Quick dynamic response method for energy storage converter based on model predictive control

Abstract

The invention relates to the technical field of energy storage converters, in particular to an energy storage converter rapid dynamic response method based on model predictive control. Comprising the following steps of system initialization, reference trajectory and disturbance feedforward generation, prediction model rolling prediction, multi-objective cost function evaluation, dynamic weight adjustment, constrained online rolling optimization and control quantity application and rolling execution. According to the invention, high-efficiency dynamic response speed and accurate control are realized, the conflict problem of multiple control targets in the dynamic process is fundamentally solved, intrinsic safety protection of millisecond-level response for controlling a bottom layer is provided for an energy storage unit, the service life of a battery is prolonged, and the intelligent network construction capability adaptive to a complex power grid environment is realized; and the engineering contradiction between high performance and low complexity is effectively balanced.

Description

Technical Field

[0001] This invention relates to the field of energy storage converter technology, specifically a fast dynamic response method for energy storage converters based on model predictive control. Background Technology

[0002] As the penetration rate of renewable energy sources such as wind and solar power in the power system continues to increase, the stability of the power grid faces new challenges. The intermittent and fluctuating energy output characteristics make the need for system frequency and voltage regulation increasingly urgent. Electrochemical energy storage systems, with their rapid power response and flexible four-quadrant operation, have become key equipment for mitigating renewable energy fluctuations and providing grid ancillary services. The energy storage converter, as the bridge connecting the energy storage battery and the grid, directly determines the strength of the energy storage system's grid support capability through its control performance.

[0003] In traditional control methods for energy storage converters, dual-loop control based on proportional-integral (PI) regulators (outer loop power / voltage loop, inner loop current loop) is widely used due to its simple design and high reliability. However, this control method has limitations. First, its control performance heavily relies on precise system parameter tuning, and its dynamic response speed and stability significantly decrease when facing conditions such as changes in grid impedance and operating point deviations. Second, PI control is essentially a "lag" correction based on past and current errors, making it difficult to cope with millisecond-level severe disturbances in the grid, such as voltage drops and frequency jumps, and failing to fully utilize the fast response of the energy storage converter. Furthermore, traditional control architectures typically separate the converter's fast power control from the long-term energy management of the energy storage battery (such as state of charge management), with intervention from the upper-level energy management system occurring at second or minute intervals. This slow intervention cannot prevent overcharging or over-discharging of the battery during instantaneous grid disturbances, posing a risk to the health and safety of the energy storage device.

[0004] To overcome the shortcomings of traditional control, Model Predictive Control (MPC), as an advanced control algorithm, has received widespread attention in the field of power electronic converters in recent years due to its ability to explicitly handle multivariable, constrained, and nonlinear system problems. MPC predicts system behavior over a future period by utilizing a discretized model of the system and calculates the optimal control action through online rolling optimization of an objective function, providing a new approach to improving the dynamic performance of energy storage converters.

[0005] However, directly applying MPC to grid-type energy storage converters and expecting them to simultaneously achieve rapid dynamic response, intrinsic safety protection, and intelligent mode switching still faces a series of challenges:

[0006] (1) Problem of singular control objectives: Most existing MPC schemes focus on improving current tracking quality or reducing switching losses, failing to organically integrate system-level objectives such as maintaining grid stability (e.g., voltage support) and ensuring the safety of energy storage (e.g., SOC overrun prevention) into the underlying fast control, resulting in multiple control objectives conflicting with each other in the dynamic process.

[0007] (2) Inflexible mode switching problem: In complex power grid environment, energy storage converters need to switch intelligently between grid-connected mode and grid-connected mode. Existing technologies usually use external logic for hard switching, which is prone to transient impact. There is a lack of a smooth internal control mechanism that can adapt to the power grid status.

