10MWh energy storage system (lithium iron phosphate), 2×3MW biomass generator sets, adjustable loads (air conditioning / water pumps / lighting) totaling approximately 5MW of adjustable capacity, an SVG capacity of ±3MVar, and on-load tap-changing transformer parameters of ±8 steps (out of 17 steps) with a transformation ratio of 110 / 10.5kV. The grid connection point voltage level is 10kV, with a normal voltage range of 0.93~1.07pu.

[0081] Communication Architecture: The local control layer employs an independent closed-loop (10ms) DSP built into the inverter / energy storage PCS. The regional coordination layer achieves peer-to-peer communication between devices within the station via GOOSE or EtherCAT high-speed LAN (single frame <1ms), with locally deployed edge computing units aggregating data from PMU, energy storage BMS, and biomass DCS. The global optimization layer aggregates operational data from the entire region via 4G / 5G or dedicated fiber optic lines through the EMS.

[0082] Control Flow: After system startup, it continuously collects PCC three-phase voltage / current and operating data of various devices. According to... Real-time voltage level determination. Under normal circumstances, it is at Level I (…). ≤0.07pu), only local adaptive droop control operates. When the photovoltaic output suddenly drops, causing the voltage to fall to level II (0.07pu < ≤0.20pu), triggering the regional coordination layer sequentially: FF-RLS identifies impedance parameters → multi-source reactive power coordination allocation → load hierarchical scheduling, while simultaneously notifying the global optimization layer to increase voltage weight. If Prediction pu / s, even Even if the voltage is still within Level I, the corresponding control layer should be notified in advance. When the voltage deteriorates to Level III ( >0.20pu), emergency re-optimization, reducing photovoltaic output under overvoltage conditions, and fully discharging energy storage while flexibly reducing load under undervoltage conditions. According to Real-time determination of GFL / GFM mode, automatically switching to GFM in weak network or isolated situations. Monitoring deviations and triggering re-optimization to restore hysteresis control degradation.

[0083] To verify the effectiveness of the above control method, this embodiment uses a typical operating condition on a certain day as an example for simulation verification. When the grid-connected voltage rises to 1.09 pu during the midday peak photovoltaic power generation period, it is determined to be a Level II warning zone. Adaptive droop control increases the droop coefficient and instructs the SVG to absorb reactive power. The voltage drops back to below 1.04 pu within a few seconds, without entering the Level III emergency limit exceedance. When cloud cover causes a sudden drop in photovoltaic output and the voltage drops to 0.96 pu, the regional coordination layer identifies the impedance through FF-RLS and coordinates the reactive power output of energy storage and biomass units to support the voltage, which rises back to above 0.98 pu. Simulation results show that the voltage exceeds the limit multiple times in the uncontrolled scenario, and the voltage fluctuation is still large in the fixed droop coefficient scenario. However, under the control of this scheme, the grid-connected voltage is maintained within the normal range of 0.93~1.07 pu throughout the day, with the highest proportion of Level I and the proportion of Level III being 0, verifying the effectiveness of the proposed multi-timescale voltage optimization control method.

[0084] Figure 5 The red dashed line, blue dotted line, and green solid line (bold) represent the grid-connected voltage waveforms under three control modes: no control, fixed droop coefficient, and the proposed scheme, respectively. The gray dotted line represents the ±7% voltage deviation limit (0.93~1.07 pu), and the gray dotted line represents the Level III emergency over-limit threshold (1.20 pu). Under the control of this scheme, the voltage is always maintained within the limit, while significant over-limits occur in both the no-control and fixed droop coefficient scenarios.

[0085] Figure 6The voltage classification ratios of the three scenarios—no control, fixed droop coefficient, and this solution—are compared: Level I (normal fluctuation range, |ΔU PCC Level 1 (≤0.07 pu), Level 2 (Warning Zone, 0.07~0.20 pu), Level 3 (Emergency Exceeding Limits, >0.20 pu). Figure 6 In the diagram, green, orange, and red represent the proportions of Level I, Level II, and Level III, respectively. This scheme shows that Level I has the highest proportion, while Level III has zero, verifying the effectiveness of the proposed multi-time-scale voltage optimization control method.

[0086] Table 1 is a reference table for setting key parameters.

