Fire storage combined frequency modulation optimization method and system for model predictive control

Abstract

The invention relates to the field of power system automation control, discloses a model predictive control-based fire-storage combined frequency modulation optimization method and system, and aims to solve the problems of inaccurate fire-storage dynamic coupling modeling, poor disturbance adaptability, insufficient multi-target collaboration and weak real-time performance in the prior art. The method comprises the following steps: constructing a thermal power generating unit nonlinear dynamic response prediction model and an energy storage system power output capability model; based on the power grid frequency deviation, generating a fire storage combined frequency modulation power distribution instruction by adopting a rolling time domain optimization strategy; and executing the instruction and correcting the deviation through a double-closed-loop feedback structure. According to the scheme, the frequency modulation precision and the response speed are remarkably improved, the unit adjustment burden and the energy storage loss are reduced, the robustness is enhanced, and collaborative optimization of economical efficiency and dynamic performance is achieved.

Description

Technical Field

[0001] This invention belongs to the field of power system automation control, specifically relating to a model predictive control method and system for combined thermal and energy storage frequency regulation optimization. Background Technology

[0002] Against the backdrop of the rapid development of new power systems, the joint participation of thermal power and energy storage in grid frequency regulation has become an important means to improve system flexibility and stability. Thermal-storage joint frequency regulation aims to achieve efficient and accurate tracking of grid frequency fluctuations by coordinating the slow dynamic response of thermal power units with the fast dynamic response capability of energy storage systems, thereby meeting increasingly stringent frequency regulation performance requirements.

[0003] Among them, model predictive control, as an advanced closed-loop optimization control strategy, has been widely explored for use in the joint frequency regulation scenario of thermal power and energy storage due to its rolling optimization, feedforward compensation and multivariate coordination capabilities. Its core lies in rolling prediction and optimization of future control behavior based on the system dynamic model, so as to take into account frequency regulation accuracy, equipment life and operating economy.

[0004] Existing technologies still face multiple challenges when applying model predictive control in combined thermal power and energy storage frequency regulation: First, the dynamic characteristics of thermal power units and energy storage systems differ significantly, and existing methods often employ simplified or decoupled models, making it difficult to accurately characterize the strong coupling relationship between the two under power allocation, response delay, and capacity constraints. Second, the predictive models are not adaptable enough to load disturbances and grid command changes, lacking an effective robust handling mechanism for uncertain disturbances, leading to easy degradation of control performance in actual operation. Third, the optimization objective function usually only focuses on frequency regulation performance indicators, neglecting the synergistic optimization of energy storage system cycle life decay and thermal power unit coal consumption costs, resulting in poor long-term operational economy. Finally, existing algorithms have high computational complexity, making it difficult to meet the real-time requirement of second-level response to frequency regulation commands under limited communication and computing resources. These problems severely restrict the large-scale application and effectiveness of combined thermal power and energy storage frequency regulation systems in grids with a high proportion of renewable energy. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a model predictive control method and system for joint frequency regulation of thermal power and energy storage, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: On one hand, a model predictive control-based thermal power-storage joint frequency regulation optimization system, comprising the following components: a thermal power unit dynamic response modeling module, used to construct a thermal power unit frequency regulation response prediction model with nonlinear dynamic characteristics based on historical operating data and real-time operating parameters of the thermal power unit; an energy storage system charge / discharge characteristic modeling module, used to establish a power output capability model of the energy storage system under different frequency regulation commands based on the state of charge, charge / discharge efficiency curves, and power constraint boundaries of the energy storage battery; a joint frequency regulation collaborative optimization controller, used to receive grid frequency deviation signals and, based on the thermal power unit frequency regulation response prediction model and the energy storage system power output capability model, generate a thermal power-storage joint frequency regulation power allocation command using a rolling time-domain optimization strategy; and a model predictive control execution unit, used to send the thermal power-storage joint frequency regulation power allocation command to the thermal power unit control system and the energy storage converter control system respectively, realizing closed-loop dynamic control of the thermal power-storage joint frequency regulation process. On the other hand, a model predictive control-based thermal power-storage joint frequency regulation optimization algorithm is proposed. The specific steps of this method are as follows: Step S110, collect historical operating data and real-time operating parameters of the thermal power unit to construct a thermal power unit frequency regulation response prediction model; Step S120, obtain the state of charge, charge / discharge efficiency curves, and power constraint boundaries of the energy storage system to establish an energy storage system power output capacity model; Step S130, receive the grid frequency deviation signal and, based on the thermal power unit frequency regulation response prediction model and the energy storage system power output capacity model, generate a thermal power-storage joint frequency regulation power allocation command using a rolling time-domain optimization strategy; Step S140, send the thermal power-storage joint frequency regulation power allocation command to the thermal power unit control system and the energy storage converter control system respectively to complete the thermal power-storage joint frequency regulation closed-loop control. Preferably, the frequency regulation response prediction model of the thermal power unit adopts a nonlinear autoregressive exogenous input model (NARX) based on state-space equations. Its input variables include unit load command, main steam pressure, turbine speed deviation and boiler combustion rate, and the output variable is the actual change in active power of the unit. The model continuously updates the model parameters through an online identification mechanism to adapt to the differences in dynamic response characteristics of the thermal power unit in different load ranges. Furthermore, the power output capability model of the energy storage system is constructed based on the equivalent circuit model of the energy storage battery, considering the nonlinear characteristics of the battery internal resistance changing with the state of charge (SOC), and introducing the influence factor of the depth of charge and discharge on the cycle life; the model dynamically calculates the upper and lower limits of the maximum adjustable power of the energy storage system under the premise of meeting the SOC safety boundary (20%≤SOC≤90%) in each control cycle. Furthermore, the rolling time-domain optimization strategy solves the following objective function minimization problem in each control cycle:

