Safety verification and correction method and system for power regulation of optical storage and charging station

Abstract

The present application belongs to the field of electric power automation and energy control, and particularly relates to a safety verification and correction method and system for power regulation of a light storage and charging station. The method comprises creating a joint simulation model for simulating the state of each object in the charging station and its interaction; obtaining operation data in a future control window and generating initial control instructions through an upper optimization algorithm; performing a rolling time sequence deduction on the charging station operating environment after the initial control instructions are executed through the joint simulation model, extracting safety indicators in the deduction data and determining whether a safety limit is triggered; generating a control mode and calculating error indicators to generate an adaptive correction step for the instructions that may exceed the limit according to the error type, and performing iterative optimization; providing a bottom measure if the optimization fails, thereby ensuring that the issued instructions meet the safety constraints. The present application solves the problem that the control of the light storage and charging station is difficult, and existing strategies cannot guarantee system safety and green electricity consumption efficiency.

Description

Technical Field

[0001] This invention belongs to the field of power automation and energy control, specifically relating to a safety verification and correction method for power regulation of photovoltaic energy storage charging stations and the corresponding control system for photovoltaic energy storage charging stations. Background Technology

[0002] A photovoltaic-storage-charging station is a new type of charging station that combines photovoltaic power generation and battery energy storage capabilities, and provides charging services through charging piles. Currently, the control of photovoltaic-storage-charging stations typically adopts an open-loop model of "upper-level optimization calculation + lower-level direct execution." The upper-level system calculates the optimal control strategy based on predicted data such as photovoltaic output and load demand, and directly sends it to the lower-level equipment for execution. Some systems introduce static power flow verification or simple rule-based judgment, but overall, it still belongs to an open-loop decision-making and direct execution mode.

[0003] Traditional photovoltaic-storage-charging station management models suffer from the following significant drawbacks: 1. Lack of pre-emptive risk assessment: Control commands are not dynamically simulated, making it impossible to detect operational risks such as voltage exceeding limits and transformer overload in advance. 2. Delayed correction: Load is passively cut off or adjusted only after a fault occurs, impacting user experience and equipment lifespan. 3. Neglecting strategy feasibility: V2G control relies on idealized performance assumptions; in actual execution, power shortages may occur due to user defaults, and the system cannot identify such risks before issuing commands. 4. Poor model update robustness: When calibrating models using historical data, they are susceptible to interference from sensor anomalies, leading to model parameter drift and affecting subsequent control accuracy. These problems are particularly prominent in scenarios with high electric vehicle (EV) penetration and high EV concurrency, severely restricting the system safety and green energy consumption efficiency of photovoltaic-storage-charging stations. Summary of the Invention

[0004] To address the problem that existing management methods cannot effectively evaluate the control effects of control commands, resulting in high control difficulty for photovoltaic-storage-charging stations and hindering system safety and green energy consumption efficiency, this invention provides a safety verification and correction method for power regulation of photovoltaic-storage-charging stations and a corresponding control system for such stations.

[0005] This invention is achieved using the following technical solution: A method for safety verification and correction of power regulation in a photovoltaic-storage-charging station includes: Create a co-simulation model to simulate the state and interactions of various objects within a charging station. This includes: a photovoltaic output model to predict photovoltaic power output; an energy storage response model to predict the charging and discharging power and energy storage efficiency of energy storage devices; an EV cluster behavior model to predict the charging and discharging demand and performance of vehicles entering the station; and a power flow model to predict the voltage of each node in the distribution network, branch current, and transformer load rate.

[0006] The system obtains the predicted photovoltaic power generation, base load, and charging demand within the future control window, and generates initial control commands through an upper-level optimization algorithm. The generated control commands include: reactive power of the photovoltaic inverter, energy storage charging and discharging power, and total charging power of the electric vehicle cluster. The values ​​of each indicator in the control commands are only valid within the control window; therefore, the command content usually also includes the corresponding valid time period.

[0007] By using a co-simulation model to perform rolling time-series simulations of the charging station's operating environment after the initial control commands, indicators related to system safety are extracted from the trajectory data generated during the simulation process, thereby determining whether a safety limit violation has been triggered. (1) If so, first select the control method according to the error type; then calculate the error index by weighted fusion of the over-limit depth of the preset constraints, and generate an adaptive correction step size based on the linear model with the error index as the variable; then iteratively optimize the instruction content of the initial control instruction with the correction step size as the adjustment amount of each round, and issue an updated control instruction that meets the requirements. (2) If not, issue the initial control command directly.

[0008] As a further improvement to this invention, the photovoltaic power output model adopts a single-diode equivalent circuit model based on physical mechanisms. The photovoltaic power output model combines on-site measured data of irradiance and temperature, and simulates the actual power output of the photovoltaic array using a maximum power point tracking algorithm, thereby predicting the power generation at future times.

[0009] As a further improvement to this invention, the energy storage response model is constructed based on the Thevenin equivalent circuit model. The energy storage response model predicts the output charging and discharging power and energy storage efficiency of the energy storage device at future times based on changes in the battery's state of charge, health status, and ambient temperature.

[0010] As a further improvement of the present invention, the EV cluster behavior model models each electric vehicle as an independent intelligent agent, and each intelligent agent has a personalized credibility score, departure time, battery capacity and maximum charging and discharging power; after pre-setting the probability distribution of user response delay and default probability, the behavior of each intelligent agent in the cluster at future moments is generated based on Monte Carlo simulation.

[0011] As a further improvement of the present invention, the power flow model of the distribution network is based on the topology of the standard distribution network test system and uses the forward-backward substitution method to calculate the three-phase power flow; and is used to output the voltage of each node, branch current and transformer load rate at future time.

[0012] As a further improvement of the present invention, the actual operating state of the charging station in the control window after each control command is issued is obtained and compared with the state at the corresponding time derived by the joint simulation model to evaluate the prediction accuracy of the joint simulation model; when the prediction accuracy is lower than the preset accuracy threshold, the update mechanism of the joint simulation model is triggered.

[0013] As a further improvement of this invention, when updating the co-simulation model, the hyperparameters of the photovoltaic power output model include the attenuation coefficient; the updating method is as follows: ; In the above formula, and These represent the decay coefficients before and after the update, respectively; This represents the photovoltaic power generation predicted by the model; This indicates the actual power generation at the corresponding moment.

[0014] As a further improvement of this invention, the hyperparameters of the energy storage response model include the energy storage efficiency and health of the energy storage device. The energy storage efficiency is updated as follows: if the deviation between the predicted and actual energy storage efficiency exceeds a deviation threshold, the predicted energy storage efficiency is replaced with the actual energy storage efficiency. The health is updated as follows: a linear decay model is used to update the health of the energy storage device based on the actual cumulative equivalent full cycle count of the energy storage system.

[0015] As a further improvement of this invention, the hyperparameters of the EV cluster behavior model include the credibility score of the agent corresponding to each user. R i If any user fails to discharge as instructed or the response delay exceeds a preset threshold, the credibility score of the corresponding agent will be reduced. R i Decrease the specified value.

[0016] As a further improvement to this invention, to enhance the derivation efficiency of the co-simulation model, the power flow model of the distribution network employs a voltage-power sensitivity matrix to perform power flow calculations. The voltage-power sensitivity matrix is ​​dynamically updated at the beginning of each control cycle; and when a change in network topology is detected or the power fluctuation of a critical branch exceeds a preset threshold, the matrix is ​​recalculated.

[0017] As a further improvement of this invention, in order to improve the derivation efficiency of the co-simulation model, the event step size is dynamically adjusted when the co-simulation model performs rolling timing derivation. In the initial stage, the default maximum event step size is used to perform coarse-grained derivation of the control window. When the power command change rate is detected to exceed a preset threshold or the simulation index enters a preset warning area, the event step size is shortened to perform fine-grained derivation of the control window.

[0018] As a further improvement of the present invention, the constraint conditions for determining whether a photovoltaic-energy storage charging pile has exceeded safety limits include: Node voltage constraint: V node (t) ∈ [V min V max ].

[0019] Transformer load factor constraint: L trafo (t) < L limit .

[0020] Energy storage SOC constraint: SOC(t) ∈ [SOC min SOC max ].

[0021] Photovoltaic reactive power constraint: Q pv (t) ∈ [Q min Q max ] .