[0087] To address the aforementioned technical problems, this invention also provides a multi-time-scale voltage optimization control system for green electricity direct-connection grid points, comprising: Data acquisition module 1 is used to acquire the three-phase voltage and current of PCC, the operating data of each controllable device, the time-of-use electricity price, and the tap position signal of the transformer tap changer. Calculation module 2 is used to calculate the actual power, voltage deviation, voltage change rate, and short-circuit ratio at the grid connection point based on the voltage and current collected by data acquisition module 1. The assessment and grading module 3 is used to assess and grade the voltage situation based on the voltage deviation calculated by the calculation module 2 according to the three-level threshold, and to notify the control layer in advance of the assessment and grading results using the voltage change rate prediction mechanism. Result matching module 4 is used to match multi-timescale control strategies based on the results obtained from evaluation and grading module 3. The energy management system 5 is used to monitor the deviation between the actual power and the predicted power, and to determine whether the deviation exceeds the threshold. The predicted power is output by the prediction model and includes photovoltaic output and load power. If the deviation is true, the result matching module 4 and the online correction module are triggered. If not, the monitoring continues. Online correction module 6 is used for online correction of line impedance parameters and prediction model coefficients in FF-RLS online impedance identification. This invention provides a multi-timescale voltage optimization control method and system for green electricity direct-connection points, achieving multi-timescale integration: from 10ms-level local fast response to minute-level global economic optimization, realizing a complete closed loop of voltage control "fast response - precise adjustment - economic operation". Adaptive droop control: reactive power margin factor. and voltage change rate factor The droop coefficient adaptively adjusts with operating conditions, avoiding oscillations or insufficient response under strong fluctuations caused by traditional fixed coefficients. FF-RLS online identification + excitation detection: passively utilizes natural load / PV fluctuations to identify impedance parameters without injecting harmonic disturbances; the excitation detection mechanism prevents steady-state divergence. Quota-aware dynamic penalty: tap changer / capacitor operating costs dynamically change with remaining quotas, automatically and intelligently allocating limited daily quotas across 96 time periods (15-minute intervals) within the day, preventing quota exhaustion in the morning and lack of options in the afternoon. GFL / GFM adaptive switching: based on a comprehensive criterion of SCR, voltage deviation, and voltage-active power sensitivity, seamless grid-connected / islanded switching is achieved. dU / dt Predictive mechanism: When the voltage deviation is still within the normal range, the deterioration trend can be detected in advance by the voltage change rate, triggering the pre-scheduling strategy of the upper-level energy management system.

[0088] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the applications disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and incorporate common knowledge or customary techniques in the art disclosed herein. The specification and examples are to be considered exemplary only, and the true scope of the invention is indicated by the claims.

[0089] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention.

Claims

1. A method for multi-time-scale voltage optimization control of green electricity direct-connection grid points, characterized in that, Includes the following steps: Step S1: Collect the three-phase voltage and current of the PCC through the PMU, and calculate the actual power, voltage deviation and voltage change rate based on the voltage and current. Simultaneously collect the operating data of each controllable device and the time-of-use electricity price. Collect the tap position signal of the transformer tap changer in real time through the on-load tap-changing transformer configured at the grid connection point, and calculate the short-circuit ratio at the grid connection point based on the voltage and current. Step S2: Based on the voltage deviation, perform voltage status assessment and classification according to three threshold levels, and notify the control layer in advance of the assessment and classification results using the voltage change rate prediction mechanism; Step S3: Match multi-timescale control strategies based on the assessment and grading results; Step S4: Implement adaptive switching between GFL / GFM modes based on the short-circuit ratio composite criterion at the grid connection point; Step S5: Monitor the deviation between the actual power and the predicted power through the energy management system, and determine whether the deviation exceeds a threshold; wherein, the predicted power is output by the prediction model, and the predicted power includes photovoltaic output and load power; If yes, return to step S3 and correct the parameters online; otherwise, return to step S5.

2. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 1, characterized in that, The voltage situation assessment and classification based on the voltage deviation using a three-level threshold specifically includes: Determine whether the voltage deviation is less than or equal to 0.

07. ; If yes, it is classified as Level I, and only the local control layer operates; otherwise, it is determined whether the voltage deviation is greater than 0.

07. And less than or equal to 0.2 ; If yes, it is classified as Level II, and the regional coordination layer is triggered sequentially, and the global optimization layer is notified to increase the voltage weight. If no, it is classified as Level III, and all control regional coordination layers are activated and the global optimization layer is urgently re-optimized.

3. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 2, characterized in that, The assessment and grading results will be communicated to the control layer in advance using the voltage change rate prediction mechanism, specifically including: Determine whether the voltage change rate is greater than zero; If not, notify the regional coordination layer to pre-schedule energy storage discharge to replenish the undervoltage; if yes, notify the regional coordination layer to pre-schedule energy storage charging to suppress overvoltage, and determine whether the voltage change rate is greater than 0.