[0007] in, Indicates the first Predicted frequency deviation and These respectively represent the thermal power unit and the energy storage system in the [missing information - likely a date or time]. The amount of power adjustment in the step. To predict the length of the time domain, , , Let be the weight coefficient, and satisfy... > > ×1.5, to prioritize frequency deviation suppression, then reduce the adjustment range of thermal power units, and finally optimize the smoothness of energy storage system operation. Preferably, the joint frequency regulation coordinated optimization controller incorporates a thermal power unit ramp rate constraint when solving the objective function. Constraints on the power change rate of energy storage systems ,in This represents the maximum ramp rate of the thermal power unit (unit: MW / min). For control period (unit: s). The power change rate coefficient of the energy storage system. This refers to the rated power of the energy storage system. Furthermore, the model predictive control execution unit adopts a dual closed-loop feedback structure. The outer loop is a power allocation command generation loop based on frequency deviation, and the inner loop is a thermal power-storage actual power tracking loop. The inner loop collects the actual output of the thermal power unit and the actual charging and discharging power of the energy storage system in real time, corrects the deviation with the allocation command, and feeds the correction amount back to the optimization calculation of the next cycle to compensate for the control error caused by model mismatch and external disturbances. Furthermore, the generation process of the joint frequency regulation power allocation command of the thermal power and energy storage adopts a hierarchical optimization architecture: the upper-level optimization determines the long-term allocation ratio of thermal power and energy storage frequency regulation capacity based on economic indicators, and the lower-level optimization performs real-time power fine-tuning based on dynamic response characteristics in each control cycle; the upper-level optimization aims to minimize the cost per unit frequency regulation mileage, and the lower-level optimization aims to minimize the integral absolute value of frequency deviation (IAE). The two are coupled through a capacity reservation mechanism. Preferably, the energy storage system charge / discharge characteristic modeling module introduces a temperature correction factor when establishing the power output capability model. This factor is based on the battery operating temperature. The dynamic adjustment of charge / discharge efficiency and maximum power output capability is functionally related as follows: ,when When the temperature exceeds the range of 15℃ to 35℃, Linear decay is used to reflect the suppressive effect of low or high temperatures on battery performance. Compared with the prior art, the present invention has the following beneficial effects: By constructing a high-precision dynamic response model for thermal power units and energy storage systems, and performing joint optimization under the model predictive control framework, the frequency response speed and accuracy of the thermal power and energy storage joint frequency regulation system were significantly improved, with the peak frequency deviation reduced by more than 35% and the regulation time shortened by 40%. By adopting a rolling time-domain optimization strategy and introducing a multi-objective weight coordination mechanism, the frequency regulation performance is guaranteed while the number and magnitude of frequent adjustments of thermal power units are effectively reduced, the lifespan of the unit equipment is extended, and the charging and discharging depth and cycle number of the energy storage system are optimized, thereby improving its economic efficiency throughout its entire life cycle. By using a dual closed-loop feedback structure and an online model update mechanism, the robustness of the system to model mismatch, parameter drift and external disturbances is enhanced, ensuring that stable frequency modulation performance can still be maintained under complex working conditions. The hierarchical optimization architecture achieves synergy between long-term economic efficiency and short-term dynamic performance, avoiding the resource waste or performance deficiencies caused by single-objective optimization in traditional methods, and reducing the overall frequency regulation cost by more than 20%. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention. Detailed Implementation

[0009] Please refer to Figure 1 To further illustrate the technical means and effects of the present invention in order to achieve the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0010] Example 1 This embodiment describes a scenario in a regional power grid where a large thermal power unit and a megawatt-level energy storage system jointly provide primary frequency regulation services. This regional power grid incorporates a high proportion of renewable energy, leading to increased frequency fluctuations and placing higher demands on the response speed and accuracy of frequency regulation resources. This invention aims to achieve coordinated and optimized frequency regulation between thermal power units and energy storage systems through a model predictive control framework to cope with complex grid frequency disturbances.