[0022] Power factor constraint: P F (t) ∈ [P Fmin P Fmax ].

[0023] If any constraint is violated, a safety overrun is determined to be triggered.

[0024] In the above formula, V node (t) represents the real-time node voltage; V min and V max These are the preset lower and upper limits of the node voltage; L trafo (t) represents the real-time load rate of the transformer; L limit The threshold representing the transformer load factor; SOC(t) represents the real-time charge of the energy storage device; SOC min and SOC max The preset lower and upper limits of the charge capacity of the energy storage device; Q pv (t) represents the real-time reactive power of the photovoltaic device; Q min and Q max This indicates the lower and upper limits of the reactive power of photovoltaic equipment; P F (t) represents the real-time power factor of the distribution network; P Fmin and P Fmax This indicates the lower and upper limits of the power factor in the distribution network.

[0025] As a further improvement of the present invention, the error types include: voltage below the lower limit, voltage above the upper limit, transformer overload, transformer reverse overload, energy storage SOC too high, and energy storage SOC too low.

[0026] As a further improvement of the present invention, the method for generating control methods based on error type includes adjusting the adjustment direction and resource call priority, and the specific strategies are as follows: (i) When the voltage is below the lower limit, establish the adjustment direction of "prioritizing reactive power support, supplementing active power discharge, and ensuring load reduction".

[0027] The system prioritizes instructing the photovoltaic inverter and energy storage to generate capacitive reactive power and increase the energy storage discharge power. After the voltage transient trend stabilizes, the system increases the discharge power of the EV to gradually replace the energy storage output and / or the discharge power of the energy storage device, thereby reducing energy storage losses. If the limit cannot be eliminated, the system finally reduces or cuts off the charging power of some EVs and increases the reactive power output of the photovoltaic system in ascending order of reliability.

[0028] (ii) When the voltage is higher than the upper limit, the adjustment direction is to prioritize reactive power absorption, active power charging and consumption, and ensure the curtailment of photovoltaic power.

[0029] First, instruct the photovoltaic inverter and energy storage to absorb inductive reactive power and increase the energy storage charging power; after the voltage transient trend stabilizes, maximize the EV charging power; if the energy storage is full and there is no load to accept, then finally execute the curtailment strategy to limit the photovoltaic active power output.

[0030] (iii) When the transformer is overloaded, establish the adjustment direction of "rapid peak shaving of energy storage, collaborative sharing of EV load, and orderly load shedding".

[0031] The system prioritizes instructing the energy storage system to increase its discharge power; after the load pressure is initially relieved, the EVs are called in to discharge; if the total load still exceeds the limit, the charging power of some EVs is reduced or cut off in ascending order of reliability.

[0032] (iv) When the transformer is overloaded in reverse, establish the adjustment direction of "full absorption of energy storage, consumption of EV expansion, and minimum protection of photovoltaic power generation".

[0033] First, instruct the energy storage system to increase its charging power; after the reverse power flow is controlled, increase the EV charging power; if the energy storage is full and the charging load has reached its limit, then finally forcibly limit the grid-connected active power of the photovoltaic inverter.

[0034] (v) When the SOC is too low, establish the adjustment direction of "blocking discharge, switching to reactive power support, and prioritizing recharge".

[0035] Immediately prohibit energy storage discharge commands and control it to switch to reactive power support mode. At the same time, reduce the charging load in the station or use the off-peak hours of the power grid for rapid recharging. If it still cannot be restored to a safe range, temporarily take the energy storage system offline and readjust it according to the current actual achievable efficiency and available capacity of the energy storage.

[0036] (vi) When the SOC is too high, establish the adjustment direction of "blocking charging, switching to reactive power support, and accelerating discharge".

[0037] Immediately prohibit energy storage charging commands and control them to switch to reactive power support mode. At the same time, actively increase EV charging power or guide EV discharge to accelerate the consumption of surplus power in the station. In extreme cases, link with photovoltaic curtailment.

[0038] (vii) Global constraint verification mechanism: When executing any strategy involving photovoltaic reactive power regulation, the system verifies in real time whether the photovoltaic reactive power and power factor meet the preset constraints; if the calculated command causes the photovoltaic reactive power and power factor to exceed the limits, the original command issuance is immediately suspended, and the constraint saturation and multi-source compensation strategy is executed instead, including: First, the photovoltaic reactive power command is clamped to the current maximum allowable safety boundary, and the resulting reactive power deficit is calculated. Then, the remaining reactive power capacity of the energy storage system and EVs is used for compensation first. If the reactive power resources of the entire site are still insufficient, the active power-reactive power replacement mechanism is activated to reduce the active power output of the photovoltaic system to release reactive power capacity until all constraints are met simultaneously.

[0039] As a further improvement to the present invention, the step size is corrected. K n The calculation formula is:

[0040] In the above formula, A and A 0 represents the base adjustment amount and preset gain of the corresponding indicator, respectively; Indicates the first i The depth at which a constraint exceeds its limit; value i Indicates the first i The predicted value of the indicator corresponding to the constraint; Indicates the first i The one-sided threshold closest to the predicted value in the constraints; the over-limit states of voltage, transformer load, and energy storage SOC when i=1, 2, 3; severity Indicates the error index.

[0041] As a further improvement of the present invention, the process of iteratively optimizing the instruction content of the initial control command includes: S1: Analyze the error types included in the current safety over-limit state and select the corresponding control method.

[0042] S2: For over-limit constraints such as voltage, transformer load rate, and energy storage SOC that require iterative correction, based on the predicted values ​​of the corresponding indicators for each constraint. value i Calculate the depth of exceeding the limit for each of the corresponding safety thresholds. Error index severity and correction step size K n .

[0043] S3; Determine the adjustment direction based on the control method, use the correction step size as the adjustment amount, generate updated values ​​for various indicators, and generate new control commands.

[0044] S4: Use a co-simulation model to perform rolling timing deduction of the new control commands, and determine whether the safety limit violation state has been lifted based on the trajectory data generated during the deduction process. (1) If so, the iteration ends and the optimized control command is output.

[0045] (2) If not, continue to determine whether the maximum number of iterations has been reached: If the target has been reached, a final control command is generated through a safety net mechanism; otherwise, the process returns to step S3 to generate a new control command.

[0046] As a further improvement to this invention, the backup mechanism first performs a first-level response, including: forcibly disconnecting the charging connection of all EVs with a reliability score below the safety threshold; switching the energy storage system to a "voltage / reactive power priority support mode"; limiting the active power output of the photovoltaic system to reserve reactive power regulation capacity; and determining whether the safety over-limit has been resolved: if yes, a final control command corresponding to the first-level response is generated; otherwise, a second-level response is performed, and a final control command corresponding to the second-level response is generated. The second-level response includes disconnecting non-critical loads, and if the risk of transformer reverse overload or frequency anomaly has not been eliminated, switching the energy storage system control mode to a "grid-type V / f control mode" to establish local voltage and frequency support, or executing a full-site emergency shutdown protection under extreme overload conditions.

[0047] This invention also includes a control system for a photovoltaic-storage-charging station, which employs the aforementioned safety verification and correction method for power regulation in photovoltaic-storage-charging stations to generate dynamically updated control commands based on the real-time operating status of the charging station, thereby maintaining the stable operation of the charging station. The control system for the photovoltaic-storage-charging station includes: a strategy generation module, a charging station twin module, a decision-making module, and a model update module.

[0048] The strategy generation module generates initial control commands based on the real-time operating status of the photovoltaic-storage-charging station. A dynamically updated co-simulation model runs within the charging station twin module. This model jointly simulates the states and interactions of the photovoltaic equipment, energy storage equipment, EV clusters, and distribution network within the charging station, enabling rolling time-series projections of the charging station's future operating status based on its current state. The decision module uses the charging station twin module to predict whether any safety limits will be exceeded after adopting arbitrary control commands. If so, it first selects a control method based on the error type; then, it weights and fuses the exceedance depths of preset constraints to calculate an error index, and uses this error index as a variable to generate an adaptive correction step size based on a linear model. Finally, it iteratively optimizes the initial control command content using the correction step size as the adjustment amount for each round, and issues the updated control command. Otherwise, the control command is issued to the designated object for execution. The model update module is used to calculate the prediction accuracy of the co-simulation model after each round of control command is issued, based on the actual operating status of the charging station and the predicted status output by the charging station twin module. When the prediction accuracy threshold is lower than the preset value, the hyperparameters of the co-simulation model are updated.

