Dynamic load source network coupling nonlinear coordination control method for thermal power generating unit

Abstract

The invention discloses a thermal power generating unit dynamic load source network coupling nonlinear coordination control method. The method is specifically implemented according to the following steps: step 1, establishing a unit operation measurement and data acquisition unit; 2, carrying out source network unified modeling and multi-working-condition linearization; step 3, designing and realizing an extended state observer (ESO); step 4, carrying out ESO-MM-EMPC slow environment control design; step 5, FOSMC + ESO fast environment control design is carried out; and step 6, performing real-time control by adopting multi-time scale cooperative control and constraint management. The problems that an existing control method is low in response speed and low in control precision are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent control method of thermal power, and particularly relates to a dynamic load source network coupling nonlinear coordination control method of thermal power unit. BACKGROUND

[0002] The thermal power unit is the main regulating power source of the current automatic generation control (AGC) of the power grid, and a coordinated control system (CCS) thereof completes load tracking and main steam parameter regulation through coordination of the boiler, the steam turbine and other actuators. After large-scale wind power and photovoltaic power are connected to the grid, stronger uncertainty and random load disturbance are faced, and higher requirements are put forward for the fast, safe and economic peak regulation capability of the unit.

[0003] The traditional unit system thermal power unit mostly adopts the boiler following mode (BF), the turbine following boiler mode (TF) and the coordination mode (CC), and the coordination control of the boiler fuel quantity, the feedwater quantity and the turbine governing valve opening degree is realized based on the conventional PID and the feedforward compensation. On the basis of the traditional CCS, intelligent feedwater, intelligent overshoot and "reverse load change" processing strategies are introduced, which improve the load change performance to a certain extent, but still belong to the linear control framework with fixed structure and parameter setting depending on experience, and the adaptability to the strong nonlinear and strong coupling characteristics of the supercritical unit is limited.

[0004] In order to reduce the mutual restriction between the main steam pressure and the load, the existing technology proposes a multivariable adaptive dynamic decoupling algorithm to decouple the sliding pressure section and the constant pressure section, and improve the coordination quality in the sliding pressure condition, but mainly aims at the internal boiler-turbine loop of the unit, does not consider the coupling characteristics with the random load on the grid side, and regards the grid load change as a given instruction or an external disturbance, lacking unified modeling and control of the source network coupling mechanism. Under the conditions of strong random load, large-scale deep peak regulation, multiple constraints of the unit (such as main steam pressure and temperature constraints, actuator saturation constraints), nonlinear, multi-time scale coupling and the like, there are problems of slow response speed and low control precision, and it is difficult to simultaneously consider fast response, robustness and operation economy. SUMMARY

[0005] The purpose of the present application is to provide a dynamic load source network coupling nonlinear coordination control method of thermal power unit, and to solve the problems of slow response speed and low control precision existing in the existing control method.

[0006] The technical solution adopted by the present application is that the dynamic load source network coupling nonlinear coordination control method of thermal power unit is implemented according to the following steps: Step 1, establishing a unit operation measurement and data acquisition unit; Step 2, unified modeling of the source network and linearization of multiple working conditions; Step 3, designing and implementing an extended state observer (ESO); Step 4, ESO-MM-EMPC slow loop control design; Step 5, FOSMC+ESO fast loop control design is performed; Step 6, real-time control is performed by using multi-time scale cooperative control and constraint management.

[0007] The application is also characterized in that, Step 1 is specifically, Step 1.1, obtaining unit operation state and power grid environment information from a distributed control system; Step 1.2, pre-processing the data collected in step 1.1 by using signal filtering and bad point elimination.

[0008] The unit operation state reflects the boiler combustion and the steam-water system state, including main steam pressure , superheated steam temperature , boiler fuel quantity , air supply quantity , feed water flow , and the electrical and mechanical quantities of the generator and turbine operation state, also including the electromagnetic power , rotational speed / frequency ω, and steam valve opening output by the generator. The power grid environment includes the power grid frequency deviation Δf from the grid side and the AGC command AGC_cmd issued by the superior dispatch.

[0009] Step 2 is specifically, Step 2.1, constructing a boiler-side dynamic equation; The boiler-side energy conversion dynamic process includes two core parameters, main steam pressure and superheated steam temperature , and the dynamic behavior of the main steam pressure is described by a nonlinear differential equation as follows: (1) Wherein, is the rate of change of the main steam pressure, which is affected by the current main steam pressure , superheated steam temperature , fuel quantity , air supply quantity , feed water flow , and a random disturbance term representing unmodeled dynamics and the main steam pressure; The dynamic behavior of the superheated steam temperature is described by a nonlinear differential equation as follows: (2) Wherein, is the rate of change of the superheated steam temperature, is the random disturbance term of the superheated steam temperature; Step 2.2, establishing a turbine-side and generator-side model; Steam turbines and generators, as core equipment for energy conversion and electrical energy output, have relatively fast dynamic response speeds, and the mechanical power output of steam turbines... It is the main steam pressure Superheated steam temperature Steam valve opening Thermal efficiency and the random disturbance term of the steam turbine The nonlinear function is used to calculate the mechanical power output of the steam turbine. for; (3); Calculate rotor motion and electromagnetic transient processes: (4) in, Let be the inertial constant of the generator. The rate of change of rotational speed, The electromagnetic power output by the generator. The damping coefficient is... For speed deviation; This represents the random disturbance term of the generator; Step 2.3: Perform grid-side disturbance modeling and source-grid coupling; Step 2.4: Perform multi-condition linearization and multi-model management.

[0010] Step 2.3 specifically involves: Step 2.3.1, AGC instruction processing; The AGC command AGC_cmd from the power grid dispatch center, after a delay representing the communication and processing time, is converted into a load command for the generating units. ; Step 2.3.2, primary frequency modulation input; The grid frequency deviation Δf and the rate of change ROCOF are the direct input signals that trigger the primary frequency regulation action of the unit, converting the frequency deviation into a functional relationship of the turbine valve opening or load reference value correction. Step 2.3.3: Perform source-network coupling; Using a governor model and steam valve channels, load commands are transmitted. and frequency deviation Δf and turbine valve opening They are coupled together to form a complete closed loop from grid-side disturbances to unit response.