[0011] The dynamic response modeling module for thermal power units is responsible for accurately depicting the dynamic behavior of thermal power units during frequency regulation. This module collects various operational data of the thermal power unit in real time, including but not limited to historical load commands, actual active power output, main steam pressure, turbine speed deviation, boiler combustion rate, feedwater flow rate, turbine valve opening, and operating parameters of each stage of heaters and condensers. This data is acquired at millisecond-level time resolution through a high-speed data acquisition system and then cleaned, denoised, outlier detected, and imputed by a dedicated data preprocessing unit. The data preprocessing unit employs a method combining sliding window mean filtering and wavelet transform to effectively filter out random sensor noise and short-term measurement spikes. After data preparation, the module constructs a frequency regulation response prediction model for the thermal power unit based on this data. This model uses a nonlinear autoregressive exogenous input model (NARX) based on state-space equations. The input variables of the NARX model include: changes in unit load command (i.e., power setpoints from the upper-level control system, typically changing at a rate of MW / min), main steam pressure deviation (instantaneous difference relative to rated pressure, in MPa), turbine speed deviation (instantaneous difference relative to synchronous speed, in rpm), and boiler combustion rate changes (i.e., adjustments to fuel supply rate or air / fuel ratio, in %). The model's output variables are precisely defined as the changes in the unit's actual active power within each control cycle (in MW). The model construction process begins with offline training using historical operating data, employing iterative least squares or Kalman filtering algorithms to estimate initial parameters and capture the nonlinear dynamic characteristics of the unit across different load ranges (e.g., 30% to 100% rated load). Upon entering the online operation phase, the module activates an online identification mechanism, continuously fine-tuning and updating the model parameters using the latest real-time operating data. The online identification mechanism employs a forgetting factor recursive least squares algorithm. By assigning greater weight to new data, it ensures the model can dynamically adapt to differences in dynamic response characteristics caused by factors such as different load levels, different operating modes (e.g., sliding pressure operation or constant pressure operation), and equipment aging. Specifically, the mechanism updates the key parameters of the model every 5 seconds, ensuring that the error between the model's predicted values ​​and the actual unit output is always controlled within an acceptable range; for example, the prediction error is less than 1% of the actual output change. This online update mechanism enables the model to capture subtle dynamic characteristics such as changes in boiler thermal storage capacity, turbine efficiency decline, or minute variations in valve characteristics, thereby greatly improving the prediction accuracy of the actual unit response.

[0012] The energy storage system charge / discharge characteristic modeling module aims to establish a power output capability model for the energy storage system under high-frequency frequency regulation commands. This module first obtains the state of charge (SOC) of each energy storage unit in real time from the battery management system (BMS) and aggregates the SOCs of all energy storage units to calculate the average SOC of the entire energy storage system. Simultaneously, the module also obtains detailed charge / discharge efficiency curves provided by the energy storage battery manufacturer. These curves typically use SOC and charge / discharge current as inputs and output the instantaneous charge / discharge efficiency of the battery. Furthermore, power constraints must be considered, including the rated power of the energy storage converter (PCS), the maximum instantaneous charge / discharge current limit of the battery, and DC bus voltage fluctuations. Based on this information, the module constructs a power output capability model for the energy storage system. This model is based on an equivalent circuit model of the energy storage battery, for example, using a first-order or second-order RC equivalent circuit model to accurately simulate the battery's internal resistance characteristics. In this equivalent circuit model, the battery's internal resistance is not a fixed value but exhibits significant nonlinear characteristics with changes in the state of charge (SOC). This module loads parameters such as battery open-circuit voltage, ohmic internal resistance, polarization internal resistance, and polarization capacitance at different SOCs into the model in real time using lookup tables or polynomial fitting. Furthermore, the model incorporates a factor influencing cycle life based on the depth of charge / discharge (DoD). This factor estimates the battery's state of health (SOH) and remaining life in real time by recording the depth of each charge / discharge cycle of the energy storage system and combining it with the DoD-cycle life curve provided by the battery manufacturer. Based on this, the maximum adjustable power is fine-tuned to avoid damage to battery life caused by overcharging and discharging. Within each control cycle (e.g., 1 second), the module dynamically calculates the upper and lower limits of the maximum adjustable power of the energy storage system while meeting the SOC safety boundary. The SOC safety boundary is strictly set between 20% and 90% to prevent irreversible damage to the battery from over-discharge and safety hazards from overcharging, while ensuring the energy storage system has sufficient frequency regulation reserve capacity. To further improve the model's accuracy and safety, the module introduces a temperature correction factor when establishing the power output capability model. This factor is based on the real-time operating temperature of the energy storage system's battery pack. Dynamically adjust charge / discharge efficiency and maximum power output. Battery operating temperature is monitored in real-time by a distributed temperature sensor array, and the average and maximum temperatures of the battery pack are calculated. When the temperature exceeds the battery's optimal operating range (typically 15°C to 35°C), the battery's internal resistance increases, the diffusion rate of active materials decreases, leading to a drop in charge / discharge efficiency and limiting maximum power output. Temperature correction factor. The functional relationship is as follows:

[0013] in, This refers to the average operating temperature of the battery pack (unit: °C). When... When the ideal operating temperature is within the range of 15℃ to 35℃, A value close to 1 has little impact on charge and discharge performance. When... When outside this range, for example, below 15°C or above 35°C, The power will decrease linearly. This means that under low-temperature conditions, the available charge and discharge power of the energy storage system will be reduced proportionally to prevent excessive current from damaging the battery in the low-temperature, high-internal-resistance state, while ensuring battery life. Under high-temperature conditions, power output will also be limited to avoid the risk of thermal runaway and maintain stable system operation. This correction factor ensures that the energy storage system can still operate safely and reliably under extreme temperature conditions and adjusts its contribution to frequency regulation according to actual performance.

[0014] The joint frequency regulation coordinated optimization controller is the core decision-making unit of this invention. This controller uses the grid frequency deviation signal issued by the grid dispatch center as its main input. This signal is typically received with a sampling period of 100 milliseconds or 1 second and undergoes digital filtering to remove high-frequency noise. Based on the aforementioned thermal power unit frequency regulation response prediction model and energy storage system power output capability model, the controller employs a rolling time-domain optimization strategy to generate joint thermal power-storage frequency regulation power allocation commands. Within each control cycle (e.g., 1 second), the rolling time-domain optimization strategy first utilizes the current system state and future... Based on the predicted information (e.g., the predicted grid frequency deviation for the next 50 seconds), an optimization problem is solved to determine the control input at the current moment. Once the control input at the current moment is executed, the system state advances forward by one control cycle, and the optimization time domain also rolls forward accordingly, repeating the above optimization process. This rolling optimization mechanism allows the controller to adapt to system dynamic changes and external disturbances in real time. Within each control cycle, the optimization controller solves the problem of minimizing the following objective function:

[0015] in, Indicates the first time domain within the rolling time domain The formula for calculating the predicted power grid frequency deviation is as follows: , The rated power grid frequency, This represents the predicted actual power grid frequency. and These respectively represent the thermal power unit and the energy storage system in the [missing information - likely a date or time]. The predicted power adjustment amount for each step. To predict the time domain length, it is usually set to 30 to 60 control cycles, depending on the grid inertia and frequency regulation response speed requirements. , , These are weighting coefficients used to balance different optimization objectives. In this embodiment, the weighting coefficients are precisely set to... =100, =30, =10. This weighting configuration ensures > > The relationship of ×1.5 is used to prioritize the rapid and effective suppression of frequency deviations, which is the primary goal of FM service; secondly, by reducing... The weights are used to limit the frequency and magnitude of adjustments of thermal power units, thereby extending the lifespan of the unit equipment and reducing fuel consumption; finally, minimizing... The aim is to optimize the smoothness of energy storage system operation, avoid frequent deep charge and discharge cycles, thereby extending the cycle life of energy storage batteries and reducing their total life-cycle operating costs. In solving the above objective function, this controller introduces several key constraints. The ramp rate constraint of the thermal power unit is strictly defined as... ,in This refers to the maximum ramp rate of the thermal power unit (unit: MW / min), for example, 10MW / min. The control period (in seconds), for example, 1 second. The power change rate constraint of the energy storage system is: ,in This is the power change rate coefficient for the energy storage system (e.g., 0.5, indicating that the maximum power change per second is 50% of the rated power). The rated power of the energy storage system (unit: MW) is also included. Furthermore, it includes upper and lower limits of the actual power output of the thermal power units and the energy storage system, as well as safety boundary constraints of the energy storage system's State of Charge (SOC). The above optimization problem is iteratively calculated using a real-time optimization solver, which typically employs efficient algorithms such as quadratic programming (QP) or interior-point methods to ensure that the solution is completed and the optimization results are output within each control cycle.