[0049] The technical solution provided by this invention has the following beneficial effects: Unlike traditional open-loop control, the technical solution provided by this invention injects candidate strategies into a joint simulation model that includes a four-dimensional model of photovoltaics, energy storage, EV clusters, and distribution networks for rolling time-series deduction before the control command is officially executed. This achieves the technical effect of "quasi-closed-loop control". By identifying risks such as voltage over-limit and transformer overload in advance, it avoids equipment tripping and grid accidents from the source and effectively ensures the safety and reliability of the optimized control command.

[0050] This invention innovatively proposes an adaptive step size correction strategy based on the depth of the limit exceedance when iteratively optimizing control commands. It realizes the calculation of the comprehensive depth of the limit exceedance based on simulation results and then dynamically determines the optimal correction step size coefficient. This enables fast regression with large step sizes when the limit exceedance is severe and fine-tuning with small step sizes when approaching the boundary. It solves the problems of slow convergence, oscillation or dead loop caused by fixed step sizes.

[0051] This invention also introduces a three-stage graded degradation fallback strategy in the system control process: constructing a three-level defense system of "normal correction → first-level response (cutting low credibility EV) → second-level response (island security protection)". After the first-level response, the simulation is re-verified. The second-level degradation is based on the microgrid island operation principle and is confirmed as an inherent safe state for direct output, ensuring that the system can still output absolutely safe conservative instructions under the condition that the algorithm cannot converge or extreme failure.

[0052] This invention also designs a lightweight event-driven simulation acceleration mechanism in the application stage of the co-simulation model: on the one hand, it uses a voltage-power sensitivity matrix to replace nonlinear power flow calculation for linearization; on the other hand, it combines a dynamic time step strategy (large step size for coarse sweep by default, and dense step size only when power changes suddenly or the index approaches the threshold) to significantly reduce the computational load and meet the requirements of short-cycle real-time control.

[0053] This invention also effectively ensures the prediction accuracy of the co-simulation model through a parameter update mechanism. This strategy uses statistical methods (such as the 3σ principle) to remove abnormal data from real operation and maintenance data for the purpose of evaluating the accuracy of the prediction data derived from the model. Then, the model hyperparameters are dynamically adjusted according to the historical error distribution to realize online learning and optimization of the model, effectively isolate sensor noise from contaminating the simulation accuracy, and improve the robustness of the system in long-term operation. Attached Figure Description

[0054] Figure 1 This is a flowchart of the safety verification and correction method for power regulation of photovoltaic energy storage charging station provided in Embodiment 1 of the present invention.

[0055] Figure 2 This is a logic block diagram of the iterative optimization of the instruction content of the control instruction in Embodiment 1 of the present invention.

[0056] Figure 3 This is a system architecture diagram of the control system for the photovoltaic energy storage charging station provided in Embodiment 2 of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] Example 1 To address the aforementioned problems existing in the operation of current photovoltaic-storage-charging stations, this embodiment provides a safety verification and correction method for power regulation in photovoltaic-storage-charging stations. This method ensures the reliability of regulation quality by performing a high-fidelity simulation of the potential regulation effects before issuing commands. Furthermore, when risks exist during the simulation, an adaptive iterative correction strategy is used to improve the effect of each generated command, ensuring strategy safety. In addition, under extreme operating conditions, this embodiment also sets a graded response mode with a fallback effect for the regulation strategy corresponding to the command. During the application phase, a lightweight model option is set to ensure the efficiency and real-time performance of the simulation process, and the co-simulation model used in the inference process is evaluated and updated to improve the quality of the simulation results.

[0059] Specifically, such as Figure 1As shown, the safety verification and correction method for power regulation of photovoltaic energy storage charging stations provided in this embodiment includes the following process: 1. Create a co-simulation model to simulate the state and interactions of various objects within the charging station.

[0060] In this embodiment, to pre-evaluate the potential regulatory effects of the generated control commands, a co-simulation model is designed. This co-simulation model is essentially a digital twin system. On one hand, it can realistically simulate the state of the photovoltaic power generation system, energy storage devices, and charging piles within the charging station, as well as the interaction processes between these devices. On the other hand, it can predict the dynamic changes in charging demand from vehicles entering the station in the near future, as well as the fulfillment of different charging vehicle demands; thus, it can simulate the interaction behavior between vehicles entering the station and the charging station. Based on this, the co-simulation model is initialized with the real-time operation and maintenance status of the charging station, and then specified control commands are input. This allows for a rough prediction of future changes in the relevant environment through rolling time-series simulations.

[0061] In practical applications, this embodiment constructs four sub-models related to photovoltaic equipment, energy storage equipment, power distribution network, and electric vehicles, and sets the interaction methods between different sub-models to build the required co-simulation model. Therefore, the co-simulation model of this embodiment includes the following components: (1) Photovoltaic power output model for predicting photovoltaic output. The photovoltaic power output model in this embodiment adopts a single diode equivalent circuit model based on physical mechanisms. The photovoltaic power output model combines on-site measured data of irradiance and temperature, and simulates the actual output of the photovoltaic array through the maximum power point tracking algorithm (MPPT) to predict the power generation P at future times. pv The output power value is valid for a specified time period. The hyperparameters of the photovoltaic output model include the attenuation coefficient. The initial value of the attenuation coefficient can be set according to the nominal parameters on the photovoltaic module's nameplate and can be dynamically updated based on the equipment's historical operating data.

[0062] (2) Energy storage response model for predicting the charging and discharging power and energy storage efficiency of the energy storage device. The energy storage response model in this embodiment is constructed based on the Thevenin equivalent circuit model. The energy storage response model predicts the output charging and discharging power P of the energy storage device at future times based on the changes in the battery state of charge (SOC), state of health (SOH), and ambient temperature of the energy storage device. essAnd the energy storage efficiency η, similarly, the output data is valid for a specified period of time. The energy storage response model incorporates the influence of ambient temperature on the charging and discharging efficiency of the equipment; in practical applications, model parameters (such as internal resistance and open-circuit voltage curves) are identified through hybrid pulse power characteristic (HPPC) test data and are periodically calibrated based on measured data during operation.

[0063] (3) EV cluster behavior model for predicting the charging and discharging demand and performance status of vehicles entering the station. Unlike photovoltaic equipment and energy storage equipment, which are fixed devices, vehicles entering the station are dynamic and uncertain. Therefore, in the EV cluster behavior model, this embodiment adopts a multi-agent modeling method, modeling each electric vehicle as an independent agent, and each agent has a personalized credibility score R. i Attributes such as departure time, battery capacity, and maximum charge / discharge power are included. Each vehicle's credibility score corresponds to the user's historical fulfillment rate and is parameter normalized, with a value range of [0, 1]. Newly registered vehicles can have their R... i =0.9. This embodiment predetermines the probability distribution of each user's response delay and the probability of default, and then uses Monte Carlo simulation to generate the behavior of each agent in the cluster at future moments.

[0064] In practical applications, the response latency of each agent in the EV cluster behavior model adopts a log-normal distribution. For example, it is fitted based on the annual operating data of a specific photovoltaic-storage-charging station in a given year, determining that the latency conforms to a mean μ=2.5 and a variance σ=0.8. Of course, in other implementations, the parameters can also be refitted and dynamically distributed based on historical data from actual sites. Furthermore, this embodiment can also introduce a communication latency module to superimpose a fixed communication latency. Its typical value is set to 2 seconds to simulate the transmission time of commands from the upper-level system to the charging pile. During the simulation, the actual response power P of the EV cluster is... ev for: ; In the above formula, t0 represents the time when the instruction is issued. Indicates response delay, C avail This indicates the available output power.

[0065] (4) Distribution network power flow model for predicting the voltage of each node, branch current and transformer load rate of the distribution network. In this embodiment, the distribution network power flow model is based on the topology of a standard distribution network test system (which can be an IEEE 13-node system) and uses the forward-backward substitution method to calculate the three-phase power flow; and is used to output the voltage V of each node at future time. node Branch current and transformer load factor L trafo In practical applications, the parameters of the model access point (such as transformer capacity and line impedance) are set according to the actual engineering drawings.