[0011] 6. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 5, characterized in that step 2.4 specifically comprises: Step 2.4.1: Select a typical load point and perform linearization; Within the actual operating range of the unit, several representative load points are selected as operating points. These operating points are then incorporated into the boiler-side dynamic equation model to establish turbine-side and generator-side models, grid-side disturbance modeling, and source-grid coupling model, resulting in an approximate linear model. Step 2.4.2: Establish the state space sub-model; For each boiler-side dynamic equation model, turbine-side and generator-side models are established, along with grid-side disturbance modeling and source-grid coupling model. A discrete-time state-space sub-model is then established, in standard form: (5) in, For operating condition index, For discrete time steps, This is a state vector, which includes the main steam pressure. Superheated steam temperature Rotational speed / frequency ω; The control input vector includes the fuel quantity. Steam turbine valve opening ; For disturbance input; (6) in, The output vector includes the electromagnetic power output by the generator. ,matrix The first The coefficient matrix of the system dynamic characteristics under each working condition index; Step 2.4.3, unify all uncertainties into a bounded set of perturbations: (7).

[0012] Step 3 specifically involves: Step 3.1, define the extended state; For the unit control system, it is represented as: (8) in, It is a state vector. It is the control input vector. It is an external disturbance; (9) in, It is the output vector, ESO will In addition to control input All other terms, including nonlinear terms, uncertain terms, and disturbance terms, are aggregated into a new state variable. The "total perturbation" is rewritten as an augmented linear system. (10) in, These are matrices of appropriate dimensions; (11) in, It is the rate of change of the total disturbance, expressed as; (12); Step 3.2, ESO update law; The initial state of the unit's control input is estimated using a state observer. Total disturbance In the discrete-time domain, the update law of ESO is expressed as: (13) (14) in, It is the system matrix of the augmented system. The observer gain matrix is... This is the actual measurement output of the system. It is the ESO's predicted output; (15) in, z k For ESO in The state estimation vector at time step (i.e., the estimated value of the original state). The estimated value of total disturbance ; The observer compares the error between the actual output and the predicted output. It continuously corrects its state estimation to achieve real-time tracking of system state and total disturbance; Step 3.3, Observer gain matrix tuning; Select the observer gain matrix The Whale Optimization Algorithm (WOA) is introduced, with the control performance of AGC as the optimization target. The control performance of AGC includes settling time, overshoot, and steady-state error. Within a preset search space, the bandwidth key parameters of ESO are calculated periodically. Step 3.4: Perform real-time disturbance estimation, and use the estimated value as feedforward compensation in slow loop and fast loop control; Step 3.4.1: Map the key physical quantities of the generating unit and the power grid into state and control quantities: The state vector is; (16); The output vector is; (17); The slow loop control quantity is; (18); The fast loop control quantity is; (19); External disturbances This includes load fluctuations, grid frequency deviation Δf, and changes in fuel quality; Step 3.4.2: Perform feedforward compensation in RMPC; In slow-loop robust model predictive control (RMPC), disturbance estimates of ESO are introduced. The prediction model is modified so that it can "sense" and compensate for model mismatch and external disturbances in real time. The revised prediction model is as follows: (20) Step 3.4.3: Perform feedforward compensation in FOSMC; In the fractional-order sliding mode control (FOSMC) of the fast loop, the disturbance estimate of the ESO is also used as feedforward compensation, and the control law of the fast loop is expressed as: (20) in, It is an equivalent control item.

[0013] Step 4 specifically involves: Step 4.1: Monitor the active power output of the generator unit. Main steam pressure And steam valve opening Key parameters determine the current operating condition range of the unit; Step 4.2, construct the Tube-RMPC optimization problem; Step 4.2.1, in the prediction time domain Internally, the slow-loop controller enables the unit's output to track the setpoint, while also considering the smoothness and economy of the control action; among which, the output reference trajectory In the AGC load command, the power component is... The main steam pressure and temperature components are the operating setpoints; Calculate the objective function objective function This includes penalties for output deviations and penalties for control increments; (twenty one) in, It is the first The predicted output of the step, It outputs the reference trajectory. It controls the increment; Step 4.2.2: Define constraints, including state constraints, control quantity constraints, control increment constraints, and robustness constraints; The state constraints require that the main steam pressure and main steam temperature must be within the upper and lower limits of safe operation, i.e. ; Control constraints ensure that fuel quantity, air supply volume, and feedwater quantity do not exceed the unit's limits, i.e. ; Incremental control constraints prevent actuators from moving too quickly by limiting the rate of change of the control variable. ; Robustness constraints, by introducing an auxiliary feedback control law, limit the amount of fuel, air, and water to a "pipeline" centered on the nominal trajectory. Step 4.2.3, Weight Matrix The elements determine the degree of importance attached to the tracking error of different output variables; the weight matrix The elements determine the penalty for controlling energy consumption. To achieve better control, the weight matrix... and Perform online optimization and selection; By constructing a comprehensive performance index This includes frequency deviation integral, dynamic indicators of adjustment time, and economic indicators of incremental coal consumption; using the particle swarm optimization (PSO) algorithm, the weight matrix is ​​optimized within a pre-defined search space. and Optimize the elements to find those that can make Minimize the weight combination to achieve adaptive adjustment of the control strategy and achieve the best balance between load response speed and economy; Step 4.3, ESO corrects the predicted trajectory; In each control cycle, the ESO unit outputs the current total disturbance estimate. The RMPC controller incorporates the total disturbance estimate into the prediction model to correct the state equations. (twenty two); Step 4.4, Online optimization solution and rolling control; Step 4.4.1: RMPC employs a Receding Horizon Control strategy, repeating the following steps at each sampling time: Measurement and estimation are performed to obtain the current state of the system and to estimate the total disturbance using ESO. Prediction and optimization: Based on the current state and the corrected model, the optimization problem is solved in the prediction time domain to obtain an optimal control sequence; Execution and rolling: Only the first control variable in the control sequence is executed. At the next sampling time, the system state is updated, and then the above process is repeated. Step 4.4.2: The slow-loop RMPC only issues the optimal control increment for the current moment in each control cycle. This is then combined with the existing fuel quantity, air supply, and water supply loops in the DCS to form a new setpoint. The system writes data into the boiler-side fuel, fan, and feedwater regulation loops via the DCS interface to achieve ESO-enhanced multi-model robust model predictive slow-loop control. The slow-loop RMPC controller, through forward-looking optimization, adjusts slow variables such as fuel and air volume in advance, providing sufficient energy support for the rapid action of the fast loop.