[0016] The generation process of the combined thermal power and energy storage frequency regulation power allocation command adopts a hierarchical optimization architecture to achieve synergy between long-term economic efficiency and short-term dynamic performance. The upper-level optimization aims to minimize the cost per unit frequency regulation mileage, i.e., minimizing the coal cost of thermal power units and the lifespan loss cost of energy storage systems while meeting the grid's frequency regulation requirements. This upper-level optimization is performed over a relatively long scheduling cycle (e.g., hours or days) to determine the long-term allocation ratio of thermal power and energy storage frequency regulation capacity and reserve capacity. For example, based on the load forecast and renewable energy forecast for the next 24 hours, the upper-level optimization model calculates an optimal percentage contribution of thermal power unit frequency regulation capacity and energy storage system frequency regulation capacity, as well as the SOC reference trajectory of the energy storage system in the next scheduling cycle. The lower-level optimization performs real-time power fine-tuning based on dynamic response characteristics within each control cycle (e.g., 1 second), with the objective function being to minimize the absolute value of the integral of frequency deviation (IAE), i.e., minimizing... This allows for rapid suppression of grid frequency fluctuations. Upper-level optimization and lower-level optimization are coupled through a capacity reservation mechanism. Based on long-term economic calculations, the upper-level optimization reserves a certain proportion of frequency regulation capacity for thermal power units and energy storage systems. The lower-level optimization then performs real-time power allocation within these reserved capacity ranges, ensuring that short-term dynamic performance is met without compromising long-term economic objectives. For example, the upper-level optimization sets the upper limit of frequency regulation capacity for thermal power units at 50MW and for energy storage systems at 20MW. The lower-level optimization, within a 1-second control cycle, converts the real-time frequency deviation signal into power demand and, within a total capacity of 70MW, performs optimal power allocation based on the real-time dynamic response capabilities and current SOC status of the thermal power units and energy storage systems. This hierarchical architecture effectively avoids the resource waste or performance deficiencies caused by single-objective optimization in traditional methods.

[0017] The Model Predictive Control Execution Unit (MPCU) is the final control link of this invention. This unit receives the combined thermal power generation and energy storage (EPG) frequency regulation power allocation command generated by the joint frequency regulation and collaborative optimization controller, and distributes it to the thermal power unit control system and the energy storage converter control system respectively, realizing closed-loop dynamic control of the combined thermal power generation and energy storage frequency regulation process. The MPCU adopts a dual closed-loop feedback structure to enhance the system's robustness and control accuracy. The outer loop is a power allocation command generation loop based on frequency deviation. This loop is responsible for generating the total power demand for combined thermal power generation and energy storage frequency regulation that meets the optimization objectives and constraints based on the grid frequency deviation and the system model, and decomposing it into allocation commands for the thermal power unit and the energy storage system. The inner loop is a thermal power-to-energy tracking loop, whose function is to ensure that the thermal power unit and the energy storage system can accurately track the received power allocation command. The inner loop collects the actual active power output signal of the thermal power unit and the actual charging and discharging power signal of the energy storage system in real time (acquired through the power sensor of the PCS), and performs real-time deviation correction between these actual power values ​​and the power allocation command issued in the current cycle. For example, if there is a positive deviation between the actual output of the thermal power unit and the command, the inner loop controller will generate a negative correction. This correction will be fed back to the optimization calculation of the joint frequency regulation co-optimization controller at the beginning of the next control cycle through the internal feedback channel. This real-time deviation correction mechanism can effectively compensate for control errors caused by external disturbances such as response delays of thermal power units or energy storage systems, model mismatch, parameter drift, and sudden changes in grid load, ensuring that the system can maintain stable frequency regulation performance under highly dynamic environments. In specific implementation, the inner loop controller can use a high-bandwidth proportional-integral (PI) controller or a more advanced adaptive controller to dynamically adjust its control parameters according to the actual power tracking error, so as to minimize static and dynamic tracking errors, thereby completing the closed-loop control of the thermal power and energy storage joint frequency regulation.