[0066] This embodiment utilizes the four sub-models described above to construct a co-simulation environment. In this practical application, this embodiment does not impose limitations on the tools used to implement the simulation environment; for example, each sub-model can be integrated within the MATLAB / Simulink platform and interact with the control system in real time via a TCP / IP interface. In other embodiments, the co-simulation environment can be a C++ / Python code module embedded within the charging station system, or it can be a standalone real-time simulation server. Communication between modules can be via TCP / IP, UDP, shared memory, or internal function calls.

[0067] Second, obtain the predicted photovoltaic power generation, base load and charging demand within the future control window, and generate initial control commands through upper-level optimization algorithms.

[0068] In this embodiment, the charging station system operates with a fixed control cycle T. cycle The operation status of the charging station is continuously controlled. The control cycle can be flexibly preset according to actual engineering needs, for example, set to 1-15 minutes. The control instructions generated by the upper-level optimization algorithm include: reactive power of the photovoltaic inverter, charging and discharging power of energy storage, and total charging power of the electric vehicle cluster. The values ​​of each indicator in the control instructions are only valid within the control window, so the instruction content usually also includes the corresponding valid time period.

[0069] Specifically, the initial control commands generated in this embodiment will not be issued directly, but will be issued after evaluation by the co-simulation model.

[0070] Third, a rolling time-series simulation of the charging station's operating environment after the initial control command is performed using a joint simulation model, and indicators related to the evaluation of system safety are extracted from the trajectory data generated during the simulation process.

[0071] In practical applications, for each control cycle, the system executes a rolling simulation with a specified future time window Δt (e.g., 15–60 minutes). Assuming the current time is T, the simulation time axis is t ∈ [T, T+Δt], where the step size of the future time window Δt can be dynamically adjusted. This rolling time-domain mechanism of "long-window simulation, short-cycle execution" can both detect long-term risks in advance and avoid strategy failure due to the accumulation of prediction errors.

[0072] This embodiment, through rolling simulation of the joint simulation model, can obtain the charging station operation data at various times within the future time window. The indicators related to evaluating system safety extracted from the operation data at each time t in this embodiment include: node voltage V. node (t), Transformer load rate L trafo(t), the charge of the energy storage device SOC(t), and the reactive power Q of the photovoltaic device. pv (t), power factor P F (t).

[0073] Fourth, based on the simulation data from the joint simulation model, assess whether the charging station will trigger a safety overrun within a specified future time window after the current control command is implemented, and then make corresponding decisions.

[0074] The constraints for determining whether a photovoltaic-storage charging pile has exceeded safety limits include: Node voltage constraint: V node (t) ∈ [V min V max ].

[0075] Transformer load factor constraint: L trafo (t) < L limit .

[0076] Energy storage SOC constraint: SOC(t) ∈ [SOC min SOC max ].

[0077] Photovoltaic reactive power constraint: Q pv (t) ∈ [Q min Q max ] .

[0078] Power factor constraint: P F (t) ∈ [P Fmin P Fmax ].

[0079] In the above formula, V node (t) represents the real-time node voltage; V min and V max These are the preset lower and upper limits of the node voltage; typical values ​​for both can be set to 0.95 and 1.05 pu, respectively. trafo (t) represents the real-time load rate of the transformer; L limit The threshold value represents the transformer load rate; a typical value can be set to 90%. SOC(t) represents the real-time charge of the energy storage device; SOC min and SOC max These are the preset lower and upper limits of charge for the energy storage device; typical values ​​can be set to 10% and 90%. Q pv (t) represents the real-time reactive power of the photovoltaic device; Q min and Q max Indicates the lower and upper limits of reactive power of photovoltaic equipment; , among which, S maxThe rated apparent power of the photovoltaic inverter is determined by the parameters on the equipment nameplate; P pv P represents the photovoltaic power generation capacity. F (t) represents the real-time power factor of the distribution network; P Fmin and P Fmax This represents the lower and upper limits of the power factor in the distribution network, with typical values ​​of 0.9 and 1.0.

[0080] Among the five constraints mentioned above, if any one constraint is violated, a safety limit violation is determined to be triggered; if all constraints are met, the control command is determined to be compliant.

[0081] Therefore, this embodiment makes the following decision based on the determination result of whether the control command triggers a safety limit violation: (1) If a safety limit violation is triggered, a control mode is generated according to the error type and the error index is calculated to generate an adaptive correction step size. Then, the instruction content of the initial control instruction is iteratively optimized, and an updated control instruction that meets the requirements is issued. (2) If not, issue the initial control command directly.

[0082] In this embodiment, as Figure 2 As shown, the process of iteratively optimizing the instruction content of the initial control command is as follows: S1: Analyze the error types included in the current safety over-limit state and select the corresponding control method.

[0083] In this embodiment, based on the aforementioned constraints, the determined error types generally include: voltage below the lower limit, voltage above the upper limit, transformer overload, transformer reverse overload, energy storage SOC too high, and energy storage SOC too low.

[0084] Different error types reflect different risk states of the charging station, thus requiring different control methods. Specifically, this embodiment's method for generating control methods based on error type includes adjusting the direction of adjustment and resource allocation priority, with the specific strategies as follows: (i) When the voltage is below the lower limit, establish the adjustment direction of "prioritizing reactive power support, supplementing active power discharge, and ensuring load reduction".

[0085] The system prioritizes instructing the photovoltaic inverter and energy storage to generate capacitive reactive power and increase the energy storage discharge power. After the voltage transient trend stabilizes, the discharge power of the EV is increased to gradually replace the energy storage output and / or the discharge power of the energy storage device, reducing energy storage losses. If exceeding the limit still cannot be eliminated, the charging power of some EVs is reduced or cut off in ascending order of reliability, and the reactive power output of the photovoltaic is increased until the lower limit is reached.

[0086] (ii) When the voltage is higher than the upper limit, the adjustment direction is to prioritize reactive power absorption, active power charging and consumption, and ensure the curtailment of photovoltaic power.

[0087] First, instruct the photovoltaic inverter and energy storage to absorb inductive reactive power and increase the energy storage charging power. After the voltage transient trend stabilizes, maximize the EV charging power. If the energy storage is full and there is no load to accept, then finally execute the curtailment strategy to limit the photovoltaic active power output.

[0088] (iii) When the transformer is overloaded, establish the adjustment direction of "rapid peak shaving of energy storage, collaborative sharing of EV load, and orderly load shedding".

[0089] The system prioritizes instructing the energy storage system to increase its discharge power. After the load pressure is initially relieved, the EVs are called in to discharge. If the total load still exceeds the limit, the charging power of some EVs is reduced or cut off in ascending order of reliability, thereby reducing the load rate of the transformer and preventing equipment damage.

[0090] (iv) When the transformer is overloaded in reverse, establish the adjustment direction of "full absorption of energy storage, consumption of EV expansion, and minimum protection of photovoltaic power generation".

[0091] First, instruct the energy storage system to increase its charging power. After the reverse power flow is controlled, increase the EV charging power. If the energy storage is full and the charging load has reached its limit, then finally, forcibly limit the grid-connected active power of the photovoltaic inverter.

[0092] (v) When the SOC is too low, establish the adjustment direction of "blocking discharge, switching to reactive power support, and prioritizing recharge".

[0093] Immediately prohibit energy storage discharge commands and control it to switch to reactive power support mode. At the same time, reduce the charging load in the station or use the off-peak hours of the power grid for rapid recharging. If it still cannot be restored to a safe range, temporarily take the energy storage system offline and readjust it according to the current actual achievable efficiency and available capacity of the energy storage.

[0094] (vi) When the SOC is too high, establish the adjustment direction of "blocking charging, switching to reactive power support, and accelerating discharge".

[0095] Immediately prohibit energy storage charging commands and control them to switch to reactive power support mode. At the same time, actively increase EV charging power or guide EV discharge to accelerate the consumption of surplus power in the station. In extreme cases, link with photovoltaic curtailment.

