Transient power fluctuation suppression process control method for integrated photovoltaic-storage-charging power plants

Abstract

This invention relates to the field of microgrid control technology, specifically disclosing a process control method for suppressing transient power fluctuations in an integrated photovoltaic-storage-charging power station. The method is applied to a DC microgrid-type integrated photovoltaic-storage-charging power station with a three-level controller architecture. This three-level controller architecture consists of an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller connected in communication. The technical solution provided by this invention establishes a three-level collaborative architecture of upper-level global optimization, middle-level hybrid energy storage allocation, and lower-level local control, enabling hierarchical response and collaborative processing of transient power fluctuations in actual operation. Based on upper-level prediction and middle-level power allocation, the lower-level controller utilizes nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control working in tandem to provide advance compensation and dynamic calculation correction for transient disturbances such as sudden load increases on charging piles, mitigating the lag phenomenon of conventional single feedback control.

Description

Technical Field

[0001] This invention relates to the field of microgrid control technology, specifically to a process control method for suppressing transient power fluctuations in an integrated photovoltaic-storage-charging power station. Background Technology

[0002] Integrated photovoltaic-storage-charging power plants typically employ a centralized energy management system combined with a control architecture for the underlying converter to balance power supply and demand between photovoltaic power generation and charging load. In normal operation, the energy management system calculates a power reference value based on the current source-load state and sends it to the underlying energy storage converter via a communication network. The underlying converter then uses a conventional proportional-integral control algorithm to perform local voltage or current closed-loop regulation to maintain the voltage state of the DC bus.

[0003] In application scenarios where multiple high-power DC charging piles start up concurrently or when there are sudden changes in environmental weather, the DC microgrid of the power station will experience drastic transient power fluctuations. Due to the presence of communication links, the existing centralized control architecture is unable to respond in a timely manner to transient disturbances ranging from microseconds to hundreds of milliseconds; at the same time, the underlying single feedback control has physical-level adjustment lag when facing nonlinear electrical disturbances.

[0004] Furthermore, when the system encounters communication interruptions or sensor anomalies of varying durations during transient fluctuations, existing technologies lack coordinated degradation protection mechanisms for different time scales and control levels, which can easily lead to bus voltage exceeding limits or cause converter protective shutdown. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a process control method for suppressing transient power fluctuations in integrated photovoltaic-storage-charging power plants, thereby improving the reliability of integrated photovoltaic-storage-charging power plants when facing complex electrical disturbances and non-ideal operating conditions.

[0006] To achieve the above objectives, a first aspect of the present invention proposes a process control method for suppressing transient power fluctuations in an integrated photovoltaic-storage-charging power station. This method is applied to a DC microgrid-type integrated photovoltaic-storage-charging power station with a three-level controller architecture. The three-level controller architecture consists of an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller connected in communication. The method includes:

[0007] The system acquires multi-source operation data of the power plant, executes multi-source fusion power prediction through the upper-level global optimization controller in a preset first control cycle, generates an optimization strategy that includes total power adjustment instructions and pre-adjustment instructions, and sends it to the middle-level hybrid energy storage controller.

[0008] The mid-layer hybrid energy storage controller executes dual-closed-loop fuzzy hybrid energy storage power allocation based on the total power adjustment command in the optimization strategy during a preset second control cycle, generating a first power reference command for power-type energy storage units and a second power reference command for energy-type energy storage units, and sends them down to the bottom-layer converter local controller.

[0009] The underlying converter local controller performs nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control based on the first power reference command and the second power reference command in a preset third control cycle to generate a pulse width modulation duty cycle signal to drive the power electronic converter to perform power regulation; wherein, the duration of the third control cycle is less than the duration of the second control cycle, and the duration of the second control cycle is less than the duration of the first control cycle.

[0010] Multi-dimensional online fault identification is performed synchronously within the aforementioned control cycle. Faults are classified into first-level, second-level, and third-level faults according to the fault duration scale. Unobstructed degradation control that matches the time scale and control level is executed to achieve closed-loop suppression of transient power fluctuations.

[0011] To achieve the above objectives, a second aspect of the present invention proposes a transient power fluctuation suppression process control system for an integrated photovoltaic-storage-charging power station, which is applied to a DC microgrid-type integrated photovoltaic-storage-charging power station. The system includes an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller that are interconnected, forming a three-level controller architecture.

[0012] The upper-level global optimization controller is used to acquire multi-source operation data of the power plant, execute multi-source fusion power prediction in a preset first control cycle, generate an optimization strategy that includes total power adjustment instructions and pre-adjustment instructions, and send it to the middle-level hybrid energy storage controller.

[0013] The mid-layer hybrid energy storage controller is used to execute dual-closed-loop fuzzy hybrid energy storage power allocation based on the total power adjustment command in the optimization strategy at a preset second control cycle, generate a first power reference command for power-type energy storage units and a second power reference command for energy-type energy storage units, and send them down to the bottom-layer converter local controller.

[0014] The underlying converter local controller is used to perform nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control based on the first power reference command and the second power reference command in a preset third control cycle, and generate a pulse width modulation duty cycle signal to drive the power electronic converter to perform power regulation; wherein, the duration of the third control cycle is less than the duration of the second control cycle, and the duration of the second control cycle is less than the duration of the first control cycle.

[0015] The three-level controller architecture is also used to synchronously perform multi-dimensional online fault identification within the aforementioned control cycle, classify faults into first-level faults, second-level faults, and third-level faults according to the fault duration scale, and perform non-disruptive degradation control that matches the time scale and control level to achieve closed-loop suppression of transient power fluctuations.

[0016] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for controlling transient power fluctuations in an integrated photovoltaic-storage-charging power station.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] The technical solution provided by this invention establishes a three-level collaborative architecture consisting of upper-level global optimization, mid-level hybrid energy storage allocation, and lower-level local control. In actual operation, this architecture enables layered response and collaborative processing of transient power fluctuations. Based on upper-level prediction and mid-level power allocation, the lower-level controller utilizes nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control in synergy. This allows for advance compensation and dynamic calculation correction of transient disturbances such as sudden load increases on charging piles, mitigating the lag phenomenon of conventional single feedback control. Simultaneously, the system synchronously performs multi-dimensional fault identification during operation and triggers matching non-disruptive degradation control based on the fault duration. This ensures that even when encountering fault conditions such as missing sensor data or inter-level communication interruptions, the power station can still maintain system operation and DC bus voltage stability through the corresponding level's takeover mechanism. This reduces the probability of overall power station shutdown and improves the reliability of the photovoltaic-energy storage-charging integrated power station in the face of complex electrical disturbances and non-ideal operating conditions. Attached Figure Description

[0019] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0020] Figure 1 This is a flowchart illustrating the transient power fluctuation suppression process control method for the integrated photovoltaic-storage-charging power station provided by the present invention.

[0021] Figure 2 This is a comparison chart of multi-source fusion power prediction and actual load fluctuation tracking in the transient power fluctuation suppression process control method for integrated photovoltaic, energy storage and charging power stations provided by this invention;

[0022] Figure 3This is a nonlinear response surface diagram of the dual-closed-loop fuzzy power allocation smoothing coefficient in the transient power fluctuation suppression process control method for integrated photovoltaic, energy storage and charging power stations provided by the present invention.

[0023] Figure 4 This is a Pareto front and fitness convergence curve of the multi-objective optimization of the transient control joint cost function in the transient power fluctuation suppression process control method of the photovoltaic-storage-charging integrated power station provided by the present invention;

[0024] Figure 5 This is a graph showing the evolution trend of equivalent ohmic internal resistance and polarization time constant of the energy storage battery throughout its entire life cycle in the transient power fluctuation suppression process control method for the integrated photovoltaic-storage-charging power station provided by this invention.

[0025] Figure 6 This is a schematic diagram of the adaptive reconstruction space mapping of feedforward gain and phase lead angle in the transient power fluctuation suppression process control method of the photovoltaic-storage-charging integrated power station provided by the present invention;

[0026] Figure 7 This is a bitmap showing the electrothermal coupling between virtual junction temperature and transient limiting safety boundary (SOA) under heavy load conditions in the transient power fluctuation suppression process control method for integrated photovoltaic, energy storage and charging power plants provided by this invention.

[0027] Figure 8 This is a schematic diagram illustrating the implementation of the transient power fluctuation suppression process control system for the integrated photovoltaic-storage-charging power station provided by the present invention.

[0028] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0029] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0030] The following description, with reference to the accompanying drawings, outlines a method, system, and electronic device for suppressing transient power fluctuations in an integrated photovoltaic, energy storage, and charging power station according to embodiments of the present invention.

[0031] Example 1:

[0032] This embodiment provides a process control method for suppressing transient power fluctuations in an integrated photovoltaic-storage-charging power station. Specifically, this method is applied to a DC microgrid-type integrated photovoltaic-storage-charging power station with a three-level controller architecture.