[0018] Example 2 Building upon Example 1, this embodiment proposes differentiated optimization strategies and modeling improvements to address the changing demand for frequency regulation resources in the power grid under specific operating modes and the performance of energy storage systems under extreme environmental conditions. The application scenario for this embodiment is a weakly synchronous power grid region with large-scale wind farm integration. In this region, due to the uncertainty of wind power output and its low inertia, the power grid frequency fluctuates dramatically, requiring higher response standards. Simultaneously, the energy storage system may face extreme climatic conditions of high or low temperatures, leading to more significant performance degradation.

[0019] Building upon the dynamic response modeling module for thermal power units in Example 1, this example further optimizes the online identification mechanism of the NARX model, making it more adaptable. In areas with weak synchronous power grids, the operating conditions of thermal power units change more drastically due to frequent and large-amplitude frequency disturbances. Therefore, the online identification mechanism no longer relies solely on the forgetting factor recursive least squares algorithm, but introduces a model parameter correction strategy based on Online Support Vector Machine (Online SVM). This strategy can detect the prediction residuals of the NARX model in real time. Once the residuals exceed a preset threshold or show a persistent deviation trend, Online SVM will trigger rapid adjustments to local model parameters, especially parameters related to transient response characteristics, such as the turbine control valve response time constant or boiler heat storage coefficient, to adapt to the nonlinear characteristic changes of the unit under extreme operating conditions (e.g., rapid deep peak shaving or significant ramp-up). Meanwhile, to cope with the frequent parameter update requirements, the module adopts a lightweight model integration technology to maintain multiple NARX model copies in parallel. Each copy is trained and updated in different operating conditions. The controller intelligently switches or weights and fuses these model outputs according to real-time operating conditions, thereby improving the robustness and prediction accuracy of the overall model.

[0020] Building upon the energy storage system charge / discharge characteristic modeling module in Example 1, this example enhances the power output capability model of the energy storage system, particularly in terms of accuracy under extreme temperature conditions. In addition to the temperature correction factor mentioned in Example 1... This embodiment further introduces a battery aging (SOH) and temperature coupling correction factor. When the energy storage system operates for a long time, the battery's state of health (SOH) decreases, and its performance degradation at extreme temperatures becomes more significant. This coupling factor is used to perform a secondary correction on the temperature correction factor by real-time monitoring of the battery pack's SOH (e.g., estimating it using coulomb counting and internal resistance growth trends) and combining it with a two-dimensional lookup table of battery temperature and SOH. For example, a battery pack with an SOH of 80% will have a significantly lower maximum power output at 0°C than a battery pack with an SOH of 100%. Furthermore, in establishing the power constraint boundary, this embodiment also considers the impact of temperature differences (Delta T) and SOC differences (Delta SOC) between battery modules on the overall available power of the system. When Delta T or Delta SOC exceeds a preset threshold, in order to ensure the healthy operation of the battery pack and avoid local overload, the system will automatically reduce the maximum charge and discharge power of the entire energy storage system, rather than solely relying on average temperature and SOC. This refined modeling ensures that the energy storage system can still provide reliable and safe frequency regulation capabilities under extreme climatic conditions and battery aging.

[0021] In this embodiment, the joint frequency regulation coordinated optimization controller adjusts the objective function weight configuration of the rolling time-domain optimization strategy to adapt to the higher frequency stability requirements of the weakly synchronous power grid. Considering the low inertia characteristics brought about by wind power grid connection, the speed at which the power grid suppresses frequency deviations becomes the primary consideration. Therefore, the objective function... The weighting coefficients in the text were adjusted to =150, =20, =15. At this time... The value is higher than that of Example 1, emphasizing frequency deviation. To minimize the frequency fluctuations that are more likely to occur in weakly synchronous power grids. The weight was slightly reduced. The weighting is slightly increased, aiming to make smoother use of the energy storage system while ensuring frequency suppression, and to moderately relax the restrictions on the adjustment range of thermal power units, allowing them to make faster responses with greater amplitude when necessary to compensate for insufficient system inertia.