[0096] (vii) Global constraint verification mechanism: When executing any strategy involving photovoltaic reactive power regulation, the photovoltaic reactive power and power factor are verified in real time to see if they meet the preset constraints. If the calculated command causes the photovoltaic reactive power and power factor to exceed the limit, the original command issuance is immediately suspended, and the constraint saturation and multi-source compensation strategy is executed instead. This includes: first, clamping the photovoltaic reactive power command to the current maximum allowable safety boundary and calculating the resulting reactive power deficit, and then prioritizing the use of the remaining reactive power capacity of the energy storage system and EV for compensation. If the reactive power resources of the entire station are still insufficient, the active-reactive power replacement mechanism is activated to reduce the photovoltaic active power output to release reactive power capacity until all constraints are met simultaneously.

[0097] S2: For over-limit constraints such as voltage, transformer load rate, and energy storage SOC that require iterative correction, based on the predicted values ​​of the corresponding indicators for each constraint. value i Calculate the depth of exceeding the limit for each of the corresponding safety thresholds. Error index severity and correction step size K n .

[0098] In this embodiment, the correction step size K n This refers to the minimum adjustment amount corresponding to any indicator in each round of control command iteration optimization. Generally speaking, the greater the deviation of the current state of the charging station from a stable and safe state, the larger the adjustment should be. Conversely, if the current state is close to a stable and safe state, the adjustment amount of each indicator in the control command should be smaller, achieving dynamic and fine adjustment. Therefore, in the practical application of this embodiment, the correction step size... K n The calculation formula is: ; In the above formula, A and A 0 represents the base adjustment amount and preset gain of the corresponding indicator, respectively; severity Indicates the error index. Indicates the first i The depth at which a constraint exceeds its limit; value i Indicates the first i The predicted value of the indicator corresponding to the constraint; Indicates the first i The one-sided threshold that is closest to the predicted value in the constraints; for example, V min and V max The values ​​are 0.95 and 1.05 pu, respectively; when V node (t) = 0.90pu, at which point it is closer to V. min ,Right now =Vmin =0.90 pu. Conversely, when V node (t) = 1.08pu, at which point it is closer to V. max ,Right now = V max =1.05 pu.

[0099] As can be seen from the above formula, the more items that violate the constraints under any working condition, or the greater the deviation of the actual indicators from the preset safety threshold, the larger the correction step size; conversely, the smaller the correction step size.

[0100] It should be noted that i=1, 2, and 3 correspond to the over-limit states of voltage, transformer load, and energy storage SOC. Photovoltaic non-functional capacity constraints and power factor constraints are physical boundary constraints of the equipment and are not included in the over-limit depth calculation. Instead, a global real-time verification mechanism ensures the execution of instructions during the correction process. In this embodiment, the over-limit weight w1 = 0.6 for voltage, w2 = 0.4 for transformer load, and w3 = 0.2 for energy storage SOC (w3 can also be dynamically increased when SOC is in extreme ranges).

[0101] S3; Determine the adjustment direction based on the control method, use the correction step size as the adjustment amount, generate updated values ​​for various indicators, and generate new control commands.

[0102] In this embodiment, a new index value is obtained by superimposing an adjustment amount of a specified direction and equal to the correction step size onto the original values ​​of each index in the original control command, thus forming the updated control command. During this process, if photovoltaic reactive power regulation is involved, the aforementioned global constraint verification strategy is executed simultaneously.

[0103] S4: Use a co-simulation model to perform rolling timing deduction of the new control commands, and determine whether the safety limit violation state has been lifted based on the trajectory data generated during the deduction process. (1) If so, the iteration ends and the optimized control command is output.

[0104] (2) If not, continue to determine whether the maximum number of iterations has been reached: If the target has been reached, a final control command is generated through a safety net mechanism; otherwise, the process returns to step S3 to generate a new control command.

[0105] In this embodiment, to prevent the charging pile from falling into a "dead loop" due to the inability to remove the safety limit exceeding the limit through conventional control methods in extreme cases, this embodiment sets a safety net mechanism for the optimization process. That is, the iteration termination condition is set to satisfy the constraints or reach the preset maximum number of iterations N. maxTypical values ​​can be set to 6-10 iterations. Correspondingly, when the number of iterations n ≥ N... max If the convergence fails to materialize by then, the safety net mechanism will be triggered immediately.

[0106] In this embodiment, the safety protection mechanism comprises two layers. Upon triggering, it first initiates a primary response, including: forcibly disconnecting the charging connections of all EVs with a credibility score below the safety threshold; switching the energy storage system to a "voltage / reactive power priority support mode"; and limiting the active power output of the photovoltaic system to reserve reactive power regulation capacity. The aforementioned control measures in the primary response are more stringent than the adaptive iterative process, but are still relatively mild, allowing for partial preservation of the charging station's operational functions. This embodiment further uses a co-simulation model to determine whether the charging station's safety over-limit state has been resolved after the primary response: if so, the primary response control is successful, generating the corresponding ultimate control command and issuing it to the corresponding control object.

[0107] Conversely, if the first-level response fails, it indicates a failure. In this case, a second-level response is initiated, generating the corresponding ultimate control command. The second-level response includes disconnecting non-critical loads and, if the risk of transformer reverse overload or frequency anomaly persists, switching the energy storage system control mode to a "grid-based V / f control mode" to establish local voltage and frequency support, or executing a station-wide emergency shutdown protection under extreme overload conditions. This second-level response strategy is based on the microgrid islanding principle. By disconnecting non-critical loads and switching to the "grid-based V / f control mode," it constructs a stable system with self-balancing power and voltage and frequency independently supported by local sources. Under this specific operating mode, the system's voltage and frequency are directly clamped by the energy storage converter, theoretically ensuring that its operating point remains within a safe range. Therefore, the charging station can be directly considered safe under this control state without the need for complex co-simulation verification, thus avoiding the risk of algorithm dead loops under extreme conditions while ensuring absolute safety.

[0108] Example 2 Building upon the safety verification and correction method for power regulation of photovoltaic-storage-charging stations provided in Example 1, this example further offers a more optimized solution. This optimization includes, on the one hand, accelerating the design of the co-simulation model to ensure its real-time performance and data processing efficiency under local deployment conditions. On the other hand, it includes optimizing the hyperparameters of the sub-models in the co-simulation model by incorporating real-time updated operational data from the charging station, so that the simulation results more closely approximate the actual operational data of the charging station.

[0109] Specifically, in practical applications, the acceleration mechanism for co-simulation model design in this embodiment includes: (i) In the power flow model of the distribution network, the voltage-power sensitivity matrix is ​​used to realize the power flow calculation.

[0110] In this mode, the voltage-power sensitivity matrix M (obtained through the power flow Jacobian matrix) is used by default instead of nonlinear power flow calculation. After setting the value of the voltage-power sensitivity matrix M, the simulation will use V... t = V base + M·ΔP t Quickly determine the voltage at each node and the load on each branch. Where V base Represents the fundamental value of the node voltage; ΔP t Vt represents the vector of changes in the injected active power at each node of the system relative to the base state at time t; Vt represents the simulated real-time voltage prediction value of each node at time t.

[0111] In practical applications, the sensitivity matrix M is updated once at the beginning of each control cycle to reflect topology changes. When a change in network topology is detected or the power fluctuation of a critical branch exceeds a preset threshold, the matrix is ​​immediately recalculated to ensure the effectiveness of the linear approximation.

[0112] (ii) When the co-simulation model performs rolling time series simulation, the event step size adopts a dynamic adjustment strategy.

[0113] In this embodiment, when the co-simulation model performs rolling timing extrapolation, for each control cycle T, the system executes rolling simulation and safety verification for a specified future time window Δt (e.g., 15-60 minutes). Let the current time be T, and the simulation time axis be t: t ∈ [T, T+Δt], with the step size dynamically adjustable. In this mode, in the initial stage, a default maximum event step size (e.g., 60s) is used to perform coarse-grained extrapolation of the control window. When the power command change rate is detected to exceed a preset threshold or the simulation index enters a preset warning zone, the event step size is shortened (down to as short as 1s) to perform refined extrapolation of the control window. In practical applications, this embodiment can also establish a mapping relationship between the degree of safety exceedance assessed by the system and the set event step size, facilitating adaptive adjustment of the event step size.

[0114] This embodiment employs a rolling time-domain mechanism of "long-window simulation and short-cycle execution," which can both identify long-term risks in advance and avoid strategy failure due to the accumulation of prediction errors. Furthermore, by default using the maximum event step size, it ensures that the co-simulation model always performs simulations in a more coarse-grained mode for most of the time when the charging station is in a non-safe, over-limit state; this significantly reduces the system's data processing scale.