[0033] In practical industrial applications, the hardware topology of this DC microgrid-based photovoltaic-storage-charging integrated power station mainly includes photovoltaic power generation units, a hybrid energy storage system, AC / DC charging pile equipment, and a common DC bus connecting each power unit. To achieve precise scheduling and disturbance-resistant control of the entire microgrid system, the system constructs the aforementioned three-level controller architecture at the cyber-physical level. This architecture, from top to bottom, consists of an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller connected in communication.

[0034] The upper-level global optimization controller is typically deployed on the power plant's edge computing server or cloud dispatch center, possessing powerful computing capabilities and responsible for long-term energy management and multi-source data fusion processing. The mid-level hybrid energy storage controller is usually a microgrid central controller built on a multi-core digital signal processor, responsible for coordinating the power output ratio of various energy storage media. The lower-level converter local controller is directly embedded inside each power electronic converter, responsible for acquiring high-frequency electrical signals and directly generating control waveforms for the driver chips.

[0035] Based on the above hardware platform and control architecture, the transient power fluctuation suppression process control method of the integrated photovoltaic-storage-charging power station in this embodiment includes the following steps:

[0036] Step 1: Acquisition of multi-source operating data and generation of global power reference instructions.

[0037] Specifically, the system first acquires multi-source operating data of the power plant, and then executes multi-source fusion power prediction through the upper-level global optimization controller in a preset first control cycle. This generates an optimization strategy that includes total power adjustment instructions and pre-adjustment instructions, which is then sent to the middle-level hybrid energy storage controller. In actual operation, the preset first control cycle is relatively long, mainly to match the energy dispatch time scale of the microgrid, in order to handle massive amounts of meteorological data and historical load characteristics.

[0038] In this step, the specific steps of the multi-source fusion power prediction include: First, acquiring historical photovoltaic power generation, local environmental meteorological data, and real-time charging status. Local environmental meteorological data includes irradiance, ambient temperature, wind speed, etc., collected by the on-site meteorological station; real-time charging status includes the current occupancy status of each charging pile, power demand, and reserved charging queue data.

[0039] Subsequently, the aforementioned operational data is input into a fusion prediction network model consisting of a deep learning time series prediction network and a gradient boosting tree model. The deep learning time series prediction network is mainly responsible for extracting the complex time-series nonlinear mapping relationship between photovoltaic power generation and meteorological conditions; the gradient boosting tree model is used to process discrete structured data containing user behavior features to improve the generalization prediction capability for sudden changes in charging load.

[0040] Through the fusion prediction network model, the system can output the photovoltaic power output prediction curve and the charging pile load prediction curve for a preset future time period, and generate the total power adjustment command based on the calculated difference between the photovoltaic power output prediction curve and the charging pile load prediction curve. When a power deficit or power surplus greater than a preset power threshold is predicted in the future, a pre-adjustment command is generated to control the energy storage unit to perform pre-discharge or pre-charge. The physical significance of this pre-adjustment mechanism is that, due to the chemical characteristics of energy storage units, which have large capacity but slow response, issuing pre-discharge or pre-charge commands in advance can guide their state of charge to enter the optimal adjustment range in advance, thereby avoiding a significant drop in bus voltage due to insufficient response during sudden power deficits.

[0041] like Figure 2 This demonstrates the operational performance of the upper-level global optimization controller in performing multi-source fusion power prediction. The horizontal axis of the graph represents time in minutes, recording the prediction time window for the next 10 minutes; the vertical axis represents power in kilowatts, reflecting the power flow level within the power plant.

[0042] The figure contains three sets of core data presentation formats. The light blue filled area represents the 95% confidence interval boundary calculated by the algorithm, the continuous black solid line represents the actual load power curve of physical acquisition, and the blue discrete dashed line represents the fused predicted power curve jointly output by the deep learning time series prediction network and the gradient boosting tree model.

[0043] Observing the waveform change trajectory in the figure, it can be seen that when the time axis advances to the 5th minute, the actual load power is affected by the sudden high-power charging request and the value jumps from the basic 200 kilowatts to around 330 kilowatts, forming a significant load peak.

[0044] Faced with this power surge, the blue fusion prediction power curve showed a stable early tracking characteristic. Its waveform rose synchronously around the 5th minute and reached a predicted peak of about 310 kilowatts. The prediction curve as a whole was reasonably enveloped within the light blue 95% confidence interval.

[0045] This graphical feature objectively reflects the system's ability to accurately fit the trends of high-frequency load surges and low-frequency weather variations after acquiring historical photovoltaic power generation, local environmental meteorological data, and real-time charging status. This prediction result enables the upper-level global optimization controller to generate total power regulation commands and pre-regulation commands containing specific values ​​before a substantial bus voltage drop occurs. This guides energy storage units to enter the corresponding charging and discharging preparation states in advance, improving the microgrid system's anti-disturbance regulation capability under complex source-load interaction scenarios.

[0046] Step 2: Dual-loop fuzzy hybrid energy storage power allocation.

[0047] For example, the mid-level hybrid energy storage controller, within a preset second control cycle, executes a dual-closed-loop fuzzy hybrid energy storage power allocation based on the total power regulation command in the optimization strategy. This generates a first power reference command for power-type energy storage units and a second power reference command for energy-type energy storage units, which are then sent to the underlying converter local controller. In this process, the duration of the preset second control cycle is shorter than the duration of the first control cycle to match the moderate response frequency of grid power dispatch.

[0048] The specific steps of the dual-closed-loop fuzzy hybrid energy storage power allocation include the following process: The total power adjustment command and the state of charge of the power-type energy storage unit are input into the first fuzzy controller, and the first fuzzy controller outputs the total smoothing coefficient. The practical significance of this process is that the hybrid energy storage system does not always fully respond to the upper-level adjustment command, but needs to dynamically reduce the power based on the remaining energy of its own power-type energy storage medium. To quantify this smoothing process, the following algorithm formula for calculating the total smoothing upper limit is given:

[0049] ;

[0050] in, This indicates the upper limit of the total output power of the hybrid energy storage system. Its physical meaning is the maximum electrical power that the hybrid energy storage system is allowed to release or absorb within the current control cycle. This represents the total smoothing coefficient output by the first fuzzy controller. This coefficient is a dimensionless scalar that characterizes the power discount ratio calculated by the system based on the battery health state and state of charge. This represents the total power adjustment command issued by the upper-level global optimization controller. Multiplying the total power adjustment command by the total damping coefficient yields the upper limit of the total output power of the hybrid energy storage.

[0051] Furthermore, the state of charge (SOC) and its rate of change of SOC of the energy-type energy storage unit are input into a second fuzzy controller, which then outputs allocation coefficients. The upper limit of the total output power of the hybrid energy storage is multiplied by these allocation coefficients to obtain the second power reference command output to the energy-type energy storage unit. Finally, the difference between the upper limit of the total output power of the hybrid energy storage and the second power reference command is defined as the first power reference command output to the power-type energy storage unit. Through this dual fuzzy allocation based on SOC and its rate of change, the system can guide high-frequency pulsating power loads to the power-type energy storage unit and low-frequency, gradual power deficits to the energy-type energy storage unit, thus extending the overall lifespan of the system.

[0052] like Figure 3This diagram illustrates the spatial mapping logic of the first fuzzy controller in a mid-level hybrid energy storage controller during dual-closed-loop fuzzy hybrid energy storage power allocation. The horizontal axis represents the state of charge (SBC) of the power-type energy storage units, in percentages ranging from 0 to 100; the vertical axis represents the total power regulation command, in kilowatts, ranging from -500 to +500; and the vertical axis represents the total output smoothing coefficient, dimensionless, limited to 0 to 1. The Chinese legend in the upper right corner identifies this graph as a smoothing coefficient response surface.

[0053] The three-dimensional nonlinear surface shown in the figure visually reflects the amplitude evolution of the smoothing coefficient under different operating conditions through a color gradient from blue to red. Observing the spatial morphology of the surface, it can be found that when the state of charge of the power storage unit is in the healthy range of 40% to 60%, regardless of whether the total power regulation command is positive charging or negative discharging, the surface height remains in the high red plateau region close to 1, indicating that the system can perform power regulation tasks at a high proportion.

[0054] When the state of charge gradually deviates from the center and approaches the full charge boundary of 100, if the total power adjustment command is a positive charging demand and gradually reaches 500 kilowatts, the height of the curved surface will show a non-linear steep drop trend and descend to the blue valley area close to 0, indicating that the system significantly reduces the upper limit of charging power to prevent the energy storage unit from being overcharged.

[0055] Conversely, if the total power regulation command is changed to a discharge demand of -500 kW, the surface height will quickly rise back to around 1, allowing the energy storage unit to discharge with a large current.

[0056] Similarly, when the state of charge approaches zero at the depletion boundary and faces a significant negative discharge command, the surface height will nonlinearly converge to a trough. This smooth nonlinear surface, formed by the interweaving of multidimensional boundary conditions, alleviates the risk of dead zone in conventional linear control under extreme conditions, ensures the physical safety boundary of power-type energy storage units, and improves the stable operation capability of integrated photovoltaic-storage-charging power stations in complex power throughput scenarios.