[0014] Step 5 specifically involves: Step 5.1: The fast-loop control unit directly acts on the unit's fast-response actuator, and its actual target is the turbine valve opening command in the DEH speed control system. The core objective is to quickly and accurately track AGC active power commands and respond rapidly to grid frequency deviations (Δf), thereby achieving high-quality primary frequency regulation. Step 5.2, the fractional-order sliding surface is designed as follows; (twenty three) in, Is the order of Fractional differential operators, These are weighting coefficients; fractional derivative terms improve the system's response speed, while fractional integral terms effectively eliminate steady-state errors. The fractional operator is implemented using the discretization approximation algorithm defined by Grünwald–Letnikov (GL). The discretization form defined by GL is as follows: (twenty four) in, It is the sampling period. These are the binomial coefficients; Step 5.3, ESO feedforward compensation; Similar to the slow loop, the ESO unit of the fast loop estimates the total disturbance on the electromechanical side in real time. , which is introduced into the control law as a feedforward compensation quantity; The equivalent control item is represented as: (25) in, It is an equivalent control term based on a simplified linear model. It is to compensate for the gain; Step 5.4, fractional sliding mode control law; Step 5.4.1, by order The equivalent control term is then obtained by solving for it. Its function is to make the system state move along the sliding surface under ideal conditions (i.e., without disturbance and with an accurate model); Step 5.4.2: The control law includes a switching term, which uses the saturated function sat with a boundary layer to replace the sign function sgn, switching the gain and the saturated function to overcome system uncertainties and disturbances, and suppress the high-frequency "chattering" phenomenon inherent in sliding mode control. The final control law is: (26) in, It's about switching the gain. It is the boundary layer thickness. When the system state is within the boundary layer, the control quantity becomes a continuous function, thereby effectively suppressing chattering. Step 5.4.3: Dynamic adjustment using an adaptive reaching law. To ensure robustness while minimizing switching gain; (27) in, It is a positive parameter. When the error is large, the gain is increased to speed up the approach speed, and when the error is small, the gain is decreased to reduce chattering, so as to achieve a good balance between dynamic performance and actuator life. Step 5.5: Implement the fast loop process to complete the fast loop control based on fractional sliding mode and ESO; The control flow of the fast loop is in each sampling cycle. Execution includes the following steps; Read real-time signals such as signals, AGC commands, frequency deviation Δf, and unit output P_e; ESO update, updating the total disturbance estimate on the electromechanical side using measurement signals. ; Calculate the sliding surface and the fractional-order sliding surface based on the current state. ; Calculate the control quantity, based on the current , and Calculate the final control quantity ; The system issues commands and sends the calculated control quantities to the DEH speed control system through the interface unit.

[0015] Step 6 specifically involves: Step 6.1, Constraint transmission from slow loop to fast loop; In each slow loop control cycle, the RMPC unit of the slow loop calculates the safe operating boundary of the unit for a future period based on the prediction results, and transmits this constraint information to the FOSMC controller of the fast loop. Step 6.1.1, the upper limit of the variable load rate can be used; The slow-loop RMPC dynamically calculates the maximum allowable load change rate of the unit based on the boiler's current heat storage status, fuel quantity level, and the stability of thermal parameters. When the fast-cycle loop executes the AGC command, the load change rate must not exceed the upper limit, thereby avoiding drastic fluctuations in boiler side parameters due to excessively rapid load changes, or even exceeding the limit. Step 6.1.2, safe operating range; The slow loop calculates the safe operating range of the main steam pressure and superheated steam temperature thermal parameters and transmits them to the fast loop. When the fast loop performs rapid frequency regulation, it ensures that the regulation behavior will not exceed the safe range. Step 6.2, feedback on the state and offset from the fast loop to the slow loop; In each fast loop control cycle, the fast loop controller records the actual active power output response and frequency deviation changes of the unit, and uses ESO to estimate the transient energy shift caused by rapid load changes. After filtering and downsampling, the high-frequency information is fed back to the slow loop. The slow loop controller uses the feedback information to correct the initial state and constraint margin of its prediction model, making the rolling optimization closer to the real-time dynamics of the unit, thereby making more accurate long-term decisions.

[0016] The beneficial effects of this invention are as follows: The coordinated control method of this invention, through precise compensation of internal and external disturbances by the ESO and the robust design of RMPC and FOSMC, can significantly improve the accuracy of load regulation. Through a multi-timescale coordination mechanism, perfect coordination between the load change rate and the main steam pressure is achieved. The rapid action of the fast loop is limited by the safety constraints calculated by the slow loop, ensuring that key thermal parameters such as the main steam pressure remain within a safe and stable range during rapid load changes. Attached Figure Description

[0017] Figure 1 This is a flowchart of the dynamic load source-grid coupling nonlinear coordinated control method for thermal power units according to the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0019] Example 1 The present invention provides a dynamic load source-grid coupled nonlinear coordinated control method for thermal power units, the process of which is as follows: Figure 1 As shown, the specific implementation steps are as follows: Step 1: Establish the unit operation measurement and data acquisition unit; Step 2: Perform unified source-network modeling and multi-condition linearization; Step 3: Design and implement the Extended State Observer (ESO); Step 4: Perform ESO-MM-EMPC slow-cycle control design; Step 5: Perform FOSMC+ESO fast environmental control design; Step 6: Implement real-time control using multi-timescale collaborative control and constraint management.

[0020] Example 2 The present invention provides a dynamic load source-grid coupled nonlinear coordinated control method for thermal power units, which is implemented according to the following steps: Step 1: Establish the unit operation measurement and data acquisition unit; Step 1.1: Obtain unit operating status and power grid environment information from the Distributed Control System (DCS); The unit's operating status reflects the state of the boiler combustion and steam-water system, including the main steam pressure. Superheated steam temperature Boiler fuel quantity Air volume Water supply flow rate ; and the electrical and mechanical quantities of the generator and turbine operating status, including the electromagnetic power output by the generator. Speed / frequency ω, valve opening ; The power grid environment includes the power grid frequency deviation Δf from the power grid side and the AGC command AGC_cmd issued by the superior dispatcher; Step 1.2 involves preprocessing the data collected in Step 1.1 using signal filtering and bad pixel removal to ensure the data is authentic and reliable.

[0021] Step 2: Perform unified source-network modeling and multi-condition linearization; Step 3: Design and implement the Extended State Observer (ESO); Step 4: Perform ESO-MM-EMPC slow-cycle control design; Step 5: Perform FOSMC+ESO fast environmental control design; Step 6: Implement real-time control using multi-timescale collaborative control and constraint management.

[0022] Example 3 The present invention provides a dynamic load source-grid coupled nonlinear coordinated control method for thermal power units, which is implemented according to the following steps: Step 1: Establish the unit operation measurement and data acquisition unit; Step 1.1: Obtain unit operating status and power grid environment information from the Distributed Control System (DCS); The unit's operating status reflects the state of the boiler combustion and steam-water system, including the main steam pressure. Superheated steam temperature Boiler fuel quantity Air volume Water supply flow rate ; and the electrical and mechanical quantities of the generator and turbine operating status, including the electromagnetic power output by the generator. Speed / frequency ω, valve opening ; The power grid environment includes the power grid frequency deviation Δf from the power grid side and the AGC command AGC_cmd issued by the superior dispatcher; Step 1.2 involves preprocessing the data collected in Step 1.1 using signal filtering and bad pixel removal to ensure the data is authentic and reliable.