[0115] In this embodiment, the method for optimizing the prediction accuracy of the co-simulation model is as follows: After each control command is issued, the actual operating status of the charging station in the control window is obtained and compared with the corresponding state derived by the co-simulation model to evaluate the prediction accuracy of the co-simulation model; when the prediction accuracy is lower than the preset accuracy threshold, the update mechanism of the co-simulation model is triggered.

[0116] In practical applications, to avoid the impact of sensor bias on the actual operating status, this experiment can also use a statistical outlier detection mechanism to remove outliers caused by sensor malfunctions. In the outlier cleaning stage, this embodiment adopts the 3σ principle.

[0117] Specifically, when triggering a co-simulation model update, for the photovoltaic output model, the relative prediction error of photovoltaic power is calculated. : ; when Exceeding the preset accuracy threshold At this time, the photovoltaic output model update mechanism is triggered. In this embodiment, The value is dynamically determined based on the historical prediction error distribution. ,in, This represents the standard deviation of the photovoltaic forecast error over the past 7 days; this design avoids accidental updates due to a single large error during periods of drastic weather fluctuations.

[0118] The hyperparameters of the photovoltaic power output model include the attenuation coefficient; its update method is as follows: ; In the above formula, and These represent the decay coefficients before and after the update, respectively; This represents the photovoltaic power generation predicted by the model; This represents the actual power generation at the corresponding moment. The attenuation coefficient update also needs to meet the safety threshold constraint of this parameter; in this embodiment, ∈[0.8, 1.0].

[0119] The hyperparameters of the energy storage response model include the energy storage efficiency and health of the energy storage device. The energy storage efficiency is updated as follows: if the deviation between the predicted and actual energy storage efficiency exceeds a deviation threshold, the predicted energy storage efficiency is replaced with the actual energy storage efficiency. Specifically, the relative deviation of the energy storage efficiency is calculated. : ; when Exceeding the preset calibration threshold At that time, the simulation preset efficiency will be used. Update to actual efficiency : In this embodiment, Based on the historical efficiency error distribution, dynamically determined, taking... ,in This represents the standard deviation of energy storage efficiency error over a period of time.

[0120] The health status is updated using a linear decay model, based on the actual cumulative equivalent full cycle count of the energy storage system. renew: ; in, and The images show the energy storage health status before and after the update. k This is the attenuation coefficient, preset according to the battery type.

[0121] The hyperparameters of the EV cluster behavior model include the credibility score of the agent corresponding to each user. R i If any user fails to discharge as instructed or the response delay exceeds a preset threshold, the credibility score of the corresponding agent will be reduced. R i Decrease the specified value.

[0122] Example 3 Based on the solutions in Embodiments 1 and 2, this embodiment further provides a control system for a photovoltaic energy storage charging station. Wherein, as... Figure 3 As shown, the control system of the photovoltaic-storage-charging station includes: a strategy generation module, a charging station twin module, a decision-making module, and a model update module.

[0123] The strategy generation module generates initial control commands based on the real-time operating status of the photovoltaic-storage-charging station. A dynamically updated co-simulation model runs within the charging station twin module. This model performs joint simulations of the states and interactions of the photovoltaic equipment, energy storage equipment, EV clusters, and power distribution network within the charging station; thereby enabling rolling time-series projections of the charging station's future operating states based on its current operating status.

[0124] The decision-making module uses the charging station twin module to predict whether the charging station will exceed safety limits after adopting any control command. If so, it first selects the control method based on the error type, then calculates the error index by weighted fusion of the exceedance depths of various preset constraints, and generates an adaptive correction step size based on a linear model using the error index as a variable. The initial control command content is then iteratively optimized using the correction step size as the adjustment amount for each round, and the updated control command is issued. Otherwise, the control command is issued to the designated object for execution. The model update module calculates the prediction accuracy of the co-simulation model after each round of control command issuance, based on the actual operating state of the charging station and the predicted state output by the charging station twin module. When the prediction accuracy threshold is lower than a preset value, the hyperparameters of the co-simulation model are updated.

[0125] The control system of the photovoltaic energy storage charging station provided in this embodiment is essentially a computer device, which includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the computer program is executed by the processor, it adopts the safety verification and correction method for power regulation of the photovoltaic energy storage charging station as in Embodiment 1 or 2, so as to generate dynamically updated control instructions based on the real-time operating status of the charging station, so as to maintain the stable operation of the charging station.

[0126] In practical applications, the computer equipment can be a standalone computer device, such as a laptop, tablet, desktop computer, or a rack server, blade server, tower server, or cabinet server (including standalone servers or server clusters composed of multiple servers) capable of executing computer programs.

[0127] The computer device in this embodiment includes, but is not limited to, a memory and a processor that can be interconnected via a system bus. In this embodiment, the memory (i.e., the readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory can be an internal storage unit of the computer device, such as the hard disk or RAM of the computer device. In other embodiments, the memory can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Of course, the memory can also include both internal storage units and external storage devices of the computer device. In this embodiment, the memory is typically used to store the operating system and various application software installed on the computer device. Furthermore, the memory can also be used to temporarily store various types of data that have been output or will be output.

[0128] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is typically used to control the overall operation of a computer device.

[0129] Simulation test

[0130] To verify the performance of the safety verification and correction method for power regulation in photovoltaic-storage charging stations provided by this invention, technicians conducted simulation tests on the relevant scheme in a small-scale integrated photovoltaic-storage charging station in an urban area. The test content is as follows: I. System configuration of photovoltaic-storage-charging station: The photovoltaic system has an installed capacity of 600 kW and an initial degradation coefficient. = 1.0. The energy storage system has a rated capacity of 250kWh, a maximum charge / discharge power of 125 kW, an initial battery health state of 100%, and a current state of charge (SOC) of 65%. 40 V2G-enabled electric vehicles are connected to the station, 8 of which are newly registered users (credibility rating R). i = 0.9), while the historical fulfillment rates of the remaining vehicles range from 0.72 to 0.96. The charging station is connected to a 10 kV distribution network, and the transformer used has a rated capacity of 800 kVA.

[0131] II. Scene Setup and Simulation Process

[0132] 2.1 System Control

[0133] At the start of the control period T = 10:00, the upper-level optimization algorithm generates the initial control strategy, including the reactive power setpoint Q of the photovoltaic inverter. pv = -80 kVar; Energy storage system discharge power command P ess cmd(t) = 100 kW; Total power command P of electric vehicle cluster ev cmd(t) = 150 kW, with a time window of [10:00, 10:15].

[0134] The system injects initial control commands into the co-simulation model for rolling timing simulation. Simulation results show that at t = 10:07, V node (t) drops to 0.93 pu (below the 0.95 pu safety lower limit), constituting a "voltage below the lower limit" error; at this time, L trafo (t) is 82%, which is within a safe range; during this period, there were 6 electric vehicles with R i < 0.8 and C avail (t) > 0, indicating that it has adjustable capability.

[0135] The system triggers a dynamic correction mechanism, correcting according to the control direction of voltage below the lower limit (reactive power support priority, active power discharge supplement, load reduction as a safety net): The over-limit depth Δ_err is calculated as (0.95 - 0.93) / 0.95 = 0.022, and the error index severity = 0.6. 0.022 = 0.0132, correction step size Kn = 0.5 x 0.0132 + 0.1 = 0.1066.

[0136] Based on the control direction of the voltage being below the lower limit, the "reactive power support priority" strategy is first activated. The current voltage gap is 0.02 pu, and the correction step size is Kn = 0.1066. The corresponding voltage rise that needs to be undertaken is ΔV = 0.1066 x (0.95 - 0.93) = 0.002132 pu.

[0137] The required reactive power adjustment is calculated based on the reactive power-voltage sensitivity Sqv at the current operating point. In this embodiment, based on the current grid topology and load level, Sqv = 0.0005 pu / kVar is calculated in real time (this value varies with operating conditions and is only an example here). Theoretically, the required reactive power adjustment is 0.002132 / 0.0005 = 4.264 kVar. The current photovoltaic command is -80 kVar, which has a reverse effect on voltage boosting. Therefore, the reactive power command needs to be adjusted to issue capacitive reactive power, with a theoretical target value of Qpv = -80 +(80+4.264) = +4.264 kVar.