[0008] (3) The problem of balancing computational complexity and performance: The computationally intensive MPC method is computationally burdensome and difficult to implement high-frequency control on low-cost microprocessors. No solution has yet been proposed to address this technical problem. Summary of the Invention

[0009] To address the problems in related technologies, this invention proposes a fast dynamic response method for energy storage converters based on model predictive control, thereby overcoming the aforementioned technical issues in existing technologies. The purpose of this invention is to achieve high-efficiency dynamic response speed and precise control, fundamentally solving the conflict problem of multiple control objectives in the dynamic process, providing intrinsic safety protection for energy storage units with millisecond-level response at the control layer, extending battery life, and possessing intelligent grid construction capabilities that adapt to complex power grid environments, effectively balancing the engineering contradiction between high performance and low complexity.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a fast dynamic response method for energy storage converters based on model predictive control, comprising the following steps:

[0011] S1, System Initialization: Preload the system parameters and safe operating boundaries of the energy storage converter;

[0012] S2, Real-time data acquisition and processing: In each control cycle, the AC side voltage and current and DC side voltage of the energy storage converter are acquired synchronously and their coordinates are transformed; at the same time, the real-time state of charge of the energy storage unit is obtained from the battery management system at a slower cycle.

[0013] S3, Reference Trajectory and Disturbance Feedforward Generation: Receive upper-level control commands, generate current and voltage reference values ​​for the current control cycle, and use the collected grid voltage as a measurable feedforward disturbance signal;

[0014] S4, Rolling Prediction Model: Based on the established discrete state-space prediction model of the energy storage converter, the system behavior state sequence at multiple future times is predicted in a rolling manner according to the current system state, candidate control sequence and disturbance prediction sequence.

[0015] S5, Multi-objective cost function evaluation: Construct a cost function to evaluate the performance of the candidate control sequence. The cost function includes at least a penalty for future tracking errors, a penalty for drastic changes in control quantity, a predictive penalty for out-of-bounds energy storage state of charge, and an evaluation of voltage support capability.

[0016] S6, Dynamic weight adjustment: Based on the real-time operating status of the power grid, the weights of each item in the cost function are dynamically adjusted. When a power grid fault is detected, current tracking and voltage support capabilities are prioritized. When the system is stable, switching smoothness is prioritized to reduce losses.

[0017] S7, Constrained Online Rolling Optimization: Under the set constraints of output current, output voltage and control increment, the optimal control sequence that minimizes the multi-objective cost function in the future control time domain is solved online.

[0018] S8, Application of Control Quantity and Rolling Execution: The first control quantity in the optimal control sequence is converted into a switching signal to drive the energy storage converter, and steps S1 to S7 are repeated in the next cycle to achieve closed-loop rolling optimization control.

[0019] Preferably, in step S5, the predictive penalty for out-of-bounds energy storage state of charge is implemented in the following manner:

[0020] Predict the trajectory of future state of charge changes caused by the current control sequence;

[0021] When it is predicted that the state of charge will exceed the preset safe operating range at a future time, a penalty term that increases sharply is generated in the cost function.

[0022] The penalty term forces the optimizer to automatically adjust the power output when calculating control commands, avoiding overcharging or over-discharging of the energy storage unit and achieving inherent safety protection at the control layer.

[0023] Preferably, in step S6, the assessment of voltage support capability and mode switching are achieved in the following manner:

[0024] A penalty term is introduced into the cost function to penalize deviations of the converter port voltage from its reference value.

[0025] By dynamically adjusting the weight of the penalty term, a smooth switching between grid-connected mode and grid-connected mode of the energy storage converter can be achieved; when the grid strength is high, the weight is reduced to prioritize precise current control; when a grid failure is detected or the device is in islanded mode, the weight is increased to force the controller to prioritize maintaining the stability of the port voltage.

[0026] Preferably, in step S4, the perturbation prediction sequence is obtained in the following manner:

[0027] Real-time identification and observation of the fundamental and harmonic components of the grid voltage;

[0028] Within the short prediction time domain, assuming that the disturbance remains unchanged at the current observation value, linear extrapolation is performed based on its changing trend.