03. If yes, then even if the voltage deviation is still within Level I, the regional coordination layer must be notified to pre-schedule energy storage charging for overvoltage suppression. If no, then return to determine whether the voltage deviation is greater than 0.

07. And less than or equal to 0.2 The steps.

4. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 2, characterized in that, The specific aspects of matching multi-timescale control strategies based on assessment and grading results include: When the condition is determined to be Level I, local adaptive droop control is executed; When the classification is Level II, FF-RLS online impedance identification and multi-source reactive power coordination are triggered. When the level is determined to be Level III, the entire control area coordination layer is activated and the global optimization layer is triggered for emergency re-optimization. The global optimization layer adopts a two-stage optimization using a day-ahead mixed integer quadratic programming scheduling model and an intraday nonlinear programming rolling correction model.

5. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 4, characterized in that, The global optimization layer employs a two-stage optimization approach, combining a day-ahead mixed-integer quadratic programming scheduling model and an intraday nonlinear programming rolling correction model. Specifically, this includes: The day-ahead mixed integer quadratic programming scheduling model optimizes the baseline curves of discrete and continuous variables for the next 24 hours with a granularity of 15 minutes. The discrete variables include the Tap position signal and the capacitor switching status. The intraday nonlinear programming rolling correction model optimizes the continuous variables for the next 4 hours with a granularity of 5 minutes. It achieves closed-loop control through model prediction and rolling time domain, advances the window every 5 minutes, and executes only the first time period instruction.

6. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 4, characterized in that, In the local adaptive droop control, the total reactive power reference value of the inverter is determined according to the following formula: in, The adaptive coefficient is the steady-state reactive power baseline value issued by the global optimization layer. , The reactive power margin factor, Voltage rate of change factor, base droop coefficient , The rated capacity of the inverter, U ref =1p.u, This represents the maximum controllable voltage deviation range of the local control layer. For voltage deviation, J u This is the virtual voltage inertia coefficient. This represents the rate of change of voltage.

7. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 1, characterized in that, The adaptive switching of GFL / GFM mode based on the short-circuit ratio composite criterion at the grid connection point specifically includes: The composite criterion for the short-circuit ratio at the grid connection point is: in, This is the system equivalent voltage at PCC. P rated For the rated capacity of grid-connected equipment, The equivalent impedance from the PCC to the system's equivalent power supply is obtained through the FF-RLS online impedance identification. For GFM switching threshold, The GFL recovery threshold; when Or when PCC disconnects, it triggers GFL→GFM switching. At that time, GFM→GFL recovery is triggered. The value is 3.

0. The value is 4.5, and the hysteresis width is ; When in the hysteresis interval At that time, maintain the original mode.

8. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 4, characterized in that, The online correction parameters specifically include: The global optimization layer is triggered to update the daily plan, and the line impedance parameters of the FF-RLS online impedance identification and the prediction model coefficients are corrected online.

9. The multi-time-scale voltage optimization control method for green electricity direct-connection grid points according to claim 1, characterized in that, The controllable equipment includes photovoltaic, energy storage, load, and reactive power compensation equipment.

10. A multi-time-scale voltage optimization control system for green electricity direct-connection grid points, characterized in that, include: The data acquisition module (1) is used to acquire the three-phase voltage and current of the PCC, the operating data of each controllable device, the time-of-use electricity price, and the tap position signal of the transformer tap changer. The calculation module (2) is used to calculate the actual power, voltage deviation, voltage change rate and grid connection point short-circuit ratio based on the voltage and current collected by the data acquisition module (1). The assessment and grading module (3) is used to assess and grade the voltage situation according to the voltage deviation calculated by the calculation module (2) based on the three-level threshold, and to notify the control layer in advance of the assessment and grading results using the voltage change rate prediction mechanism. The result matching module (4) is used to match multi-timescale control strategies based on the results obtained from the evaluation and grading module (3); The energy management system (5) is used to monitor the deviation between the actual power and the predicted power, and to determine whether the deviation exceeds the threshold. The predicted power is output by the prediction model and includes photovoltaic output and load power. If the deviation is exceeded, the result matching module (4) and the online correction module (6) are triggered. If the deviation is exceeded, the monitoring continues. The online correction module (6) is used to correct the line impedance parameters and prediction model coefficients of FF-RLS online impedance identification.

Images