[0057] Step 3: Low-level nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control.

[0058] It is important to note that, through the underlying converter local controller in a preset third control cycle, based on the first power reference command and the second power reference command, nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control are executed to generate a pulse width modulation duty cycle signal to drive the power electronic converter for power regulation. The duration of the third control cycle is shorter than the duration of the second control cycle, and the duration of the second control cycle is shorter than the duration of the first control cycle. The underlying control is the core execution link for transient power fluctuation suppression, and its response speed determines the DC bus voltage's resistance to impact.

[0059] The specific steps for implementing nonlinear electric disturbance feedforward compensation and fuzzy adaptive feedback control include: real-time acquisition of the common DC bus voltage sag rate and charger current mutation rate of the integrated photovoltaic-storage-charging power station. The bus voltage sag rate reflects the macroscopic state of the overall power imbalance of the system, while the charger current mutation rate directly characterizes the microscopic impact intensity of the disturbance source. Subsequently, the bus voltage sag rate and charger current mutation rate are input into a nonlinear electric disturbance state observer to generate a feedforward compensation voltage command. To clarify the feedforward compensation quantization relationship of this observer, the following algorithm formula is given:

[0060] ;

[0061] in, This indicates the generated feedforward compensation voltage command, which is used to feed forward directly to the control loop input to counteract the hysteresis effect caused by external electrical disturbances. This represents the built-in bus voltage sag feedforward gain coefficient of the observer; This represents the voltage sag rate of the common DC bus obtained in real time. This represents the feedforward gain coefficient for the charger current surge built into the observer; This represents the real-time current fluctuation rate of the charger. This algorithm allows for the injection of a reverse compensation control quantity before the charger causes substantial damage to the bus voltage.

[0062] Simultaneously, in the feedback loop, the voltage error amplitude between the actual instantaneous voltage of the common DC bus and the preset common DC bus reference voltage, as well as the rate of change of this voltage error amplitude, are calculated. The voltage error amplitude and the rate of change of the voltage error amplitude are then input into a fuzzy adaptive proportional-integral-derivative (PID) controller to generate a feedback compensation voltage command. Finally, the feedforward compensation voltage command and the feedback compensation voltage command are superimposed to generate the pulse width modulation duty cycle signal. This signal is converted into on / off commands for high-power IGBTs or SiC switching devices through the underlying hardware's drive circuitry, completing the physical exchange of energy.

[0063] Step 4: Online identification and classification of multi-dimensional faults.

[0064] Optionally, multi-dimensional online fault identification is performed synchronously within the aforementioned control cycle. Faults are classified into first-level, second-level, and third-level faults according to their duration, and non-disruptive degradation control matching the time scale and control level is executed to achieve closed-loop suppression of transient power fluctuations. Industrial environments are complex, and various electromagnetic interferences or hardware anomalies occur frequently; therefore, an online fault classification mechanism is essential.

[0065] Specifically, the multi-dimensional online fault identification classification criteria include: faults with a duration less than a first preset time threshold are classified as first-level faults, including instantaneous sensor faults and electromagnetic interference. These faults typically manifest as spikes or momentary loss of sampled data and are self-healing. Faults with a duration greater than or equal to the first preset time threshold but less than a second preset time threshold are classified as second-level faults, including communication interruptions between control levels and energy storage converter faults. These faults indicate that some physical links or hardware have been substantially damaged. Faults with a duration greater than or equal to the second preset time threshold are classified as third-level faults, including meteorological platform data acquisition interruptions. These faults represent a lack of macroscopic information at the system level; wherein, the first preset time threshold is less than the second preset time threshold.

[0066] Step 5: Disruptive degradation control based on time and hierarchy matching.

[0067] Specifically, the system designs differentiated response logic for different levels of faults. In the disturbance-free degradation control, the handling logic for the first level of fault is as follows: when a momentary fault of the sensor is identified through residual analysis, the local controller of the underlying converter activates the fault-tolerant state observer and uses the system mathematical model to estimate the output value of the faulty sensor in real time, so as to replace the physical acquisition signal for underlying control.

[0068] This estimation mechanism based on internal state equations can provide virtual data with high confidence in a short time, maintaining the normal closed loop of the underlying control loop. When the physical acquisition signal is detected to have returned to normal and the duration of maintaining the normal state exceeds the preset stability time threshold, the system uses the physical acquisition signal as the control input to replace the estimated value, thereby ending the underlying degradation state.

[0069] For example, in the disturbance-free degradation control, the handling logic for second-level and third-level faults is as follows: When a communication interruption between the control levels is detected in a second-level fault, the local controller of the bottom-level converter independently switches to a pure electrical operation mode and independently schedules each energy storage unit according to a pre-set reserve power allocation ratio. In this mode, the bottom-level controller no longer waits for the middle-level command refresh, but instead independently executes droop control or constant voltage control based on locally measured high-frequency bus voltage and current to ensure the power plant does not shut down. When a meteorological platform data acquisition interruption is detected in a third-level fault, the upper-level global optimization controller switches to a pure historical data prediction mode, using historical operating data from the same period of the past cycle cached by the system to generate a prediction curve to replace real-time data prediction, thereby reducing the negative impact of external data link breakage on global scheduling.

[0070] Step Six: Weighted Overlay Transition of Operating Mode Switching.

[0071] Specifically, to prevent secondary voltage surges caused by sudden changes in control mode, the disturbance-free degradation control employs a weighted superposition disturbance-free parameter switching algorithm during the switching of operating modes. The final control quantity output at each moment is calculated using the following general algorithm formula:

[0072] ;

[0073] in, This represents the final control output calculated by the algorithm at the current moment; This indicates the preset transition switching period duration that matches the current fault level, used to set the time span for the alternation of old and new parameters; This represents the elapsed time since the start of the state transition, and is subject to the following constraints. Time window; This indicates the old control output used by the system before the state transition started; This indicates the new control output that the system will adopt after the state transition is complete; This represents the bus voltage feedforward compensation correction amount used to smooth out transient voltages. It should be noted that the correction amount... It is only during the transition switching period. The incremental terms that are temporarily activated within the internal system are independent of the regular feedforward compensation instructions in terms of physical scope and lifecycle, thus avoiding control integral saturation.

[0074] In the initial stage of the switchover, the algorithm prioritizes the old control output parameters. As time progresses, the weight of the new control output parameters gradually increases until the transition is completed smoothly, which greatly improves the power quality during the switchover process.

[0075] Step 7: Graded verification of self-healing control and parameter self-tuning.

[0076] It should also be noted that after the unobstructed degradation control is executed, the method further includes a graded verification self-healing control step: continuously monitoring the system fault status signal, and when the signal is detected to have recovered and the verification pass condition is met within a continuous preset time period, the system determines that the fault has been physically recovered. After determining that the fault has been physically recovered and stable operation has reached the preset verification delay time, the system reactivates the aforementioned weighted superposition unobstructed parameter switching algorithm, and gradually switches the control mode from the unobstructed degradation control state back to the normal three-level cooperative control state from the bottom up according to the preset algorithm, thereby forming a complete closed loop for fault handling.

[0077] Optionally, during normal operation, the method automatically performs a control parameter self-tuning step at a preset parameter tuning cycle to adapt to device aging and environmental changes that occur during long-term system operation. Specifically, using actual operating dynamic data within a preset historical time period as training samples, a linear joint cost function is constructed, including the maximum bus voltage drop variable and the voltage recovery time variable. To quantify this tuning objective, the algorithm formula for the cost function is given:

[0078] ;

[0079] in, This represents the overall evaluation value of the linear joint cost function; the smaller the value, the better the transient control performance of the system. This indicates the set drop amplitude weighting coefficient, which is used to adjust the proportion of voltage drop depth in the overall evaluation; This represents the maximum voltage drop amplitude variable extracted from actual operational dynamic data; This represents the set recovery time weighting coefficient; This represents the voltage recovery time variable extracted from actual operational dynamic data.

[0080] A genetic algorithm is used to iteratively solve for the feedforward compensation gain parameter of the linear electrical disturbance state observer and the proportional-integral parameter combination of the fuzzy adaptive feedback controller, with the goal of minimizing the linear joint cost function. After the iteration terminates, the feedforward compensation gain parameter and the proportional-integral parameter combination are output, and the current operating control parameters are overwritten and updated, giving the control system the ability to continuously evolve.

[0081] like Figure 4This diagram illustrates the optimization process of the system automatically executing control parameter self-tuning steps at a preset parameter tuning cycle during normal operation. The diagram consists of two sub-plots: the left sub-plot's horizontal axis represents the maximum bus voltage drop in volts, and the vertical axis represents the voltage recovery time in milliseconds; the right sub-plot's horizontal axis represents the iteration number in cycles, and the vertical axis represents the optimal fitness of the joint cost function in dimensionless units.