[0029] Preferably, in step S7, the constrained online rolling optimization adopts a hierarchical optimization approach to balance computational speed and control accuracy:

[0030] The first layer of optimization involves quickly selecting several candidate switching states with the best performance from the limited set of discrete switching states of the energy storage converter.

[0031] The second layer of optimization involves using a continuous optimization algorithm to determine the precise duration of each state in the next control cycle, based on the selected candidate switch states, thereby achieving a balance between precise control and a fixed switching frequency.

[0032] Preferably, in step S3, the reference trajectory and disturbance feedforward generate the accepted upper-level control command, which is generated by the dynamic fusion of grid-based control and grid-following control; the fusion ratio is dynamically adjusted according to the real-time assessed grid connection strength.

[0033] To achieve the above objectives, the present invention also provides the following technical solution:

[0034] A fast dynamic response system for an energy storage converter based on model predictive control for performing the above method, the system comprising:

[0035] The data acquisition and processing module is used to execute the real-time data acquisition and processing steps;

[0036] A reference generation and feedforward module is used to perform the reference trajectory and disturbance feedforward generation steps;

[0037] The prediction model module is used to store the discrete state-space model of the energy storage converter and execute the rolling prediction step of the prediction model.

[0038] A multi-objective optimization calculation module is used to construct the multi-objective cost function and execute the dynamic weight adjustment step;

[0039] A constrained rolling optimizer is used to perform the constrained online rolling optimization steps;

[0040] The modulation and driving module is used to execute the control quantity application and rolling execution steps.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] (1) This invention is a fast dynamic response method for energy storage converters based on model predictive control. By rolling prediction based on discrete state space model, the control action that can optimally guide the future state of the system is actively calculated, so that the response speed of energy storage converter to millisecond-level disturbances such as grid frequency fluctuations and voltage sags reaches the theoretical limit. At the same time, through hierarchical optimization, precise fine-tuning of switching state is achieved while ensuring real-time calculation, effectively reducing output current ripple and improving power quality.

[0043] (2) This invention is a fast dynamic response method for energy storage converter based on model predictive control. It unifies current tracking, switching loss, voltage support and energy storage SOC into a multi-objective cost function. Through the dynamic weight adjustment in steps S5 and S6, it prioritizes grid stability and battery health during grid faults, fundamentally solving the conflict problem of multiple control objectives in the dynamic process and achieving the optimization of the system's global performance.

[0044] (3) The present invention is a fast dynamic response method for energy storage converter based on model predictive control. By predictive penalty for the out-of-bounds state of charge of energy storage in step S5, the SOC protection is deeply embedded into the bottom fast control loop. When calculating each switch signal, the impact of the current action on the future SOC trajectory is evaluated, providing the energy storage unit with intrinsic safety protection of millisecond-level response at the control bottom layer, which greatly extends the battery life.

[0045] (4) The present invention is a fast dynamic response method for energy storage converter based on model predictive control. By dynamically adjusting the voltage support weight in step S6, the energy storage converter can be made to seamlessly, smoothly and adaptively switch between grid-connected and grid-connected modes. Based on the real-time assessment of the grid strength, it can be made to operate stably and efficiently in various complex environments from the main grid to the end microgrid.

[0046] (5) The present invention is a fast dynamic response method for energy storage converter based on model predictive control. By setting the hierarchical optimization method in step S5, the first layer performs rapid initial selection in a limited set, which ensures the speed of control; the second layer performs continuous optimization in the selected range, which ensures the accuracy of control. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0048] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0049] Example

[0050] Please see Figure 1 This invention proposes a technical solution for a fast dynamic response method for energy storage converters based on model predictive control: the fast dynamic response method for energy storage converters based on model predictive control includes the following steps:

[0051] S1, System Initialization: Preload the system parameters and safe operating boundaries of the energy storage converter;

[0052] S2, Real-time data acquisition and processing: In each control cycle, the AC side voltage and current and DC side voltage of the energy storage converter are acquired synchronously and their coordinates are transformed; at the same time, the real-time state of charge of the energy storage unit is obtained from the battery management system at a slower cycle.