[0023] Step 2: Perform unified source-network modeling and multi-condition linearization; Step 2.1, construct the dynamic equations on the boiler side; Step 2.2: Establish the turbine side and generator side models; Step 2.3: Perform grid-side disturbance modeling and source-grid coupling; Step 2.4: Perform multi-condition linearization and multi-model management.

[0024] Step 3: Design and implement the Extended State Observer (ESO); Step 4: Perform ESO-MM-EMPC slow-cycle control design; Step 5: Perform FOSMC+ESO fast environmental control design; Step 6: Implement real-time control using multi-timescale collaborative control and constraint management.

[0025] Example 4 The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units of the present invention is implemented according to the following steps: Step 1: Establish the unit operation measurement and data acquisition unit; Step 1.1: Obtain unit operating status and power grid environment information from the Distributed Control System (DCS); The unit's operating status reflects the state of the boiler combustion and steam-water system, including the main steam pressure. Superheated steam temperature Boiler fuel quantity Air volume Water supply flow rate ; and the electrical and mechanical quantities of the generator and turbine operating status, including the electromagnetic power output by the generator. Speed / frequency ω, valve opening ; The power grid environment includes the power grid frequency deviation Δf from the power grid side and the AGC command AGC_cmd issued by the superior dispatcher; Step 1.2 involves preprocessing the data collected in Step 1.1 using signal filtering and bad pixel removal to ensure the data is authentic and reliable.

[0026] Step 2: Perform unified source-network modeling and multi-condition linearization; Step 2.1, construct the dynamic equations on the boiler side; Step 2.2: Establish the turbine side and generator side models; Step 2.3: Perform grid-side disturbance modeling and source-grid coupling; Step 2.4: Perform multi-condition linearization and multi-model management.

[0027] Step 3: Design and implement the Extended State Observer (ESO); Step 3.1, define the extended state; Step 3.2, ESO update law; Step 3.3, Observer gain matrix tuning; Step 3.4: Perform real-time disturbance estimation, and use the estimated value as feedforward compensation in slow loop and fast loop control.

[0028] Step 4: Perform ESO-MM-EMPC slow-cycle control design; Step 5: Perform FOSMC+ESO fast environmental control design; Step 6: Implement real-time control using multi-timescale collaborative control and constraint management.

[0029] Example 5 The present invention provides a dynamic load source-grid coupled nonlinear coordinated control method for thermal power units, which is implemented according to the following steps: Step 1: Establish the unit operation measurement and data acquisition unit; Step 1.1: Obtain unit operating status and power grid environment information from the Distributed Control System (DCS); The unit's operating status reflects the state of the boiler combustion and steam-water system, including the main steam pressure. Superheated steam temperature Boiler fuel quantity Air volume Water supply flow rate ; and the electrical and mechanical quantities of the generator and turbine operating status, including the electromagnetic power output by the generator. Speed / frequency ω, valve opening ; The power grid environment includes the power grid frequency deviation Δf from the power grid side and the AGC command AGC_cmd issued by the superior dispatcher; Step 1.2 involves preprocessing the data collected in Step 1.1 using signal filtering and bad pixel removal to ensure the data is authentic and reliable.

[0030] Step 2: Perform unified source-network modeling and multi-condition linearization; Step 2.1, construct the dynamic equations on the boiler side; The dynamic process of energy conversion on the boiler side includes main steam pressure and superheated steam temperature Two core parameters describe the dynamic behavior of the main steam pressure using nonlinear differential equations: (1) in, The rate of change of main steam pressure is affected by the current main steam pressure. Superheated steam temperature Fuel quantity Air volume Water supply flow rate And a random perturbation term representing the unmodeled dynamics and main steam pressure. The impact; The dynamic behavior of superheated steam temperature can be described by nonlinear differential equations: (2) in, The rate of change of superheated steam temperature. This is a random disturbance term for the superheated steam temperature.

[0031] Step 2.2: Establish the turbine side and generator side models; Steam turbines and generators, as core equipment for energy conversion and electrical energy output, have relatively fast dynamic response speeds, and the mechanical power output of steam turbines... It is the main steam pressure Superheated steam temperature Steam valve opening Thermal efficiency and the random disturbance term of the steam turbine The nonlinear function is used to calculate the mechanical power output of the steam turbine. for; (3); Calculate rotor motion and electromagnetic transient processes: (4) in, Let be the inertial constant of the generator. The rate of change of rotational speed, The electromagnetic power output by the generator. The damping coefficient is... For speed deviation; This is the random disturbance term of the generator; the imbalance between the mechanical power output by the turbine and the electromagnetic power output by the generator causes changes in the generator speed, which is the basis for analyzing the stability of the power system.

[0032] Step 2.3: Perform grid-side disturbance modeling and source-grid coupling; Step 2.3.1, AGC instruction processing; The AGC command AGC_cmd from the power grid dispatch center, after a delay representing the communication and processing time, is converted into a load command for the generating units. ; Step 2.3.2, primary frequency modulation input; The grid frequency deviation Δf and the rate of change ROCOF are the direct input signals that trigger the primary frequency regulation action of the unit, converting the frequency deviation into a functional relationship of the turbine valve opening or load reference value correction. Step 2.3.3: Perform source-network coupling; Using a governor model and steam valve channels, load commands are transmitted. and frequency deviation Δf and turbine valve opening They are coupled together to form a complete closed loop from grid-side disturbances to unit response.

[0033] Step 2.4: Perform multi-condition linearization and multi-model management; Step 2.4.1: Select a typical load point and perform linearization; Within the actual operating range of the unit, several representative load points (e.g., 30%, 50%, 75%, 100% of rated load) are selected as operating points. These operating points are then incorporated into the boiler-side dynamic equation model to establish turbine-side and generator-side models, grid-side disturbance modeling, and source-grid coupling model, resulting in an approximate linear model.

[0034] Step 2.4.2: Establish the state space sub-model; For each boiler-side dynamic equation model, turbine-side and generator-side models are established, along with grid-side disturbance modeling and source-grid coupling model. A discrete-time state-space sub-model is then established, in standard form: (5) in, For operating condition index, For discrete time steps, This is a state vector, which includes the main steam pressure. Superheated steam temperature Rotational speed / frequency ω; The control input vector includes the fuel quantity. Steam turbine valve opening ; For disturbance input; (6) in, The output vector includes the electromagnetic power output by the generator. ,matrix The first The coefficient matrix of the system dynamic characteristics under each working condition index; Step 2.4.3, unify all uncertainties into a bounded set of perturbations: (7).