[0082] In the Pareto front scatter plot on the left, the blue scatter points represent the numerous individuals in the population with various control parameter combinations generated by the genetic algorithm during the iteration process. These individuals are widely distributed in a two-dimensional space with voltage drop amplitudes of 20 volts to 50 volts and recovery times of 40 milliseconds to 140 milliseconds. The red broken lines and asterisks indicate the Pareto front formed after multi-objective screening.

[0083] The distribution trend of the red leading edge curve shows that there is a mutually restrictive game relationship between the two technical indicators of maximum bus voltage drop and voltage recovery time. When the voltage drop is limited to about 20 volts, the system's voltage recovery time will be extended to about 100 milliseconds. However, when a faster recovery time of 40 milliseconds is desired, the voltage drop will be relaxed to about 50 volts. This provides the system with a variety of optimal compromise control parameter combinations for different operating conditions.

[0084] In the fitness convergence curve plot on the right, the continuous solid green line represents the evolution trajectory of the genetic algorithm towards minimizing the joint cost function over 100 iterations. The green curve shows a rapid decline during the initial 1st to 20th generations, with the fitness value dropping rapidly from an initial 120 to around 35, indicating that the algorithm eliminated inferior proportional-integral parameter combinations and feedforward compensation gain parameters during this stage. From the 20th to the 100th generation, the rate of decline of the green curve gradually flattens and eventually stabilizes around 25.

[0085] This convergence process objectively proves that the genetic algorithm can stably solve for the control parameters that minimize the linear joint cost function, giving the underlying nonlinear electric disturbance state observer and fuzzy adaptive feedback controller the ability to continuously evolve.

[0086] Addressing the limitations of existing microgrid control technologies, such as centralized communication delays, lagging response of lower-level controls, and insufficient capacity of single feedback mechanisms to handle complex operating conditions, this embodiment discloses a process control method for suppressing transient power fluctuations in integrated photovoltaic-storage-charging power plants. This method provides a three-dimensional collaborative solution encompassing a three-level spatial architecture and multiple temporal scales. Through accurate prediction using upper-level deep learning and gradient boosting trees, combined with mid-level dual-closed-loop fuzzy state guidance, and nonlinear superposition calculation of feedforward and feedback compensation voltage commands at the lower level, the system can proactively implement physical interception under conditions such as sudden load changes, significantly reducing the voltage drop depth and recovery time of the bus voltage.

[0087] Meanwhile, with the addition of multi-dimensional fault identification based on time thresholds, non-disruptive parameter weighted switching algorithms, and parameter self-tuning mechanisms driven by joint cost functions, the entire integrated power station can still ensure the safe and reliable output of the underlying power conversion even when experiencing equipment aging or sudden link failures, effectively improving the stable operation capability and long-term economic benefits of the new energy microgrid system in extreme operating environments.

[0088] Example 2:

[0089] In the long-term practical operation of integrated photovoltaic-storage-charging power stations, the hybrid energy storage system, including power-type and energy-type energy storage units, inevitably undergoes aging of its internal physicochemical structure after tens of thousands of charge-discharge cycles. This physical aging is mainly manifested in the thickening of the solid electrolyte interface on the electrode surface, the dissolution of active materials, and the consumption of electrolyte. The degradation of these internal chemical mechanisms directly leads to an increase in the battery's equivalent ohmic internal resistance and a decrease in its internal ion diffusion capacity.

[0090] Because the nonlinear electrical disturbance state observer relied upon by the local controller of the underlying converter in Embodiment 1 to perform microsecond-level control is constructed based on the physical model used during the initial system setup, as the physical components age, the originally accurate built-in model parameters of the observer will deviate significantly from the actual physical device states. This technical problem of model mismatch in the nonlinear electrical disturbance state observer will cause the feedforward compensation command calculated by the underlying controller to exhibit severe lag or attenuation in response amplitude and phase when facing sudden transient power fluctuations, thereby triggering DC bus voltage instability.

[0091] To address the control mismatch problem caused by this long-term evolution, this embodiment discloses an observer adaptive reconstruction step based on the evolution of health status.

[0092] Specifically, the method provided in this embodiment includes a low-frequency state monitoring and parameter identification link that runs parallel to high-frequency real-time control. Within a preset state monitoring period, the upper-level global optimization controller extracts the historical voltage and current sequences of the power-type and energy-type energy storage units during the steady-state charging and discharging phases.

[0093] In practical engineering applications, the preset state monitoring period is typically set to encompass several complete charge-discharge depth cycles to ensure the statistical significance and stability of the acquired data. The steady-state charge-discharge phase is strictly defined as the operating range where the rate of change of charge-discharge current is extremely low and the system is not subjected to external transient power surges. Data extraction within this phase effectively filters out electromagnetic interference signals caused by high-frequency switching actions of the power electronic converter and spike voltage data caused by transient capacitor charging and discharging, thereby ensuring that the recorded historical voltage and current sequences purely and accurately reflect the slowly changing electrochemical polarization and ohmic impedance characteristics within the battery. The upper-level global optimization controller caches this time-series data through its internal large-capacity storage module, providing high signal-to-noise ratio data samples for subsequent offline or semi-online parameter identification.

[0094] For example, after obtaining a high-quality data sequence, the historical voltage and current sequences are input into a preset equivalent circuit identification model. The model is then used to solve online for the current identification values ​​of the equivalent ohmic internal resistance and polarization time constant of the power-type and energy-type energy storage units. The preset equivalent circuit identification model is primarily based on fractional or integer-order Thevenin equivalent topology, which abstracts the complex internal chemical reactions of the battery into a dynamic network of resistors and capacitors.

[0095] For example, the algorithmic formula for calculating the discrete-time states within the pre-set equivalent circuit identification model is given here:

[0096] ;

[0097] In the formula: Indicates the first The estimated terminal voltage variables for each discrete sampling point are calculated from the equivalent circuit model. Indicates the first The open-circuit voltage scalar obtained by mapping open-circuit voltage to state of charge at each discrete sampling point; Indicates the first The actual input current values ​​extracted from the historical current sequence by a discrete sampling point; This indicates the current identified value of the equivalent ohmic internal resistance that the model needs to iteratively approximate. Its physical meaning is the pure ohmic impedance generated by the current collector, electrolyte and separator inside the battery. Indicates the first -1 discrete sampling points to derive polarization voltage state variables; This indicates the discrete sampling time step used by the upper-level global optimization controller when processing historical sequences; The polarization time constant, which the model needs to iteratively approximate, is currently identified and its physical meaning is an inertial time parameter characterizing the internal ion diffusion and charge transfer rate. This represents the polarization resistance auxiliary variable obtained synchronously by the identification algorithm.

[0098] The upper-level global optimization controller continuously adjusts the algorithm using the least squares method or particle swarm optimization. and The value of makes the estimated terminal voltage variable in the model output [the value of is missing]. The root mean square error between the actual historical voltage sequence and the error is minimized. When the error converges to within the tolerance band, the accurate extraction of the current identified value of the equivalent ohmic internal resistance and the current identified value of the polarization time constant is completed.

[0099] like Figure 5 This figure illustrates the physical aging characteristics of a hybrid energy storage system during long-term service. The horizontal axis represents the number of charge-discharge cycles, ranging from 0 to 5000 cycles; the left vertical axis represents the equivalent ohmic internal resistance, in milliohms; and the right vertical axis represents the polarization time constant, in seconds.

[0100] The figure contains two core evolution curves. The continuous solid red line represents the evolution trajectory of the equivalent ohmic internal resistance with the number of cycles, while the discrete dashed blue line represents the evolution trajectory of the polarization time constant with the number of cycles.

[0101] Observing the waveform changes of the two curves, it can be found that in the 0 to 1000 cycle range during the initial operation of the power station, both the red solid line and the blue dashed line remain at a relatively flat low level, the equivalent ohmic internal resistance is stable at about 15 milliohms, and the polarization time constant is maintained at about 20 seconds, indicating that the electrochemical reaction activity inside the battery is in a healthy operating range.

[0102] As the number of charge-discharge cycles gradually increases and continues to advance towards 5000 cycles, the red solid line shows a significant nonlinear upward acceleration trend due to the physical effects of the thickening of the solid electrolyte phase interface on the electrode surface and the dissolution of active materials, eventually climbing to around 45 milliohms; at the same time, the blue dashed line also shows a synchronous nonlinear steep increase, eventually climbing to around 80 seconds, objectively reflecting the decrease in the ion diffusion capacity inside the battery and the enhancement of the control response hysteresis effect.