[0053] S3, Reference Trajectory and Disturbance Feedforward Generation: Receive upper-level control commands, generate current and voltage reference values ​​for the current control cycle, and use the collected grid voltage as a measurable feedforward disturbance signal;

[0054] S4, Rolling Prediction Model: Based on the established discrete state-space prediction model of the energy storage converter, the system behavior state sequence at multiple future times is predicted in a rolling manner according to the current system state, candidate control sequence and disturbance prediction sequence.

[0055] S5, Multi-objective cost function evaluation: Construct a cost function to evaluate the performance of the candidate control sequence. The cost function includes at least a penalty for future tracking errors, a penalty for drastic changes in control quantity, a predictive penalty for out-of-bounds energy storage state of charge, and an evaluation of voltage support capability.

[0056] S6, Dynamic weight adjustment: Based on the real-time operating status of the power grid, the weights of each item in the cost function are dynamically adjusted. When a power grid fault is detected, current tracking and voltage support capabilities are prioritized. When the system is stable, switching smoothness is prioritized to reduce losses.

[0057] S7, Constrained Online Rolling Optimization: Under the set constraints of output current, output voltage and control increment, the optimal control sequence that minimizes the multi-objective cost function in the future control time domain is solved online.

[0058] S8, Application of Control Quantity and Rolling Execution: The first control quantity in the optimal control sequence is converted into a switching signal to drive the energy storage converter, and steps S1 to S7 are repeated in the next cycle to achieve closed-loop rolling optimization control.

[0059] Furthermore, in step S5, the predictive penalty for out-of-bounds energy storage state of charge is implemented in the following way:

[0060] Predict the trajectory of future state of charge changes caused by the current control sequence;

[0061] When it is predicted that the state of charge will exceed the preset safe operating range at a future time, a penalty term that increases sharply is generated in the cost function.

[0062] The penalty term forces the optimizer to automatically adjust the power output when calculating control commands, avoiding overcharging or over-discharging of the energy storage unit and achieving inherent safety protection at the control layer.

[0063] Furthermore, in step S6, the assessment of voltage support capability and mode switching are achieved in the following manner:

[0064] A penalty term is introduced into the cost function to penalize deviations of the converter port voltage from its reference value.

[0065] By dynamically adjusting the weight of the penalty term, a smooth switching between grid-connected mode and grid-connected mode of the energy storage converter can be achieved; when the grid strength is high, the weight is reduced to prioritize precise current control; when a grid failure is detected or the device is in islanded mode, the weight is increased to force the controller to prioritize maintaining the stability of the port voltage.

[0066] Furthermore, in step S4, the perturbation prediction sequence is obtained in the following manner:

[0067] Real-time identification and observation of the fundamental and harmonic components of the grid voltage;

[0068] Within the short prediction time domain, assuming that the disturbance remains unchanged at the current observation value, linear extrapolation is performed based on its changing trend.

[0069] Furthermore, in step S7, the constrained online rolling optimization adopts a hierarchical optimization approach to balance computational speed and control accuracy:

[0070] The first layer of optimization involves quickly selecting several candidate switching states with the best performance from the limited set of discrete switching states of the energy storage converter.

[0071] The second layer of optimization involves using a continuous optimization algorithm to determine the precise duration of each state in the next control cycle, based on the selected candidate switch states, thereby achieving a balance between precise control and a fixed switching frequency.

[0072] Furthermore, in step S3, the reference trajectory and disturbance feedforward generate the accepted upper-level control command, which is dynamically generated by the fusion of grid-based control and grid-following control; the fusion ratio is dynamically adjusted according to the real-time assessed grid connection strength.