[0035] Step 3: Design and implement the Extended State Observer (ESO); Step 3.1, define the extended state; For the unit control system, it is represented as: (8) in, It is a state vector. It is the control input vector. It is an external disturbance; (9) in, It is the output vector, ESO will In addition to control input All other terms, including nonlinear terms, uncertain terms, and disturbance terms, are aggregated into a new state variable. The "total perturbation" is rewritten as an augmented linear system. (10) in, These are matrices of appropriate dimensions; (11) in, It is the rate of change of the total disturbance, expressed as; (12).

[0036] Step 3.2, ESO update law; The initial state of the unit's control input is estimated using a state observer. Total disturbance In the discrete-time domain, the update law of ESO is expressed as: (13) (14) in, It is the system matrix of the augmented system. The observer gain matrix is... This is the actual measurement output of the system. It is the ESO's predicted output; (15) in, z k For ESO in The state estimation vector at time step (i.e., the estimated value of the original state). The estimated value of total disturbance ; The observer compares the error between the actual output and the predicted output. It continuously corrects its state estimation to achieve real-time tracking of system state and total disturbance.

[0037] Step 3.3, Observer gain matrix tuning; Select the observer gain matrix The Whale Optimization Algorithm (WOA) is introduced, using the control performance of AGC as the optimization objective. The control performance of AGC includes settling time, overshoot, and steady-state error. Within a preset search space, the bandwidth of ESO (and its relation to the target value) is periodically optimized. Calculate key parameters that are closely related to each other.

[0038] Observer gain matrix The choice of the observer gain matrix directly determines the performance of the ESO, including the speed and accuracy of perturbation estimation and the ability to suppress measurement noise; A balance must be struck between speed and robustness; in actual operation, the unit's operating conditions and disturbance characteristics are time-varying, not a fixed... It may not be able to remain optimal in all situations.

[0039] Step 3.4: Perform real-time disturbance estimation, and use the estimated value as feedforward compensation in slow loop and fast loop control.

[0040] Step 3.4.1: Map the key physical quantities of the generating unit and the power grid into state and control quantities: The state vector is; (16); The output vector is; (17); The slow loop control quantity is; (18); The fast loop control quantity is; (19); External disturbances This includes load fluctuations, grid frequency deviation Δf, and changes in fuel quality.

[0041] Step 3.4.2: Perform feedforward compensation in RMPC; In slow-loop robust model predictive control (RMPC), disturbance estimates of ESO are introduced. The prediction model is modified so that it can "sense" and compensate for model mismatch and external disturbances in real time. The revised prediction model is as follows: (20) Step 3.4.3: Perform feedforward compensation in FOSMC; In the fractional-order sliding mode control (FOSMC) of the fast loop, the disturbance estimate of the ESO is also used as feedforward compensation, and the control law of the fast loop is expressed as: (20) in, It is an equivalent control item.

[0042] Step 4: Design the ESO-MM-RMPC slow loop control; Step 4.1: Monitor the active power output of the generator unit. Main steam pressure And steam valve opening Key parameters determine the current operating condition range of the unit; Step 4.2, construct the Tube-RMPC optimization problem; Step 4.2.1, in the prediction time domain Internally, the slow-loop controller causes the unit's output (such as...) The system tracks the setpoint while maintaining smoothness and economy in its control actions; the output reference trajectory is included. In the AGC load command, the power component is... The main steam pressure and temperature components are the operating setpoints; Calculate the objective function objective function This includes penalties for output deviations and penalties for control increments (or control values); (twenty one) in, It is the first The predicted output of the step, It outputs the reference trajectory (i.e., main steam pressure and temperature setpoint). It controls the increment; and It is a weight matrix used to balance tracking accuracy and control cost.

[0043] Step 4.2.2: Define constraints, including state constraints, control quantity constraints, control increment constraints, and robustness constraints; The state constraints require that the main steam pressure and main steam temperature must be within the upper and lower limits of safe operation, i.e. ; Control constraints ensure that fuel quantity, air supply volume, and feedwater quantity do not exceed the unit's limits, i.e. ; Incremental control constraints prevent actuators from moving too quickly by limiting the rate of change of the control variable. ; Robustness constraints (Tube constraints) introduce an auxiliary feedback control law to limit the amount of fuel, air, and water to a "tube" centered on the nominal trajectory, ensuring that the system state will not violate the constraints under all possible disturbances.

[0044] Step 4.2.3, Weight Matrix The elements determine the degree of importance attached to the tracking error of different output variables; the weight matrix The elements determine the penalty for controlling energy consumption. To achieve better control, the weight matrix... and Perform online optimization and selection; By constructing a comprehensive performance index This includes frequency deviation integral, dynamic indicators of adjustment time, and economic indicators of incremental coal consumption; using the particle swarm optimization (PSO) algorithm, the weight matrix is ​​optimized within a pre-defined search space. and Optimize the elements to find those that can make By minimizing the weight combination, the control strategy can be adaptively adjusted to achieve the best balance between load response speed and economy.

[0045] Step 4.3, ESO corrects the predicted trajectory; In each control cycle, the ESO unit outputs the current total disturbance estimate. The RMPC controller incorporates the total disturbance estimate into the prediction model to correct the state equations. (twenty two); Step 4.4, Online optimization solution and rolling control; Step 4.4.1: RMPC employs a Receding Horizon Control strategy, repeating the following steps at each sampling time: Measurement and estimation are performed to obtain the current state of the system and to estimate the total disturbance using ESO. Prediction and optimization: Based on the current state and the corrected model, the optimization problem is solved in the prediction time domain to obtain an optimal control sequence; Execution and rolling: Only the first control variable in the control sequence is executed. At the next sampling time, the system state is updated, and then the above process is repeated.

[0046] Step 4.4.2: The slow-loop RMPC only issues the optimal control increment for the current moment in each control cycle. This is then combined with the existing fuel quantity, air supply, and water supply loops in the DCS to form a new setpoint. The system writes data into the boiler-side fuel, fan, and feedwater regulation loops via the DCS interface to achieve ESO-enhanced multi-model robust model predictive slow-loop control. The slow-loop RMPC controller, through forward-looking optimization, adjusts slow variables such as fuel and air volume in advance, providing sufficient energy support for the rapid action of the fast loop.

[0047] Step 5: Perform FOSMC+ESO fast loop control; Step 5.1: The fast-loop control unit directly acts on the unit's fast-response actuator, and its actual target is the turbine valve opening command in the DEH speed control system. The core objective is to quickly and accurately track AGC active power commands and respond rapidly to grid frequency deviations (Δf), thereby achieving high-quality primary frequency regulation. Step 5.2, the fractional-order sliding surface is designed as follows; (twenty three) in, Is the order of Fractional differential operators, These are weighting coefficients. Fractional derivative terms improve the system's response speed, while fractional integral terms effectively eliminate steady-state errors. The fractional operator is implemented using the discretization approximation algorithm defined by Grünwald–Letnikov (GL). The discretization form defined by GL is as follows: (twenty four) in, It is the sampling period. It is the binomial coefficient.