[0103] This long-term evolutionary data correlation characteristic intuitively demonstrates the engineering necessity of introducing low-frequency state monitoring and equivalent circuit identification links in the control system. The upper-level global optimization controller accurately extracts and records these internal resistance and time constant data that drift slowly with the number of cycles, and then maps them to generate the feedforward gain compensation coefficient and phase lead angle compensation coefficient required by the lower-level control. This enables the lower-level converter local controller to adaptively reconstruct the internal matrix of the observer according to the aging evolution law revealed in the figure, effectively overcoming the control mismatch phenomenon that occurs in the later stages of the battery life cycle of the traditional fixed parameter observer model, and ensuring the transient suppression capability of the photovoltaic-storage-charging integrated power station to cope with sudden power surges throughout the entire life cycle.

[0104] Optionally, after extracting the physical state parameters, the system needs to map these low-frequency physical parameters characterizing the battery's health state into mathematical adjustment parameters in the high-frequency control loop. In this step, the mid-layer hybrid energy storage controller receives the current identified values ​​of the equivalent ohmic internal resistance and polarization time constant, inputs them into a preset parameter correction mapping matrix, calculates and generates feedforward gain compensation coefficients characterizing the compensation amplitude, and phase lead angle compensation coefficients characterizing the compensation response speed, and sends them to the underlying converter local controller. The preset parameter correction mapping matrix is ​​a multi-dimensional linear or nonlinear spatial mapping transformation operator, whose core function is to decouple the complex cross-influence between physical variables and control variables.

[0105] For example, to demonstrate the core logic of this mapping calculation, the algorithm formula for the parameter-corrected mapping matrix is ​​given:

[0106] ;

[0107] In the formula: The feedforward gain compensation coefficient, calculated by mapping, is used to calibrate the output strength of the underlying feedforward control. The phase lead angle compensation coefficient, calculated by mapping, is used to adjust the time axis offset of the underlying feedforward control output command. This represents the basic feedforward gain calibration value determined through system commissioning during the initial stage of power plant grid connection; This represents the basic phase offset value determined at the initial stage of power plant grid connection; This represents the first positive real weight on the main diagonal of the mapping matrix, used to quantify the sensitivity of impedance drift to gain requirements; It represents the second positive real weight on the main diagonal of the mapping matrix, and the negative operation is used to construct the reverse compensation logic in control theory; This is the current identification value of the equivalent ohmic internal resistance passed from the upper layer; This represents the initial reference value of the equivalent ohmic internal resistance recorded in the controller's non-volatile memory; This is the current identified value of the polarization time constant passed from the upper layer; This represents the initial reference value of the polarization time constant recorded in memory.

[0108] It is important to note that when the underlying converter local controller performs the nonlinear electrical disturbance feedforward compensation, it dynamically adjusts the amplitude of the internal feedforward gain matrix of the nonlinear electrical disturbance state observer using the feedforward gain compensation coefficient, and dynamically corrects the output phase of the feedforward compensation voltage command using the phase lead angle compensation coefficient. In the microsecond-level real-time calculation process at the underlying level, the nonlinear electrical disturbance state observer generates a compensation signal by observing the state error. Its internal state transition matrix and input-output matrix determine the system's sensitivity to grid disturbances. The underlying controller directly injects the received feedforward gain compensation coefficient as a scalar multiplier into the gain matrix of the observer's computational core.

[0109] Meanwhile, before the bottom-level controller outputs the final feedforward compensation voltage command into the pulse width modulation module, it uses the phase lead angle compensation coefficient to perform discrete domain shifting operations on the command sequence or uses a coordinate rotation digital computer algorithm to perform vector angle deflection, thereby achieving microsecond-level lead action.

[0110] Specifically, the directional constraint relationship between parameters can be clearly derived from the algebraic structure of the above parameter correction mapping matrix algorithm formula: the value of the feedforward gain compensation coefficient is positively correlated with the change of the current identified value of the equivalent ohmic internal resistance, and the value of the phase lead angle compensation coefficient is negatively correlated with the change of the current identified value of the polarization time constant.

[0111] like Figure 6 This diagram illustrates the three-dimensional spatial mapping logic of a mid-layer hybrid energy storage controller, which transforms low-frequency physical parameters of battery health status into high-frequency control loop adjustment parameters. The horizontal axis represents the current identified equivalent ohmic internal resistance in milliohms, ranging from 15 to 45; the vertical axis represents the current identified polarization time constant in seconds, ranging from 20 to 80; and the vertical axis represents the compensation coefficient value, in dimensionless or degree units, used to uniformly measure the two different compensation parameters.

[0112] The figure contains two three-dimensional surfaces exhibiting nonlinear evolution characteristics. The red surface, which is located above and has an upward convex trend, represents the feedforward gain compensation coefficient, while the blue surface, which is located below and has a downward concave trend, represents the phase lead angle compensation coefficient.

[0113] Observing the spatial waveform transformation of the graph, it can be seen that at the starting position of the coordinate axis, that is, in the healthy state with an equivalent internal resistance of 15 milliohms and a polarization time constant of 20 seconds, the red surface is at the reference position with a height of 1.0, and the blue surface is at the reference position with a height of 0, indicating that the system does not need to perform adaptive reconstruction correction at this time.

[0114] As the physical aging of the battery intensifies, and the values ​​on the horizontal and vertical axes gradually shift towards 45 milliohms and 80 seconds, the red surface exhibits a significant positive correlation and rises along the spatial diagonal, eventually reaching a region with a height close to 2.5.

[0115] Meanwhile, the blue surface exhibits a negative correlation and a downward trend, with its value gradually converging to around -15. This hyperboloid spatial structure, with its upper and lower surfaces separated in opposite directions, objectively confirms the system's multi-dimensional mapping decoupling mechanism.

[0116] The system utilizes a positively increasing red surface to amplify the observer's output response amplitude to offset the energy loss caused by increased internal resistance, while simultaneously employing a negatively decreasing blue surface to provide a more significant phase lead bias to counteract polarization hysteresis. This synchronously executed amplitude amplification and phase lead adjustment mechanism ensures that the underlying converter maintains stable suppression performance when dealing with transient power fluctuations.

[0117] To illustrate this with a practical physical application: As batteries age, their actual physical equivalent ohmic internal resistance increases. This means that when outputting the same transient compensation power to the external grid, a larger voltage drop occurs inside the battery. If the gain of the underlying controller remains constant, the actual compensation current generated by its driving converter will fail to meet the expected target. Therefore, the system uses positive correlation mapping logic to ensure... As the internal resistance increases, the amplitude is amplified synchronously, thereby forcibly increasing the output response amplitude of the observer to offset the energy loss caused by the increase in internal resistance.

[0118] Similarly, when the polarization time constant increases, it means that the tailing effect of the internal chemical reaction of the battery is enhanced, and the response of the external output power becomes sluggish. If the compensation signal is still issued according to the phase at the initial site setup, the actual current waveform will inevitably show phase lag. In order to counteract this lag characteristic caused by electrochemical polarization, the system, based on negative correlation logic, reduces or even assigns a more extreme reverse compensation value to the phase lead angle compensation coefficient as the polarization time constant increases. This forces the underlying controller to start the ramp-up calculation of the control command earlier on the time axis, using the time advance of the control algorithm to compensate for the time lag of the physical devices.

[0119] A comprehensive evaluation of existing technologies reveals that current microgrid centralized control strategies or conventional observer control methods often treat the controlled object of the underlying converter—the energy storage battery—as an ideal voltage source with constant parameters or a linear object with a fixed first-order inertial element. This technological bias leads to a significant reduction in the system's ability to suppress transient fluctuations after several years of operation, especially when facing aging energy storage battery arrays. Furthermore, it frequently causes more severe bus voltage oscillations due to command overshoot or response lag.

[0120] The method provided in this embodiment constructs an adaptive dynamic reconfiguration mechanism that spans time scales and control levels. Physical parameters are obtained through lossless state identification at a large time scale at the upper level, the transformation of mechanistic characteristics into control variables is achieved through matrix mapping at the middle level, and finally, high-frequency reconfiguration is performed in the real-time computation loop at the microsecond level at the lower level. This approach enables the entire photovoltaic-storage-charging integrated power station to maintain extremely stable transient disturbance suppression performance throughout its decades-long lifespan. It not only effectively improves power quality and grid interaction capabilities but also significantly reduces the operation and maintenance costs associated with manually periodically shutting down the system to calibrate parameters, demonstrating significant engineering practical value and real-world significance.

[0121] Example 3:

[0122] To address the technical problem that transient power fluctuation suppression under heavy load conditions can easily trigger the hardware thermal protection shutdown of power electronic converters, the method described in this embodiment further includes transient current limiting and load compensation steps based on electrothermal coupling dynamic safety boundaries.

[0123] In real-world industrial operating scenarios, such as during the high temperatures of summer when multiple AC / DC charging piles in a power station are operating at full capacity concurrently, the power devices inside the power electronic converter of the hybrid energy storage system, such as insulated-gate bipolar transistors (IGBTs) or silicon carbide metal-oxide-semiconductor field-effect transistors (MOSFETs), are already in a state of continuous high heat generation. If a sudden transient power disturbance occurs in the DC microgrid at this time, the nonlinear electrical disturbance feedforward compensation algorithm in the aforementioned embodiment will instantly issue a current compensation command with an extremely large amplitude. If the underlying controller blindly executes this command, the heat generated by the huge transient current will instantly break through the physical junction temperature limit of the power devices, triggering hardware-level thermal overload protection, causing the converter to be forcibly shut down, and consequently resulting in severe instability of the entire DC microgrid bus.