[0073] In this embodiment, the method actively anticipates the system's behavior at multiple future moments through a predictive model (step S4), enabling the controller to take optimal actions in advance, fundamentally avoiding overshoot and oscillation, and achieving near-theoretical speed. The closed-loop optimization is not a one-time long-term optimization, but rather a re-execution of the prediction-optimization process (steps S4-S7) in each control cycle, and the first step of the optimization result is implemented (step S8), enabling the controller to continuously absorb the latest state information of the system and exhibiting strong robustness to model mismatch and unknown disturbances.

[0074] Steps S1 and S3 provide accurate data for model predictive control. The introduction of model predictive control correlates the slow variable of the energy state of the energy storage converter with fast power control. The grid voltage, as a feedforward disturbance, directly offsets the main impact of external grid changes on the system, simplifying the optimizer's workload.

[0075] The following is the workflow in a typical dynamic scenario:

[0076] Scenario: When a microgrid is in operation, a sudden increase in power load occurs, and then it disconnects from the main grid and enters islanding mode.

[0077] 1. The moment the disturbance occurs (sudden load increase):

[0078] System response: The upper-level control detects a frequency drop and rapidly increases the active power reference value. In step S3, the current reference value increases accordingly.

[0079] MPC Intelligent Decision Making:

[0080] The optimizer (S7) receives new data and quickly calculates a set of switching sequences that can raise the current to the new reference value with the fastest speed and the least overshoot based on the prediction model (S4).

[0081] The current tracking term in the cost function (S5) plays a dominant role at this point, ensuring a fast response.

[0082] The constraint (S7) also ensures that the output current does not exceed the converter's safety limit.

[0083] 2. Continuous operation and status alerts (SOC approaching its limit):

[0084] System status: Due to continuous discharge, the SOC of the energy storage is constantly rising, and the predictive model predicts that the SOC will exceed the safe limit in the next few steps.

[0085] MPC Intelligent Approach:

[0086] The SOC penalty term (S5) is activated and increases sharply. When the optimizer seeks to minimize the cost function, it will find that if it continues to discharge at high power as originally planned, it will fail.

[0087] Result: The optimizer automatically and smoothly calculates a new control sequence with slightly lower power, without the need for intervention from the upper-level EMS, thus achieving autonomous protection at the lower level.

[0088] 3. Mode switching moment (main network disconnected, entering island mode):

[0089] System detection: The control system detected an abnormal grid voltage and determined that the system was in an islanded state.

[0090] MPC Intelligent Decision Making:

[0091] The dynamic weight adjustment (S6) takes immediate action, significantly increasing the weight of the voltage support term in the cost function, while potentially reducing the weight of the current tracking term.

[0092] Optimizer behavior shift: At this point, maintaining stable port voltage becomes a more important task for the optimizer than accurately tracking the current reference. Its calculated control sequence prioritizes the stability of the output voltage amplitude and frequency, automatically and seamlessly switching from grid-following mode to grid-building mode to establish a stable voltage platform for critical loads within the island.

[0093] A comparison of the prior art with the technology of this invention is shown in Table 1 below:

[0094]