[0138] Before implementing photovoltaic reactive power regulation, initiate a global constraint verification mechanism: The current photovoltaic active power Ppv = 590 kW, rated capacity Smax = 600kVA, calculated maximum reactive power Qmax = (600^2 - 590^2)^(1 / 2) ≈109kVar, theoretical target value Qpv = +4.26 is within the range of [-109, +109], and has not exceeded the limit.

[0139] Power factor calculation: Active power: Ptotal = -590 + (-100) + (+150) = -540 kW

[0140] Reactive power Qtotal = +4.264 + 0 + 0 = +4.264 kVar

[0141] PF = 540 / (540^2 + 4.264^2)^(1 / 2) ≈ 0.997, which is within the range of [0.9, 1.0] and meets the requirements.

[0142] Verification passed, new strategy generated: S1 = { Qpv = +4.264 kVar, Pess,cmd = 100 kW, Pev,cmd = 150kW}.

[0143] The system will re-inject the new strategy S1 into the simulation model for verification: Simulation results: At t = 10:07, V node (t) rose to 0.96 pu, L trafo (t) rose to 85%, still within a safe range.

[0144] All safety constraints are removed, and the iteration terminates.

[0145] Determine the final strategy and issue it to each controlled object for execution.

[0146] After the control cycle ends (10:15), the system collects actual operating data. Actual output of photovoltaic power kW, actual energy storage power kW, EV cluster actual power kW, actual node voltage PU, actual load rate of transformer .

[0147] 2.2 Model Update

[0148] The system starts a self-learning update process for parameters, which incorporates measured data. Compared with simulation predictions (Photovoltaic power forecast) kW, energy storage predicted power kW, EV cluster predicted power The parameters of the model in the co-simulation environment are compared with those of the kW model, and online calibration is performed based on the deviation. (1) Photovoltaic model calibration Calculate the relative error of photovoltaic output:

[0149] in, ( (This represents the standard deviation of the photovoltaic forecast error over the past 7 days). Therefore, the degradation coefficient is updated as follows:

[0150] And limited to a preset physical range Inside.

[0151] (2) EV credibility calibration

[0152] According to the communication logs, the response delays of three electric vehicles exceeded a preset threshold. (This threshold is dynamically determined based on the historical response delay distribution.) Their credibility scores were lowered accordingly: ; The credibility of user i before and after the update is shown below; the credibility of other vehicles remains unchanged.

[0153] (3) Energy storage health status update

[0154] The actual average efficiency is calculated based on the cumulative charge and discharge energy of the energy storage system during the control cycle. With simulation prediction efficiency The deviation is Calibration threshold In this embodiment, we take ,but .because The triggering conditions have not been met, therefore the efficiency parameters will not be updated at this time. The energy storage health update uses a linear decay model: In this embodiment, a lithium iron phosphate battery is used. Calculate the additional equivalent number of full loops. ,but .

[0155] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for safety verification and correction of power regulation in a photovoltaic-storage-charging station, characterized in that, It includes: A joint simulation model is created to simulate the state and interactions of various objects within the charging station, including: a photovoltaic output model to predict photovoltaic output; an energy storage response model to predict the charging and discharging power and energy storage efficiency of energy storage devices; an EV cluster behavior model to predict the charging and discharging demand and performance status of vehicles entering the station; and a power flow model to predict the voltage of each node in the distribution network, branch current, and transformer load rate. The system obtains the predicted photovoltaic power generation, base load, and charging demand within the future control window, and generates initial control commands through upper-level optimization algorithms. The control commands include: photovoltaic inverter reactive power, energy storage charging and discharging power, and total charging power of the electric vehicle cluster. By using a joint simulation model to perform rolling time-series simulation of the charging station operating environment after the initial control command, extract the indicators related to the safety of the system contained in the trajectory data generated during the simulation process, and then determine whether the safety limit is triggered: (1) If yes, first select the control mode according to the error type, then calculate the error index by weighted fusion of the limit-breaking depth of the preset constraints, and generate an adaptive correction step size based on the linear model with the error index as the variable, and then iteratively optimize the instruction content of the initial control command with the correction step size as the adjustment amount of each round, and issue the updated control command; (2) If no, issue the initial control command directly.

2. The safety verification and correction method for power regulation of photovoltaic energy storage charging station as described in claim 1, characterized in that: The photovoltaic power output model adopts a single diode equivalent circuit model based on physical mechanisms. The photovoltaic power output model combines on-site measured data of irradiance and temperature, and simulates the actual output of the photovoltaic array through the maximum power point tracking algorithm, thereby predicting the power generation at future moments. And / or, the energy storage response model is constructed based on the Thevenin equivalent circuit model; the energy storage response model predicts the output charging and discharging power and energy storage efficiency of the energy storage device at future times based on the changes in the battery state of charge, health state and ambient temperature of the energy storage device. And / or, the EV cluster behavior model models each electric vehicle as an independent intelligent agent, and each intelligent agent has personalized credibility score, departure time, battery capacity, and maximum charging and discharging power; After pre-setting the probability distribution of user response delay and the probability of default, the behavior of each agent in the cluster at future moments is generated based on Monte Carlo simulation. And / or, the power flow model of the distribution network is based on the topology of the standard distribution network test system and uses the forward-backward substitution method to calculate the three-phase power flow, and is used to output the voltage of each node, branch current and transformer load rate at future times.

3. The safety verification and correction method for power regulation of photovoltaic energy storage charging stations as described in claim 2, characterized in that: After each control command is issued, the actual operating status of the charging station in the control window is obtained and compared with the corresponding state derived by the co-simulation model to evaluate the prediction accuracy of the co-simulation model; when the prediction accuracy is lower than the preset accuracy threshold, the update mechanism of the co-simulation model is triggered. And / or, when updating the co-simulation model, the hyperparameters of the photovoltaic output model include the attenuation coefficient; the update method is as follows: ； In the above formula, and These represent the decay coefficients before and after the update, respectively; This represents the photovoltaic power generation predicted by the model; This indicates the actual power generation at the corresponding moment; And / or, the hyperparameters of the energy storage response model include the energy storage efficiency and health of the energy storage device; the energy storage efficiency is updated as follows: if the deviation between the predicted energy storage efficiency and the actual energy storage efficiency exceeds the deviation threshold, the predicted energy storage efficiency is replaced with the actual energy storage efficiency; the health is updated as follows: the health of the energy storage device is updated using a linear decay model based on the actual cumulative equivalent full cycle count of the energy storage system. And / or, the hyperparameters of the EV cluster behavior model include the credibility score of the agent corresponding to each user. R i If any user fails to discharge as instructed or the response delay exceeds a preset threshold, the credibility score of the corresponding agent will be reduced. R i Decrease the specified value.

4. The safety verification and correction method for power regulation of photovoltaic energy storage charging station as described in claim 2, characterized in that: The power flow model of the distribution network uses a voltage-power sensitivity matrix to calculate the power flow; the voltage-power sensitivity matrix is ​​dynamically updated at the beginning of each control cycle; and when a change in network topology is detected or the power fluctuation of a key branch exceeds a preset threshold, the matrix is ​​recalculated. And / or, when the co-simulation model performs rolling timing simulation, the event step size adopts a dynamic adjustment strategy; In the initial stage, the control window is coarsely simulated using the default maximum event step size. When the power command change rate is detected to exceed the preset threshold or the simulation index enters the preset warning area, the event step size is shortened to perform fine simulation of the control window.

5. The safety verification and correction method for power regulation of photovoltaic energy storage charging stations as described in claim 1, characterized in that, The constraints for determining whether a photovoltaic-storage charging pile has exceeded safety limits include: Node voltage constraint: V node (t) ∈ [V min V max ]; Transformer load factor constraint: L trafo (t) < L limit ; Energy storage SOC constraint: SOC(t) ∈ [SOC min SOC max ]; Photovoltaic reactive power constraint: Q pv (t) ∈ [Q min Q max ] ; Power factor constraint: P F (t) ∈ [P Fmin P Fmax ]; If any constraint condition is violated, a safety overrun is determined to have occurred. In the above formula, V node (t) represents the real-time node voltage; V min and V max These are the preset lower and upper limits of the node voltage; L trafo (t) represents the real-time load rate of the transformer; L limit The threshold representing the transformer load factor; SOC(t) represents the real-time charge of the energy storage device; SOC min and SOC max The preset lower and upper limits of the charge capacity of the energy storage device; Q pv (t) represents the real-time reactive power of the photovoltaic device; Q min and Q max This indicates the lower and upper limits of the reactive power of photovoltaic equipment; P F (t) represents the real-time power factor of the distribution network; P Fmin and P Fmax This indicates the lower and upper limits of the power factor in the distribution network.