[0124] To address this technical problem of electrothermal coupling limit mismatch, this embodiment discloses the following:

[0125] Specifically, the underlying converter local controller acquires the substrate temperature of the power electronic converter in real time, and combines it with a preset dynamic thermal resistance network model and real-time current. The system continuously estimates the instantaneous virtual junction temperature of the power devices inside the power electronic converter online. In hardware structures, temperature sensors are typically mounted on the surface of heat sinks or power module substrates, resulting in significant physical heat conduction delays and failing to represent the true transient temperature of the internal semiconductor bare die. Therefore, an algorithm must be introduced for online state estimation.

[0126] It is important to note that in the programming specifications for precision digital control systems of power electronic converters, to ensure the rigor of mathematical logic and the uniqueness of engineering implementation, the real-time current used to characterize the physical feedback signal and the actual current used to control the desired output after algorithm processing must be strictly separated and independently defined. This mandatory symbolic isolation is to avoid memory variable reuse and conceptual confusion in high-frequency interrupt calculations in multi-threaded systems; otherwise, it will lead to uncontrollable duty cycles in the controller output, causing extremely serious catastrophic consequences such as system failure.

[0127] Therefore, in this embodiment, the real-time current is explicitly defined. The current is the instantaneous feedback current of the physical circuit acquired by the sensor within the current control sampling period, which only represents the objective physical state that has already occurred; the actual current defined later... This represents the action instruction to be executed after logical judgment. Based on the above physical parameters, the underlying converter local controller uses the following preset dynamic thermal resistance network model algorithm formula to derive and estimate the instantaneous virtual junction temperature online:

[0128] ;

[0129] In the formula: The calculated instantaneous virtual junction temperature is used to characterize the transient temperature level of the chip core inside the power device. This indicates the temperature of the substrate measurement point obtained through a high-precision thermistor; This represents the linear correlation coefficient of switching losses extracted in advance by fitting data from the device datasheet. This refers to the real-time current as strictly defined above; This represents the pre-fitted quadratic correlation coefficient of conduction loss; This represents the transient thermal impedance parameter used to characterize the transient thermal conduction hysteresis characteristics of the packaging material.

[0130] Using this formula, the underlying controller can map the actual temperature evolution trajectory inside the device in the microcontroller's memory at a microsecond-level refresh rate, without the need to implant expensive bare-chip sensors, by utilizing known measurement point temperatures and electrical loss mechanisms.

[0131] For example, after obtaining the real-time virtual junction temperature, the system subtracts the real-time virtual junction temperature from the preset device junction temperature alarm limit. The strictly non-negative numerical difference was calculated and defined as the dynamic heat margin. Subsequently, the dynamic thermal margin was... The input of a preset electrothermal dimension reduction mapping function is transformed into the maximum transient compensation current threshold that the power electronic converter is allowed to output within the preset third control cycle. The physical essence of this step is to equivalently reduce the remaining safety space in the temperature domain to the operating boundary of the current domain.

[0132] For example, to clearly illustrate this mapping mechanism, the algorithmic formula for the electrothermal dimensionality reduction mapping function is given:

[0133] ;

[0134] In the formula: The calculated maximum transient compensation current threshold is physically defined as the upper limit of the instantaneous overload current that the device can withstand without triggering hardware over-temperature shutdown. This refers to the dynamic thermal margin calculated above, i.e., the preset device junction temperature alarm limit value and... The difference.

[0135] This formula uses the root-finding logic of a quadratic equation to derive, in reverse, the critical current value that can be injected in the current control cycle under a given short-time thermal impedance and remaining temperature margin, providing an accurate numerical reference for subsequent limiting operations.

[0136] It is also important to note that the system subsequently compares the equivalent current of the control command generated by the nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control with the maximum transient compensation current threshold. The equivalent current of the control command refers to the absolute value of the ideal compensation current calculated by the upper-level algorithm purely from the perspective of grid voltage stabilization requirements. This comparison process constitutes a dynamic constraint relationship between grid-side requirements and physical hardware limits.

[0137] like Figure 7 This diagram illustrates the hardware protection and current limiting / cutoff process of the underlying converter's local controller when facing extreme heavy loads and transient power disturbances. The horizontal axis of the graph represents time in microseconds, with the observation window recording the transient control range from 0 to 500 microseconds; the vertical axis represents current in amperes, ranging from zero to 600 amperes; the color bar on the right indicates the real-time virtual junction temperature represented by the background, in degrees Celsius.

[0138] The background of the image is a pseudo-color thermal bitmap reflecting the internal temperature distribution of the power device. The color gradually transitions from dark blue at the bottom to dark red at the top, visually representing the heat that accumulates rapidly in the semiconductor die core as the current increases. The dark red area corresponds to the alarm limit danger zone that may trigger the hardware over-temperature protection shutdown.

[0139] Three waveform curves with clear physical meanings are superimposed on the thermal bitmap. The red discrete dashed line represents the real-time current feedback from the physical loop, the black dotted line represents the maximum transient compensation current threshold transformed by the electrothermal dimension reduction mapping function, and the blue continuous solid line represents the actual current that is finally released after dynamic safety boundary judgment.

[0140] Observing the waveform evolution trajectory in the figure, it can be found that around 250 microseconds, due to the transient impact of the microgrid, the unrestricted real-time current rapidly surged and exceeded 500 amperes, directly penetrating the dark red high-risk area at the top of the heat map. If the instruction is directly allowed according to the conventional feedback logic, there is an extremely high risk of causing hardware thermal breakdown failure.

[0141] The underlying controller continuously estimates the dynamic thermal margin online, defining a safe boundary for the maximum transient compensation current threshold in the range of approximately 450 amperes. This threshold exhibits a slightly decreasing dynamic tightening trend over time.

[0142] When the real-time red current exceeds the threshold boundary, the control logic automatically triggers the current limiting mechanism, forcibly clamping the actual output current to the black threshold boundary, thus forming a blue truncated waveform with the top limited.

[0143] This electrothermal coupling-based dimensional reduction and limiting operation ensures that the initial transient voltage supports the response speed while effectively isolating the actual output current within the hardware safety operating range. This reduces the probability of the energy storage converter being accidentally disconnected from the grid due to overheating protection from the physical level. Furthermore, the reduced transient power deficit is transferred to the charging pile side for collaborative digestion through load compensation requests by the system's middle-level controller, significantly improving the operational stability of the photovoltaic-storage-charging integrated power station under extreme and harsh conditions.

[0144] Optionally, when the determination result is that the equivalent current of the control command does not exceed the maximum transient compensation current threshold, the equivalent current of the control command is directly used as the actual current. Output; wherein, the actual current Defined as the actual physical output current that ultimately acts on the pulse width modulation duty cycle generation stage after safety boundary determination. Under normal transient fluctuation scenarios, the underlying power electronic converter has sufficient thermal capacity margin, so it can directly and normally execute the anti-disturbance compensation command issued by the system to smooth out the voltage fluctuation of the DC bus as quickly as possible.

[0145] Specifically, when the determination result indicates that the equivalent current of the control command exceeds the maximum transient compensation current threshold, it means that executing the original command will inevitably lead to hardware thermal breakdown. In this case, the underlying converter local controller will adjust the actual current... The system forcibly clamps and limits the current to the maximum transient compensation current threshold, calculates the transient power deficit based on the numerical difference between the equivalent current of the control command and the maximum transient compensation current threshold, and simultaneously generates a load compensation request signal carrying the transient power deficit and reports it to the mid-layer hybrid energy storage controller.

[0146] Under this extreme operating condition, the primary task of the underlying controller shifts from voltage regulation and degradation to ensuring the operation of the hardware devices, by forcibly reducing the physical output current. Suppressed at the safety threshold The following ensures that the energy storage converter itself will not disconnect from the grid due to thermal protection. At the same time, the original suppressed power loss caused by the clamping operation is not ignored, but is quantified as a transient power deficit and sent to the upper-level controller as a compensation request signal via the high-speed industrial fieldbus.

[0147] Optionally, after receiving the load compensation request signal, the mid-level hybrid energy storage controller reduces the current load output command of the AC / DC charging pile according to the transient power deficit ratio. In the application ecosystem of integrated photovoltaic-energy storage-charging power stations, the power battery of electric vehicles is an interruptible load with extremely high flexibility and adjustability potential. Since electric vehicle charging is a long-cycle process lasting tens of minutes or even hours, fine-tuning its charging current within a transient range of hundreds of milliseconds has no substantial impact on the charging experience of end users or the health of vehicle batteries. Based on the received transient power deficit, the mid-level controller issues short-term power reduction commands to each charging pile through the digital communication interface, forcing the DC-DC converter inside the charging pile to reduce its duty cycle. This method of compensating for the power deficiency caused by source-end thermal limiting through active load reduction on the load side achieves dynamic power replacement in physical space.