[0095] Table 1

[0096] This method achieves the theoretical limit in response speed, effectively suppresses transient oscillations in the power grid, and provides high-quality frequency and voltage support services. It can automatically make optimal choices under complex operating conditions and provide the deepest level of safety protection for energy storage converters. Through adaptive mode switching and constraint processing, it ensures that the converter operates stably without disconnecting from the grid under various power grid conditions. The hierarchical optimization method ensures the feasibility of advanced algorithms on existing hardware platforms, paving the way for large-scale commercial applications.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fast dynamic response method for energy storage converters based on model predictive control, characterized in that, Includes the following steps: S1, System Initialization: Preload the system parameters and safe operating boundaries of the energy storage converter; S2, Real-time data acquisition and processing: In each control cycle, the AC side voltage and current and DC side voltage of the energy storage converter are acquired synchronously and their coordinates are transformed; at the same time, the real-time state of charge of the energy storage unit is obtained from the battery management system at a slower cycle. S3, Reference Trajectory and Disturbance Feedforward Generation: Receive upper-level control commands, generate current and voltage reference values ​​for the current control cycle, and use the collected grid voltage as a measurable feedforward disturbance signal; S4, Rolling Prediction Model: Based on the established discrete state-space prediction model of the energy storage converter, the system behavior state sequence at multiple future times is predicted in a rolling manner according to the current system state, candidate control sequence and disturbance prediction sequence. S5, Multi-objective cost function evaluation: Construct a cost function to evaluate the performance of the candidate control sequence. The cost function includes at least a penalty for future tracking errors, a penalty for drastic changes in control quantity, a predictive penalty for out-of-bounds energy storage state of charge, and an evaluation of voltage support capability. S6, Dynamic weight adjustment: Based on the real-time operating status of the power grid, the weights of each item in the cost function are dynamically adjusted. When a power grid fault is detected, current tracking and voltage support capabilities are prioritized. When the system is stable, switching smoothness is prioritized to reduce losses. S7, Constrained Online Rolling Optimization: Under the set constraints of output current, output voltage and control increment, the optimal control sequence that minimizes the multi-objective cost function in the future control time domain is solved online. S8, Application of Control Quantity and Rolling Execution: The first control quantity in the optimal control sequence is converted into a switching signal to drive the energy storage converter, and steps S1 to S7 are repeated in the next cycle to achieve closed-loop rolling optimization control.

2. The fast dynamic response method for energy storage converters based on model predictive control according to claim 1, characterized in that, In step S5, the predictive penalty for out-of-bounds state of charge of the energy storage is implemented in the following way: Predict the trajectory of future state of charge changes caused by the current control sequence; When it is predicted that the state of charge will exceed the preset safe operating range at a future time, a penalty term that increases sharply is generated in the cost function. The penalty term forces the optimizer to automatically adjust the power output when calculating control commands, avoiding overcharging or over-discharging of the energy storage unit and achieving inherent safety protection at the control layer.

3. The fast dynamic response method for energy storage converters based on model predictive control according to claim 1, characterized in that, In step S6, the assessment of voltage support capability and mode switching are achieved in the following ways: A penalty term is introduced into the cost function to penalize deviations of the converter port voltage from its reference value. By dynamically adjusting the weight of the penalty term, a smooth switching between grid-connected mode and grid-connected mode of the energy storage converter can be achieved; when the grid strength is high, the weight is reduced to prioritize precise current control; when a grid failure is detected or the device is in islanded mode, the weight is increased to force the controller to prioritize maintaining the stability of the port voltage.

4. The fast dynamic response method for energy storage converters based on model predictive control according to claim 1, characterized in that, In step S4, the perturbation prediction sequence is obtained in the following manner: Real-time identification and observation of the fundamental and harmonic components of the grid voltage; Within the short prediction time domain, assuming that the disturbance remains unchanged at the current observation value, linear extrapolation is performed based on its changing trend.

5. The fast dynamic response method for energy storage converters based on model predictive control according to claim 1, characterized in that, In step S7, the constrained online rolling optimization adopts a hierarchical optimization approach to balance computational speed and control accuracy: The first layer of optimization involves quickly selecting several candidate switching states with the best performance from the limited set of discrete switching states of the energy storage converter. The second layer of optimization involves using a continuous optimization algorithm to determine the precise duration of each state in the next control cycle, based on the selected candidate switch states, thereby achieving a balance between precise control and a fixed switching frequency.

6. The fast dynamic response method for energy storage converters based on model predictive control according to claim 1, characterized in that, In step S3, the reference trajectory and disturbance feedforward generate the accepted upper-level control command, which is dynamically generated by the fusion of grid-based control and grid-following control; the fusion ratio is dynamically adjusted according to the real-time assessment of the grid connection strength.

Images