6. The safety verification and correction method for power regulation of photovoltaic energy storage charging stations as described in claim 5, characterized in that, The error types include: voltage below the lower limit, voltage above the upper limit, transformer overload, transformer reverse overload, energy storage SOC too high, and energy storage SOC too low. And / or, the method for generating control methods based on error type includes adjusting the direction of adjustment and the priority of resource allocation, with specific strategies as follows: (i) When the voltage is below the lower limit, establish the adjustment direction of "prioritizing reactive power support, supplementing active power discharge, and ensuring load reduction as a safety net"; The system prioritizes instructing the photovoltaic inverter and energy storage to generate capacitive reactive power and increase the energy storage discharge power. After the voltage transient trend stabilizes, the system increases the discharge power of the EV to gradually replace the energy storage output and / or the discharge power of the energy storage device, thereby reducing energy storage losses. If the limit cannot be eliminated, the system finally reduces or cuts off the charging power of some EVs and increases the reactive power output of the photovoltaic system in ascending order of reliability until the lower limit is reached. (ii) When the voltage is higher than the upper limit, the adjustment direction should be "prioritizing reactive power absorption, consuming active power, and ensuring a minimum level of photovoltaic curtailment"; First, instruct the photovoltaic inverter and energy storage to absorb inductive reactive power and increase the energy storage charging power; after the voltage transient trend stabilizes, maximize the EV charging power; if the energy storage is full and there is no load to accept, then finally execute the curtailment strategy to limit the photovoltaic active power output. (iii) When the transformer is overloaded, establish the adjustment direction of "rapid peak shaving by energy storage, collaborative sharing by EVs, and orderly load shedding"; The system prioritizes instructing the energy storage system to increase its discharge power; after the load pressure is initially relieved, the EVs are called in to discharge; if the total load still exceeds the limit, the charging power of some EVs is reduced or cut off in ascending order of reliability. (iv) When the transformer is overloaded in reverse, establish the adjustment direction of "full absorption of energy storage, expansion of EV capacity for consumption, and limited photovoltaic power generation as a safety net"; First, instruct the energy storage system to increase its charging power; after the reverse power flow is controlled, increase the EV charging power; if the energy storage is full and the charging load has reached its limit, then finally forcibly limit the grid-connected active power of the photovoltaic inverter. (v) When the SOC is too low, establish the adjustment direction of "blocking discharge, switching to reactive power support, and prioritizing recharge"; Immediately prohibit the energy storage discharge command and control it to switch to reactive power support mode. At the same time, reduce the charging load in the station or use the off-peak electricity period of the grid for rapid recharging. If it still cannot be restored to the safe range, temporarily take the energy storage system offline and readjust the two according to the current actual achievable efficiency and available capacity of the energy storage. (vi) When the SOC is too high, establish the adjustment direction of "blocking charging, switching to reactive power support, and accelerating discharge"; Immediately prohibit energy storage charging commands and control them to switch to reactive power support mode. At the same time, actively increase EV charging power or guide EV discharge to accelerate the consumption of surplus power in the station. In extreme cases, link up with photovoltaic curtailment. (vii) Global constraint verification mechanism: When executing any strategy involving photovoltaic reactive power regulation, the system verifies in real time whether the photovoltaic reactive power and power factor meet the preset constraints; if the calculated command causes the photovoltaic reactive power and power factor to exceed the limits, the original command issuance is immediately suspended, and the constraint saturation and multi-source compensation strategy is executed instead, including: First, the photovoltaic reactive power command is clamped to the current maximum allowable safety boundary, and the resulting reactive power deficit is calculated. Then, the remaining reactive power capacity of the energy storage system and EV is used to compensate for it. If the reactive power resources of the whole station are still insufficient, the active power-reactive power replacement mechanism is activated to reduce the photovoltaic active power output to release reactive power capacity until all constraints are met at the same time.

7. The safety verification and correction method for power regulation of photovoltaic energy storage charging stations as described in claim 6, characterized in that, Correction step size K n The calculation formula is: ； In the above formula, A and A 0 represents the base adjustment amount and preset gain of the corresponding indicator, respectively; Indicates the first i The depth at which a constraint exceeds its limit; value i Indicates the first i The predicted value of the indicator corresponding to the constraint; Indicates the first i The one-sided threshold closest to the predicted value in the constraints; the over-limit states of voltage, transformer load, and energy storage SOC when i=1, 2, 3; severity Indicates the error index.

8. The safety verification and correction method for power regulation of photovoltaic energy storage charging station as described in claim 7, characterized in that: The process of iteratively optimizing the content of the initial control commands includes: S1: Analyze the error types included in the current safety limit violation state and select the corresponding control method; S2: Based on the predicted values ​​of voltage, transformer load factor, and energy storage SOC-related constraints. value i Calculate the depth of exceeding the limit for each of the corresponding safety thresholds. Error indicators severity and correction step size K n ; S3; Determine the adjustment direction based on the control method, use the correction step size as the adjustment amount, generate updated values ​​for various indicators, and generate new control commands; S4: Use a co-simulation model to perform rolling timing deduction of the new control commands, and determine whether the safety limit violation state has been lifted based on the trajectory data generated during the deduction process. (1) If so, end the iteration and output the optimized control command; (2) If not, continue to determine whether the maximum number of iterations has been reached: if yes, generate the final control command through the bottom-line mechanism; otherwise, return to step S3 to generate a new control command.

9. The safety verification and correction method for power regulation of photovoltaic energy storage charging station as described in claim 8, characterized in that: The safety net mechanism first initiates a level-one response, including: forcibly disconnecting the charging connections of all EVs with a credibility score below the safety threshold; switching the energy storage system to "voltage / reactive power priority support mode"; limiting the active power output of photovoltaic systems to reserve reactive power regulation capacity; and determining whether the safety limit breach has been resolved. If yes, generate the ultimate control command corresponding to the first-level response; otherwise, perform a second-level response and generate the ultimate control command corresponding to the second-level response. The Level 2 response includes disconnecting non-critical loads, and if the risk of transformer reverse overload or frequency anomaly is not eliminated, switching the energy storage system control mode to "grid-type V / f control mode" to establish local voltage and frequency support, or executing a station-wide emergency shutdown protection under extreme overload conditions.

10. A control system for a photovoltaic-storage-charging station, characterized in that: It employs the safety verification and correction method for power regulation of the photovoltaic-storage-charging station as described in any one of claims 1-9, to generate dynamically updated control commands based on the real-time operating status of the charging station, thereby maintaining the stable operation of the charging station; the control system of the photovoltaic-storage-charging station includes: The strategy generation module is used to generate initial control commands based on the real-time operating status of the photovoltaic energy storage charging station; The charging station twin module runs a dynamically updated co-simulation model. The co-simulation model is used to jointly simulate the state of photovoltaic equipment, energy storage equipment, EV clusters, and power distribution network within the charging station and their interaction processes; thereby enabling rolling time-series extrapolation of the future operating state of the charging station based on its current operating state. The decision module is used to predict whether the charging station will exceed safety limits after adopting any control command through the charging station twin module. If so, it first selects the control method according to the error type, then calculates the error index by weighted fusion of the exceedance depth of the preset constraints, and generates an adaptive correction step size based on a linear model with the error index as the variable. Then, it iteratively optimizes the command content of the initial control command with the correction step size as the adjustment amount in each round, and issues the updated control command; otherwise, the control command is issued to the specified object for execution. The model update module is used to calculate the prediction accuracy of the co-simulation model after each round of control command is issued, based on the actual operating status of the charging station and the predicted status output by the charging station twin module, and to update the hyperparameters of the co-simulation model when the prediction accuracy threshold is lower than the preset value.

Images