[0148] A comprehensive evaluation based on existing technologies reveals that current microgrid control systems typically isolate the voltage regulation control loop from the hardware thermal protection loop. Their operational logic often involves the voltage regulation control loop drawing power indefinitely until a hardware temperature threshold is triggered, at which point it is forcibly shut down. This crude protection mechanism is highly susceptible to triggering cascading system shutdowns under heavy loads.

[0149] The method provided in this embodiment introduces a collaborative compensation mechanism between electrothermal coupling dynamic safety boundaries and spatial load. By establishing a rigorous real-time virtual junction temperature estimation and electrothermal dimensionality reduction mapping model, the system can dynamically transform the originally independent thermal limit boundaries into real-time limiting commands for high-frequency current control loops, effectively balancing transient voltage support capabilities and the survivability of the underlying hardware. Simultaneously, it strictly distinguishes between... and The variable definitions avoid the risk of data cross-contamination at the software architecture level. After the safety limit is triggered, this solution does not passively bear the consequences of voltage drop, but instead utilizes the inherent source-load interaction attribute of the photovoltaic-storage-charging integrated power station to transfer the transient power deficit that cannot be borne by the energy storage unit to the flexible charging load side for collaborative compensation at a speed of milliseconds.

[0150] This overall technical solution not only ensures the safe operation of the core power electronic converter under extreme heavy load conditions, but also significantly enhances the overall system-level capability of the power plant to withstand large-scale transient power surges without increasing any additional hardware investment through load-side spatial compensation, thus possessing extremely outstanding engineering application value.

[0151] Example 4:

[0152] This embodiment discloses a transient power fluctuation suppression process control system for an integrated photovoltaic, energy storage and charging power station. The control system is applied to a DC microgrid integrated photovoltaic, energy storage and charging power station to solve the technical problems of transient impact on the common DC bus voltage caused by sudden changes in photovoltaic output, high-power load impact of fast charging piles, and multi-dimensional faults under non-ideal operating conditions.

[0153] In actual power plant physical engineering, the control system is closely connected with the photovoltaic inverter equipment in the DC microgrid, the hybrid energy storage bidirectional converter, the bidirectional active bridge converter inside the AC / DC charging pile, and the high-precision voltage and current sensors deployed at various measuring points on the common DC bus, and realizes closed-loop control of the data link through the industrial network.

[0154] Specifically, the transient power fluctuation suppression process control system of the integrated photovoltaic-storage-charging power station includes an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller that are interconnected, forming a vertically coordinated three-level controller architecture. At the physical cabling and topology connection level, the upper-level global optimization controller and the middle-level hybrid energy storage controller, as well as the middle-level hybrid energy storage controller and the lower-level converter local controller, are typically connected via bidirectional dual-communication links using fiber optic Ethernet switch networks, industrial RS485 buses, or controller area network (CAN) buses to carry the transmission of control parameters at different time scales.

[0155] Specifically, the upper-level global optimization controller is manifested in hardware as an edge computing server or industrial-grade computer deployed in the power plant's central control room. This upper-level global optimization controller acquires multi-source operating data from the power plant, performs multi-source fusion power prediction in a preset first control cycle, generates an optimization strategy including total power adjustment instructions and pre-adjustment instructions, and sends it to the middle-level hybrid energy storage controller. In actual industrial operations, the Ethernet communication interface of the upper-level global optimization controller is sequentially connected to the database server of an external meteorological service platform, the AC / DC charging pile operation management system within the power plant, and the monitoring terminal of the photovoltaic power generation unit via network protocols to collect historical photovoltaic power, real-time temperature, irradiance, fast-charging load demand, and user reservation data as multi-source operating data.

[0156] Since long-term macroscopic power prediction involves a large number of nonlinear floating-point matrix operations, the upper-level global optimization controller is equipped with a large-capacity random access memory and a graphics processing chip. It calls a pre-set timing prediction algorithm model in its internal memory every preset first control cycle (e.g., 1 minute) to calculate the multi-source fusion power prediction curve for the future time period. Based on the difference between this curve and the calculated bus balance equation, the upper-level global optimization controller calculates and generates the total power adjustment command capable of addressing macroscopic power deficits, as well as the pre-adjustment command for slow-speed batteries. These commands are then encapsulated into standardized data frames and sent to the middle-level hybrid energy storage controller.

[0157] For example, the mid-level hybrid energy storage controller is manifested at the hardware level as a microgrid central controller (MGCC) or a high-performance programmable logic controller (PLC) independently installed in a control cabinet. The mid-level hybrid energy storage controller is used to execute dual-closed-loop fuzzy hybrid energy storage power allocation based on the total power adjustment command in the optimization strategy during a preset second control cycle. This generates a first power reference command for power-type energy storage units and a second power reference command for energy-type energy storage units, and sends these commands to the local controller of the underlying converter. The hardware core of the mid-level hybrid energy storage controller consists of a multi-core digital signal processor (DSP) or an advanced reduced instruction set machine (ARM) chip, possessing medium-scale fast data processing capabilities.

[0158] After receiving the optimization strategy transmitted from the upper layer, the mid-layer hybrid energy storage controller reads the instantaneous state-of-charge variables of the power-type and energy-type energy storage media currently uploaded via the CAN bus within a preset second control cycle (e.g., 100 milliseconds) driven by its internal discrete control timer. Subsequently, it calls its built-in fuzzy control rule library table to perform a dual-closed-loop fuzzy hybrid energy storage power allocation calculation, dynamically decoupling the first power reference command belonging to the power-type energy storage unit (such as a lithium iron phosphate battery pack or supercapacitor cabinet) and the second power reference command belonging to the energy-type energy storage unit (such as a vanadium redox flow battery pack). These power reference commands are refreshed at high frequency via the fieldbus and distributed to the underlying power electronic converters.

[0159] It is also important to note that the underlying converter local controller is directly manifested at the hardware level as a dedicated control board embedded within each bidirectional power storage converter (PCS) and charging converter. This underlying converter local controller, based on the first and second power reference commands, performs nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control at a preset third control cycle to generate a pulse-width modulation duty cycle signal to drive the power electronic converter for power regulation. The duration of the third control cycle is shorter than the duration of the second control cycle, and the duration of the second control cycle is shorter than the duration of the first control cycle. The core microprocessor of the underlying converter local controller is typically a field-programmable gate array (FPGA) or a high-speed microcontroller operating at frequencies above several hundred megahertz. It is directly connected to a high-speed analog-to-digital converter (ADC) in the hardware sampling circuit, directly reading the real-time voltage feedback of the common DC bus and the physical current feedback of each branch at a preset third control cycle (e.g., 100 microseconds).

[0160] Based on these high-frequency measurement signals, the local controller of the underlying converter executes a nonlinear electrical disturbance feedforward compensation algorithm in its internal multiply-accumulate unit to generate a feedforward term to offset transient impacts. Simultaneously, it utilizes the voltage error to execute a fuzzy adaptive feedback control algorithm, generating a feedback correction term. After superimposing the feedforward term and the feedback correction term, the hardware timer inside the underlying controller directly outputs a pulse width modulation (PWM) signal with the corresponding duty cycle. This signal, through an opto-isolated drive circuit, directly acts on the gate of the power electronic converter's switching transistor, which is composed of an insulated-gate bipolar transistor (IGBT) or a silicon carbide metal-oxide-semiconductor field-effect transistor (SiCMOSFET), thereby completing the physical throughput regulation of transient power imbalances within a time window of tens of milliseconds or even milliseconds.

[0161] Specifically, in order to cope with sudden non-ideal operating conditions such as abnormal electromagnetic environment or deterioration of communication physical link, the three-level controller architecture in the system is also used to synchronously perform multi-dimensional online fault identification within the above control cycle, classify faults into first-level faults, second-level faults and third-level faults according to the fault duration scale, and perform non-disruptive degradation control that matches the time scale and control level to achieve closed-loop suppression of transient power fluctuations.

[0162] In the software-safe thread operation of the control system, controllers at all levels monitor the status of hardware interfaces in real time through internal timer interrupts and heartbeat packet verification algorithms. Once a sensor signal sampling anomaly, disconnection, or data packet loss between control levels is detected, the system immediately triggers fault identification logic. If the fault duration is determined to be a first-level fault of transient interference (less than 100 milliseconds), the local controller of the underlying converter activates a fault-tolerant state observer to perform local virtual signal replacement, preventing distortion in the control loop.

[0163] If the duration of the fault is determined to be a second-level fault of hierarchical interruption, such as between 100 milliseconds and 1 minute, the local controller of the bottom converter will autonomously degrade to break away from the middle layer instruction dependency and switch to pure electric autonomous mode operation, or the power distribution matrix of the remaining healthy units will be reconstructed by the middle layer hybrid energy storage controller.