[0022] Furthermore, the generation process of the combined thermal power and energy storage frequency regulation power allocation command in this embodiment introduces a more dynamic capacity reservation mechanism within the hierarchical optimization architecture. The upper-level optimization aims to minimize the cost per unit frequency regulation mileage, but based on this, it dynamically adjusts the long-term allocation ratio of thermal power and energy storage frequency regulation capacity according to real-time changes in wind power forecasting errors. For example, when wind power forecasting uncertainty is high in the future, the upper-level optimization will proactively increase the frequency regulation capacity reservation of the energy storage system, because the energy storage system has a fast response speed and unrestricted ramp rate, making it more suitable for dealing with frequency fluctuations caused by high uncertainty. The lower-level optimization, during each control cycle, performs real-time power fine-tuning based on minimizing the integral absolute value of frequency deviation (IAE), and incorporates the grid frequency change rate (RoCoF) signal. When RoCoF exceeds a threshold, the lower-level optimization prioritizes the rapid response capability of the energy storage system and temporarily relaxes its smoothness constraints (i.e., dynamically adjusts). The system uses a dynamic capacity reservation and RoCoF-sensitive lower-level optimization strategy to suppress frequency drops or rises as quickly as possible, avoiding frequency instability. This dynamic capacity reservation and RoCoF-sensitive lower-level optimization strategy enable the system to respond more flexibly and quickly to frequency disturbances in scenarios with drastic fluctuations in wind power output and insufficient grid inertia.

[0023] In this embodiment, the model predictive control execution unit further enhances its robustness and adaptability. In addition to the deviation correction mechanism described in Embodiment 1, the inner-loop thermal-storage actual power tracking loop also incorporates an adaptive gain scheduling controller. This controller dynamically adjusts the proportional and integral gains of the PI controller based on the actual operating conditions of the thermal power unit and energy storage system (e.g., unit load rate, energy storage SOC, battery temperature, etc.). For example, when the thermal power unit is under low load or in a heating phase, its response may be slower; in this case, the inner-loop controller will appropriately reduce its gain to avoid oscillations. When the energy storage system SOC deviates from the ideal range, the inner-loop controller will adjust its gain to ensure that it tracks commands as accurately as possible without exceeding the SOC safety boundary. Furthermore, to enhance robustness against unknown external disturbances, the execution unit also integrates a compensation module based on sliding mode control (SMC). The SMC module can perform high-frequency compensation for residual control errors that model predictive control cannot completely eliminate. Especially in scenarios with sudden large disturbances (such as large-scale wind turbine disconnection) that are common in weakly synchronous power grids, the SMC module can provide fast and robust tracking performance, further improving the overall stability and reliability of the system.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A model predictive control-based combined thermal power and energy storage frequency regulation optimization system, characterized in that, Includes the following parts: The dynamic response modeling module for thermal power units is used to construct a frequency regulation response prediction model for thermal power units with nonlinear dynamic characteristics based on historical operating data and real-time operating parameters of thermal power units. The energy storage system charge and discharge characteristic modeling module is used to establish a power output capability model of the energy storage system under different frequency regulation commands based on the state of charge, charge and discharge efficiency curves and power constraint boundaries of the energy storage battery. The joint frequency regulation collaborative optimization controller is used to receive the grid frequency deviation signal and generate the joint frequency regulation power allocation command of thermal power unit based on the thermal power unit frequency regulation response prediction model and the power output capacity model, using a rolling time domain optimization strategy. The model predictive control execution unit is used to send the combined thermal power and energy storage frequency regulation power allocation command to the thermal power unit control system and the energy storage converter control system respectively, so as to realize closed-loop dynamic control of the combined thermal power and energy storage frequency regulation process. The frequency regulation response prediction model for thermal power units adopts a nonlinear autoregressive exogenous input model based on state-space equations. Its input variables include unit load command, main steam pressure, turbine speed deviation, and boiler combustion rate, while the output variable is the actual change in active power of the unit. The power output capability model is constructed based on the equivalent circuit model of the energy storage battery, considering the nonlinear characteristics of the battery's internal resistance changing with the state of charge, and dynamically calculating the upper and lower limits of the maximum adjustable power of the energy storage system under the premise of meeting the SOC safety boundary in each control cycle. The rolling time-domain optimization strategy solves the objective function minimization problem in each control cycle to generate a power allocation command that meets the requirements of frequency deviation suppression, thermal power regulation amplitude limitation, and energy storage action smoothness.

2. The model predictive control-based combined thermal power and energy storage frequency regulation optimization system according to claim 1, characterized in that, The dynamic response modeling module of the thermal power unit continuously updates the parameters of the frequency regulation response prediction model of the thermal power unit through an online identification mechanism. The online identification mechanism adopts the forgetting factor recursive least squares algorithm to fine-tune the key parameters of the model every 5 seconds to ensure that the model prediction error is less than 1% of the actual output change.

3. The model predictive control-based combined thermal and energy storage frequency regulation optimization system according to claim 1, characterized in that, The energy storage system charge / discharge characteristic modeling module introduces a temperature correction factor. This temperature correction factor is based on the battery operating temperature. The dynamic adjustment of charging / discharging efficiency and maximum power output capability has the following functional relationship: when When the temperature exceeds the range of 15℃ to 35℃, Linear decay is used to reflect the suppressive effect of low or high temperatures on battery performance.