[0164] If the duration of the fault is determined to be a Level 3 fault, which is a macro-chain failure, such as more than 1 minute, the upper-level global optimization controller will trigger an offline historical data reconstruction mechanism to replace the missing real-time external meteorological information.

[0165] All degradation switching processes are smoothly completed in the software state machines of the upper, middle and lower level controllers through a weighted superposition interpolation algorithm, thereby avoiding secondary electrical shocks to the common DC bus caused by sudden changes in control parameters.

[0166] Based on the control system disclosed in this embodiment, which consists of various practical industrial control devices and bus networks, the integrated photovoltaic-storage-charging power station possesses extremely high disturbance rejection redundancy through cross-level physical connections at the hardware level and multi-time-scale algorithm embedding at the software level. This control system seamlessly connects the long-cycle macroscopic prediction advantages of the upper layer, the dynamic power shunting characteristics of the middle layer, and the microsecond-level nonlinear electrical control response of the lower layer. While achieving high-performance power voltage regulation and suppression, it constructs a multi-layered hardware and software fault-tolerant protection barrier, ensuring the overall operational safety of the DC microgrid system.

[0167] Example 5:

[0168] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0169] like Figure 9 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0170] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0171] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0172] The memory 103 stores a computer program corresponding to the transient power fluctuation suppression process control method for the integrated photovoltaic-storage-charging power station of the above embodiments of the present invention. This computer program is controlled and executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the aforementioned method embodiments.

[0173] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 9 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0174] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for controlling transient power fluctuation suppression in an integrated photovoltaic-storage-charging power station, characterized in that, An integrated photovoltaic-storage-charging power station with a DC microgrid architecture featuring a three-level controller architecture, wherein the three-level controller architecture consists of an upper-level global optimization controller, a middle-level hybrid energy storage controller, and a lower-level converter local controller connected in communication; the method includes: The system acquires multi-source operation data of the power plant, executes multi-source fusion power prediction through the upper-level global optimization controller in a preset first control cycle, generates an optimization strategy that includes total power adjustment instructions and pre-adjustment instructions, and sends it to the middle-level hybrid energy storage controller. The mid-layer hybrid energy storage controller executes dual-closed-loop fuzzy hybrid energy storage power allocation based on the total power adjustment command in the optimization strategy during a preset second control cycle, generating a first power reference command for power-type energy storage units and a second power reference command for energy-type energy storage units, and sends them down to the bottom-layer converter local controller. The underlying converter local controller performs nonlinear electrical disturbance feedforward compensation and fuzzy adaptive feedback control based on the first power reference command and the second power reference command in a preset third control cycle to generate a pulse width modulation duty cycle signal to drive the power electronic converter to perform power regulation; wherein, the duration of the third control cycle is less than the duration of the second control cycle, and the duration of the second control cycle is less than the duration of the first control cycle. Multi-dimensional online fault identification is performed synchronously within the aforementioned control cycle. Faults are classified into first-level, second-level, and third-level faults according to the fault duration scale. Unobstructed degradation control that matches the time scale and control level is executed to achieve closed-loop suppression of transient power fluctuations.

2. The method according to claim 1, characterized in that, The specific steps of the multi-source fusion power prediction include: Acquire historical photovoltaic power generation, local environmental meteorological data, and real-time charging status; Input the above running data into the fusion prediction network model composed of a deep learning temporal prediction network and a gradient boosting tree model; The fusion prediction network model outputs a photovoltaic power output prediction curve and a charging pile load prediction curve for a preset future time period, and generates the total power adjustment command based on the calculated difference between the photovoltaic power output prediction curve and the charging pile load prediction curve. When a power deficit or surplus greater than a preset power threshold is predicted in the future, a pre-adjustment command is generated to control the energy storage unit to perform pre-discharge or pre-charge.

3. The method according to claim 1, characterized in that, The specific steps for the dual-closed-loop fuzzy hybrid energy storage power allocation include: The total power adjustment command and the state of charge of the power-type energy storage unit are input into the first fuzzy controller, and the total suppression coefficient is output through the first fuzzy controller; the total power adjustment command is multiplied by the total suppression coefficient to obtain the upper limit of the total output power of the hybrid energy storage. The state of charge and its rate of change of the state of charge of the energy storage unit are input into the second fuzzy controller, and the allocation coefficient is output through the second fuzzy controller. Multiply the upper limit of the total output power of the hybrid energy storage by the allocation coefficient to obtain the second power reference command output to the energy-type energy storage unit; The difference between the upper limit of the total output power of the hybrid energy storage and the value of the second power reference command is defined as the first power reference command output to the power-type energy storage unit.

4. The method according to claim 1, characterized in that, The specific steps for performing nonlinear electric disturbance feedforward compensation and fuzzy adaptive feedback control include: Real-time acquisition of the common DC bus voltage sag rate and charger current mutation rate of the integrated photovoltaic-storage-charging power station; The bus voltage drop rate and charger current mutation rate are input into the nonlinear electrical disturbance state observer to generate a feedforward compensation voltage command. Calculate the voltage error amplitude between the actual instantaneous voltage of the common DC bus and the preset common DC bus reference voltage, as well as the rate of change of the voltage error amplitude. Input the voltage error amplitude and the rate of change of the voltage error amplitude into a fuzzy adaptive proportional-integral-derivative controller to generate a feedback compensation voltage command. The feedforward compensation voltage command and the feedback compensation voltage command are superimposed to generate the pulse width modulation duty cycle signal.

5. The method according to claim 1, characterized in that, The classification criteria for the multi-dimensional online fault identification include: Faults whose duration is less than a first preset time threshold are classified as first-level faults, and first-level faults include sensor instantaneous faults and electromagnetic interference. Faults whose duration is greater than or equal to the first preset time threshold and less than the second preset time threshold are classified as second-level faults. The second-level faults include communication interruption between control levels and energy storage converter faults. Faults whose duration is greater than or equal to the second preset time threshold are classified as the third-level faults, and the third-level faults include interruption of meteorological platform data acquisition. Wherein, the first preset time threshold is less than the second preset time threshold.

6. The method according to claim 5, characterized in that, In the aforementioned non-disruptive degradation control, the processing logic for the first-level fault is as follows: When a transient fault of the sensor is identified through residual analysis, the local controller of the underlying converter activates the fault-tolerant state observer and uses the system mathematical model to estimate the output value of the faulty sensor in real time, so as to replace the physical acquisition signal for underlying control. When the system detects that the physical acquisition signal has returned to normal and has maintained a normal state for a duration exceeding a preset stability time threshold, the system will use the physical acquisition signal as a control input instead of the estimated value.

7. The method according to claim 5, characterized in that, In the aforementioned non-disruptive degradation control, the processing logic for second-level and third-level faults is as follows: When a communication interruption between the control levels is detected in the second-level fault, the local controller of the underlying converter independently switches to a pure electric operation mode and independently schedules each energy storage unit according to a pre-set reserve power allocation ratio. When the interruption of meteorological platform data acquisition is detected in the third-level fault, the upper-level global optimization controller switches to pure historical data prediction mode and uses the historical operating data of the same period in the past cycle cached by the system to generate a prediction curve to replace the real-time data prediction.

8. The method according to claim 1, characterized in that, The disturbance-free degradation control employs a weighted superposition disturbance-free parameter switching algorithm during the switching of operating modes, and its timing... The final control output The following general formula can be used to calculate: ； in, The preset transition switching period duration is matched with the fault level; This is the elapsed time since the start of the state transition, and ; The old control output parameters before the state transition starts; The new control output parameters after the state transition is completed; This is the bus voltage feedforward compensation correction amount used to smooth out transient voltages.

9. The method according to claim 1, characterized in that, After the non-disruptive degradation control is executed, the method further includes a graded verification self-healing control step: The system continuously monitors fault status signals. When the system detects that the signal has recovered and meets the verification conditions within a consecutive preset time period, it determines that the fault has been physically recovered. After determining that the fault has been physically restored and stable operation has reached the preset verification delay time, the system reactivates the weighted superposition non-disturbance parameter switching algorithm, and gradually switches the control mode from the non-disturbance degradation control state back to the normal three-level collaborative control state from the bottom up according to the preset algorithm.

10. The method according to claim 1, characterized in that, During normal operation, the method automatically performs control parameter self-tuning steps at a preset parameter tuning cycle: Using actual operational dynamic data within a preset historical time period as training samples, a linear joint cost function is constructed that includes the variable of maximum bus voltage drop and the variable of voltage recovery time. A genetic algorithm is used to minimize the linear joint cost function, and the feedforward compensation gain parameter of the nonlinear electric disturbance state observer and the proportional-integral parameter combination of the fuzzy adaptive feedback controller are obtained by iterative solution. After the iteration terminates, the combination of feedforward compensation gain parameters and proportional-integral parameters is output, and the current operating control parameters are overwritten and updated.

Images