4. The model predictive control-based combined thermal power and energy storage frequency regulation optimization system according to claim 1, characterized in that, The joint frequency regulation and collaborative optimization controller incorporates the thermal power unit ramp rate constraint when solving the objective function. Constraints on the power change rate of energy storage systems ,in This represents the maximum ramp rate of the thermal power unit. To control the cycle, The power change rate coefficient of the energy storage system. This refers to the rated power of the energy storage system.

5. The model predictive control-based combined thermal power and energy storage frequency regulation optimization system according to claim 1, characterized in that, The objective function solved by the rolling time-domain optimization strategy is: in, Indicates the first Predicted frequency deviation and These respectively represent the thermal power unit and the energy storage system in the [missing information - likely a date or time]. The amount of power adjustment in the step. To predict the length of the time domain, , , Let be the weight coefficient, and satisfy... > > ×1.

5.

6. The model predictive control-based combined thermal and energy storage frequency regulation optimization system according to claim 1, characterized in that, The model predictive control execution unit adopts a dual closed-loop feedback structure. The outer loop is a power allocation command generation loop based on frequency deviation, and the inner loop is a thermal power-storage actual power tracking loop. The inner loop collects the actual output of the thermal power unit and the actual charging and discharging power of the energy storage system in real time, corrects the deviation with the allocation command, and feeds the correction amount back to the optimization calculation of the next cycle.

7. The model predictive control-based combined thermal and energy storage frequency regulation optimization system according to claim 1, characterized in that, The generation process of the joint frequency regulation power allocation command of the thermal power and energy storage adopts a hierarchical optimization architecture: the upper-level optimization determines the long-term allocation ratio of thermal power and energy storage frequency regulation capacity based on economic indicators, and the lower-level optimization performs real-time power fine-tuning based on dynamic response characteristics in each control cycle; the upper-level optimization aims to minimize the cost per unit frequency regulation mileage, and the lower-level optimization aims to minimize the absolute value of the integral of frequency deviation. The two are coupled through a capacity reservation mechanism.

8. An optimization method for a combined thermal power and energy storage frequency regulation optimization system applied to model predictive control as described in any one of claims 1-7, characterized in that, Includes the following steps: Step S110: Collect historical operating data and real-time operating parameters of thermal power units, and construct a frequency regulation response prediction model for thermal power units; Step S120: Obtain the state of charge, charge / discharge efficiency curves, and power constraint boundaries of the energy storage system, and establish a power output capability model; Step S130: Receive the grid frequency deviation signal, and based on the thermal power unit frequency regulation response prediction model and the power output capability model, generate a joint thermal power and energy storage frequency regulation power allocation command using a rolling time-domain optimization strategy. Step S140: The combined thermal power generation and energy storage frequency regulation power allocation command is sent to the thermal power unit control system and the energy storage converter control system respectively to complete the closed-loop control of the combined thermal power generation and energy storage frequency regulation. The frequency regulation response prediction model for thermal power units adopts a nonlinear autoregressive exogenous input model based on state-space equations. Its input variables include unit load command, main steam pressure, turbine speed deviation, and boiler combustion rate, while the output variable is the actual change in active power of the unit. The power output capability model is constructed based on the equivalent circuit model of the energy storage battery, considering the nonlinear characteristics of the battery's internal resistance changing with the state of charge, and dynamically calculating the upper and lower limits of the maximum adjustable power of the energy storage system under the premise of meeting the SOC safety boundary in each control cycle. The rolling time-domain optimization strategy solves the objective function minimization problem in each control cycle to generate a power allocation command that meets the requirements of frequency deviation suppression, thermal power regulation amplitude limitation, and energy storage action smoothness.

9. The model predictive control-based joint frequency regulation optimization method for thermal power and energy storage according to claim 8, characterized in that, In step S140, the model predictive control execution unit adopts a dual closed-loop feedback structure. The inner loop collects the actual output of the thermal power unit and the actual charging and discharging power of the energy storage system in real time, corrects the deviation with the allocation command, and feeds the correction amount back to the optimization calculation of the next cycle. At the same time, the generation of the thermal power and energy storage joint frequency regulation power allocation command adopts a hierarchical optimization architecture. The upper-level optimization aims to determine the long-term capacity allocation ratio with the goal of minimizing the cost per unit frequency regulation mileage, while the lower-level optimization aims to perform real-time power fine-tuning with the goal of minimizing the absolute value of the integral of the frequency deviation.

Images