[0048] Step 5.3, ESO feedforward compensation; Similar to the slow loop, the ESO unit of the fast loop estimates the total disturbance on the electromechanical side in real time. , which is introduced into the control law as a feedforward compensation quantity; The equivalent control item is represented as: (25) in, It is an equivalent control term based on a simplified linear model. It is to compensate for the gain; Step 5.4, fractional sliding mode control law; Step 5.4.1, by order The equivalent control term is then obtained by solving for it. Its function is to make the system state move along the sliding surface under ideal conditions (i.e., without disturbance and with an accurate model); Step 5.4.2: The control law includes a switching term, which uses the saturated function sat with a boundary layer to replace the sign function sgn, switching the gain and the saturated function to overcome system uncertainties and disturbances, and suppress the high-frequency "chattering" phenomenon inherent in sliding mode control. The final control law is: (26) in, It's about switching the gain. It is the boundary layer thickness. When the system state is within the boundary layer, the control quantity becomes a continuous function, thereby effectively suppressing chattering.

[0049] Step 5.4.3: Dynamic adjustment using an adaptive reaching law. To ensure robustness while minimizing switching gain; (27) in, It is a positive parameter. When the error is large, the gain is increased to speed up the approach speed, and when the error is small, the gain is decreased to reduce chattering, so as to achieve a good balance between dynamic performance and actuator life.

[0050] Step 5.5: Implement the fast loop process to complete the fast loop control based on fractional sliding mode and ESO.

[0051] The control flow of the fast loop is in each sampling cycle. Execution takes 200 milliseconds and includes the following steps; Read real-time signals such as signals, AGC commands, frequency deviation Δf, and unit output P_e; ESO update, updating the total disturbance estimate on the electromechanical side using measurement signals. ; Calculate the sliding surface and the fractional-order sliding surface based on the current state. ; Calculate the control quantity, based on the current , and Calculate the final control quantity ; The system issues commands and sends the calculated control quantities to the DEH speed control system through the interface unit.

[0052] Step 6: Perform multi-timescale collaborative control and constraint management.

[0053] Step 6.1, Constraint transmission from slow loop to fast loop; In each slow loop control cycle (1 second), the RMPC unit of the slow loop calculates the safe operating boundary of the unit for a future period based on the prediction results, and transmits this constraint information to the FOSMC controller of the fast loop. Step 6.1.1, the upper limit of the variable load rate can be used; The slow-loop RMPC dynamically calculates the maximum allowable load change rate of the unit based on the boiler's current heat storage status, fuel quantity level, and the stability of thermal parameters. When the fast-acting boiler executes the AGC command, the load change rate must not exceed the upper limit, thereby avoiding drastic fluctuations in boiler-side parameters or even exceeding the limit due to excessively rapid load changes.

[0054] Step 6.1.2, safe operating range; The slow loop calculates the safe operating range of the main steam pressure and superheated steam temperature thermal parameters and transmits them to the fast loop. When the fast loop performs rapid frequency regulation, it ensures that the regulation behavior will not exceed the safe range.

[0055] Step 6.2, feedback on the state and offset from the fast loop to the slow loop; In each fast loop control cycle (millisecond level), the fast loop controller records the unit's actual active power output response and frequency deviation changes, and uses ESO to estimate transient energy shifts caused by rapid load changes (such as instantaneous changes in boiler heat storage). After filtering and downsampling, the high-frequency information is fed back to the slow loop controller. The slow loop controller uses the feedback information to correct the initial state and constraint margin of its prediction model, making the rolling optimization closer to the unit's real-time dynamics, thereby making more accurate long-term decisions.

[0056] Example 6 Overall structure and final control output of the control system The source-network coupled coordinated control system of this invention works in conjunction with the existing DCS / DEH architecture in engineering implementation, and the closed-loop signal flow is as follows: (1) Measurement and command: Data is collected through the unit's DCS / DEH. Unit status parameters and grid frequency deviation AGC load command .

[0057] (2) Slow-loop ESO-MM-Tube-RMPC: Taking the current state and the total disturbance estimate of the ESO output as input, the optimal control sequence is solved by rolling based on the multi-condition linear sub-model, and the setpoints of fuel quantity, air supply quantity and water supply quantity are output. By writing data into the boiler-side actuators via DCS, predictive regulation of main steam pressure, superheated steam temperature, and feedwater can be achieved.

[0058] (3) Fast Loop ESO+FOSMC: Using AGC load commands Frequency deviation Actual electrical power The total disturbance estimate on the electromechanical side of the ESO is used as input to calculate the valve opening command. The DEH speed control system acts on the steam turbine to quickly track AGC commands and participate in primary frequency regulation.

[0059] (4) Multi-timescale coordination: The slow loop recalculates the maximum allowable load change rate and the main steam parameter safety boundary every 1 second and transmits it to the fast loop as a constraint; the fast loop records the actual active response and transient heat storage offset every 200ms, and feeds it back to the slow loop after filtering and downsampling, so as to correct the prediction model and constraint margin for the next cycle.

[0060] This invention presents a dynamic load source-grid coupled nonlinear coordinated control method for thermal power units. It decouples the complex dynamic process of the unit into two control loops with different time scales: fast and slow. Targeted advanced control algorithms are then used to optimize each loop. The slow time scale processes, such as boiler-side fuel, air, and feedwater regulation, are handled by multi-model robust model predictive control (ESO-MM-RMPC) enhanced by an extended state observer (ESO) to achieve predictive and economical regulation of future load demand and ensure the stability of key thermal parameters. The fast time scale processes, such as turbine control valves, are dominated by fractional-mode sliding mode control (FOSMC+ESO) based on ESO to achieve millisecond-level fast and robust response to AGC commands and grid frequency disturbances. Through multi-time scale coordination and constraint management units, information exchange and constraint transmission between the fast and slow loops are achieved, forming an organic and intelligent coordinated control system. Ultimately, this achieves the comprehensive goal of improving unit AGC performance, primary frequency regulation capability, and operational economy.

Claims

1. A dynamic load source-grid coupled nonlinear coordinated control method for thermal power units, characterized in that, The specific steps are as follows: Step 1: Establish the unit operation measurement and data acquisition unit; Step 2: Perform unified source-network modeling and multi-condition linearization; Step 3: Design and implement the Extended State Observer (ESO); Step 4: Perform ESO-MM-EMPC slow-cycle control design; Step 5: Perform FOSMC+ESO fast environmental control design; Step 6: Implement real-time control using multi-timescale collaborative control and constraint management.

2. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 1, characterized in that, Step 1 specifically involves: Step 1.1: Obtain unit operating status and power grid environment information from the distributed control system; Step 1.2 involves preprocessing the data collected in Step 1.1 using signal filtering and bad pixel removal.

3. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 2, characterized in that, The unit's operating status reflects the state of the boiler combustion and steam-water system, including the main steam pressure. Superheated steam temperature Boiler fuel quantity Air volume Water supply flow rate ; In addition to the electrical and mechanical quantities of the generator and turbine operating status, it also includes the electromagnetic power output by the generator. Speed / frequency ω, valve opening ; The power grid environment includes the power grid frequency deviation Δf from the power grid side and the AGC command AGC_cmd issued by the superior dispatcher.

4. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 3, characterized in that, Step 2 specifically involves: Step 2.1, construct the dynamic equations on the boiler side; The dynamic process of energy conversion on the boiler side includes main steam pressure and superheated steam temperature Two core parameters describe the dynamic behavior of the main steam pressure using nonlinear differential equations: （1） in, The rate of change of main steam pressure is affected by the current main steam pressure. Superheated steam temperature Fuel quantity Air volume Water supply flow rate And a random perturbation term representing the unmodeled dynamics and main steam pressure. The impact; The dynamic behavior of superheated steam temperature can be described by nonlinear differential equations: （2） in, The rate of change of superheated steam temperature. This represents the random disturbance term for the superheated steam temperature; Step 2.2: Establish the turbine side and generator side models; Steam turbines and generators, as core equipment for energy conversion and electrical energy output, have relatively fast dynamic response speeds, and the mechanical power output of steam turbines... It is the main steam pressure Superheated steam temperature Steam valve opening Thermal efficiency and the random disturbance term of the steam turbine The nonlinear function is used to calculate the mechanical power output of the steam turbine. for; （3）； Calculate rotor motion and electromagnetic transient processes: （4） in, Let be the inertial constant of the generator. The rate of change of rotational speed, The electromagnetic power output by the generator. The damping coefficient is... For speed deviation; This represents the random disturbance term of the generator; Step 2.3: Perform grid-side disturbance modeling and source-grid coupling; Step 2.4: Perform multi-condition linearization and multi-model management.

5. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 4, characterized in that, Step 2.3 specifically involves: Step 2.3.1, AGC instruction processing; The AGC command AGC_cmd from the power grid dispatch center, after a delay representing the communication and processing time, is converted into a load command for the generating units. ; Step 2.3.2, primary frequency modulation input; The grid frequency deviation Δf and the rate of change ROCOF are the direct input signals that trigger the primary frequency regulation action of the unit, converting the frequency deviation into a functional relationship of the turbine valve opening or load reference value correction. Step 2.3.3: Perform source-network coupling; Using a governor model and steam valve channels, load commands are transmitted. and frequency deviation Δf and turbine valve opening They are coupled together to form a complete closed loop from grid-side disturbances to unit response.

6. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 5, characterized in that, Step 2.4 specifically involves, Step 2.4.1: Select a typical load point and perform linearization; Within the actual operating range of the unit, several representative load points are selected as operating points. These operating points are then incorporated into the boiler-side dynamic equation model to establish turbine-side and generator-side models, grid-side disturbance modeling, and source-grid coupling model, resulting in an approximate linear model. Step 2.4.2: Establish the state space sub-model; For each boiler-side dynamic equation model, turbine-side and generator-side models are established, along with grid-side disturbance modeling and source-grid coupling model. A discrete-time state-space sub-model is then established, in standard form: （5） in, For operating condition index, For discrete time steps, This is a state vector, which includes the main steam pressure. Superheated steam temperature Rotational speed / frequency ω; The control input vector includes the fuel quantity. Steam turbine valve opening ; For disturbance input; （6） in, The output vector includes the electromagnetic power output by the generator. ,matrix The first The coefficient matrix of the system dynamic characteristics under each working condition index; Step 2.4.3, unify all uncertainties into a bounded set of perturbations: （7）。 7. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 6, characterized in that, Step 3 specifically involves: Step 3.1, define the extended state; For the unit control system, it is represented as: （8） in, It is a state vector. It is the control input vector. It is an external disturbance; （9） in, It is the output vector, ESO will In addition to control input All other terms, including nonlinear terms, uncertain terms, and disturbance terms, are aggregated into a new state variable. The "total perturbation" is rewritten as an augmented linear system. （10） in, These are matrices of appropriate dimensions; （11） in, It is the rate of change of the total disturbance, expressed as; （12）； Step 3.2, ESO update law; The initial state of the unit's control input is estimated using a state observer. Total disturbance In the discrete-time domain, the update law of ESO is expressed as: （13） （14） in, It is the system matrix of the augmented system. The observer gain matrix is... This is the actual measurement output of the system. It is the ESO's predicted output; （15） in, z k For ESO in The state estimation vector at time step (i.e., the estimated value of the original state). The estimated value of total disturbance ; The observer compares the error between the actual output and the predicted output. It continuously corrects its state estimation to achieve real-time tracking of system state and total disturbance; Step 3.3, Observer gain matrix tuning; Select the observer gain matrix The Whale Optimization Algorithm (WOA) is introduced, with the control performance of AGC as the optimization target. The control performance of AGC includes settling time, overshoot, and steady-state error. Within a preset search space, the bandwidth key parameters of ESO are calculated periodically. Step 3.4: Perform real-time disturbance estimation, and use the estimated value as feedforward compensation in slow loop and fast loop control; Step 3.4.1: Map the key physical quantities of the generating unit and the power grid into state and control quantities: The state vector is; （16）； The output vector is; （17）； The slow loop control quantity is; （18）； The fast loop control quantity is; （19）； External disturbances This includes load fluctuations, grid frequency deviation Δf, and changes in fuel quality; Step 3.4.2: Perform feedforward compensation in RMPC; In slow-loop robust model predictive control (RMPC), disturbance estimates of ESO are introduced. The prediction model is modified so that it can "sense" and compensate for model mismatch and external disturbances in real time. The revised prediction model is as follows: （20） Step 3.4.3: Perform feedforward compensation in FOSMC; In the fractional-order sliding mode control (FOSMC) of the fast loop, the disturbance estimate of the ESO is also used as feedforward compensation, and the control law of the fast loop is expressed as: （20） in, It is an equivalent control item.

8. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 7, characterized in that, Step 4 specifically involves: Step 4.1: Monitor the active power output of the generator unit. Main steam pressure And steam valve opening Key parameters determine the current operating condition range of the unit; Step 4.2, construct the Tube-RMPC optimization problem; Step 4.2.1, in the prediction time domain Internally, the slow-loop controller causes the unit's output (such as...) The system tracks the setpoint while maintaining smoothness and economy in its control actions; the output reference trajectory is included. In the AGC load command, the power component is... The main steam pressure and temperature components are the operating setpoints; Calculate the objective function objective function This includes penalties for output deviations and penalties for control increments; （21） in, It is the first The predicted output of the step, It outputs the reference trajectory. It controls the increment; Step 4.2.2: Define constraints, including state constraints, control quantity constraints, control increment constraints, and robustness constraints; The state constraints require that the main steam pressure and main steam temperature must be within the upper and lower limits of safe operation, i.e. ; Control constraints ensure that fuel quantity, air supply volume, and feedwater quantity do not exceed the unit's limits, i.e. ; Incremental control constraints prevent actuators from moving too quickly by limiting the rate of change of the control variable. ; Robustness constraints, by introducing an auxiliary feedback control law, limit the amount of fuel, air, and water to a "pipeline" centered on the nominal trajectory. Step 4.2.3, Weight Matrix The elements determine the degree of importance attached to the tracking error of different output variables; the weight matrix The elements determine the penalty for controlling energy consumption. To achieve better control, the weight matrix... and Perform online optimization and selection; By constructing a comprehensive performance index This includes frequency deviation integral, dynamic indicators of adjustment time, and economic indicators of incremental coal consumption; using the particle swarm optimization (PSO) algorithm, the weight matrix is ​​optimized within a pre-defined search space. and Optimize the elements to find those that can make Minimize the weight combination to achieve adaptive adjustment of the control strategy and achieve the best balance between load response speed and economy; Step 4.3, ESO corrects the predicted trajectory; In each control cycle, the ESO unit outputs the current total disturbance estimate. The RMPC controller incorporates the total disturbance estimate into the prediction model to correct the state equations. （22）； Step 4.4, Online optimization solution and rolling control; Step 4.4.1: RMPC employs a Receding Horizon Control strategy, repeating the following steps at each sampling time: Measurement and estimation are performed to obtain the current state of the system and to estimate the total disturbance using ESO. Prediction and optimization: Based on the current state and the corrected model, the optimization problem is solved in the prediction time domain to obtain an optimal control sequence; Execution and rolling: Only the first control variable in the control sequence is executed. At the next sampling time, the system state is updated, and then the above process is repeated. Step 4.4.2: The slow-loop RMPC only issues the optimal control increment for the current moment in each control cycle. This is then combined with the existing fuel quantity, air supply, and water supply loops in the DCS to form a new setpoint. The system writes data into the boiler-side fuel, fan, and feedwater regulation loops via the DCS interface to achieve ESO-enhanced multi-model robust model predictive slow-loop control. The slow-loop RMPC controller, through forward-looking optimization, adjusts slow variables such as fuel and air volume in advance, providing sufficient energy support for the rapid action of the fast loop.

9. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 8, characterized in that, Step 5 specifically involves: Step 5.1: The fast-loop control unit directly acts on the unit's fast-response actuator, and its actual target is the turbine valve opening command in the DEH speed control system. The core objective is to quickly and accurately track AGC active power commands and respond rapidly to grid frequency deviations (Δf), thereby achieving high-quality primary frequency regulation. Step 5.2, the fractional-order sliding surface is designed as follows; （23） in, Is the order of Fractional differential operators, These are weighting coefficients; fractional derivative terms improve the system's response speed, while fractional integral terms effectively eliminate steady-state errors. The fractional operator is implemented using the discretization approximation algorithm defined by Grünwald–Letnikov (GL). The discretization form defined by GL is as follows: （24） in, It is the sampling period. These are the binomial coefficients; Step 5.3, ESO feedforward compensation; Similar to the slow loop, the ESO unit of the fast loop estimates the total disturbance on the electromechanical side in real time. , which is introduced into the control law as a feedforward compensation quantity; The equivalent control item is represented as: （25） in, It is an equivalent control term based on a simplified linear model. It is to compensate for the gain; Step 5.4, fractional sliding mode control law; Step 5.4.1, by order The equivalent control term is then obtained by solving for it. Its function is to make the system state move along the sliding surface under ideal conditions (i.e., without disturbance and with an accurate model); Step 5.4.2: The control law includes a switching term, which uses the saturated function sat with a boundary layer to replace the sign function sgn, switching the gain and the saturated function to overcome system uncertainties and disturbances, and suppress the high-frequency "chattering" phenomenon inherent in sliding mode control. The final control law is: （26） in, It's about switching the gain. It is the boundary layer thickness. When the system state is within the boundary layer, the control quantity becomes a continuous function, thereby effectively suppressing chattering. Step 5.4.3: Dynamic adjustment using an adaptive reaching law. To ensure robustness while minimizing switching gain; （27） in, It is a positive parameter. When the error is large, the gain is increased to speed up the approach speed, and when the error is small, the gain is decreased to reduce chattering, so as to achieve a good balance between dynamic performance and actuator life. Step 5.5: Implement the fast loop process to complete the fast loop control based on fractional sliding mode and ESO; The control flow of the fast loop is in each sampling cycle. Execution includes the following steps; Read real-time signals such as signals, AGC commands, frequency deviation Δf, and unit output P_e; ESO update, updating the total disturbance estimate on the electromechanical side using measurement signals. ; Calculate the sliding surface and the fractional-order sliding surface based on the current state. ; Calculate the control quantity, based on the current , and Calculate the final control quantity ; The system issues commands and sends the calculated control quantities to the DEH speed control system through the interface unit.

10. The dynamic load source-grid coupled nonlinear coordinated control method for thermal power units according to claim 9, characterized in that, Step 6 specifically involves: Step 6.1, Constraint transmission from slow loop to fast loop; In each slow loop control cycle, the RMPC unit of the slow loop calculates the safe operating boundary of the unit for a future period based on the prediction results, and transmits this constraint information to the FOSMC controller of the fast loop. Step 6.1.1, the upper limit of the variable load rate can be used; The slow-loop RMPC dynamically calculates the maximum allowable load change rate of the unit based on the boiler's current heat storage status, fuel quantity level, and the stability of thermal parameters. When the fast-cycle loop executes the AGC command, the load change rate must not exceed the upper limit, thereby avoiding drastic fluctuations in boiler side parameters due to excessively rapid load changes, or even exceeding the limit. Step 6.1.2, safe operating range; The slow loop calculates the safe operating range of the main steam pressure and superheated steam temperature thermal parameters and transmits them to the fast loop. When the fast loop performs rapid frequency regulation, it ensures that the regulation behavior will not exceed the safe range. Step 6.2, feedback on the state and offset from the fast loop to the slow loop; In each fast loop control cycle, the fast loop controller records the actual active power output response and frequency deviation changes of the unit, and uses ESO to estimate the transient energy shift caused by rapid load changes. After filtering and downsampling, the high-frequency information is fed back to the slow loop. The slow loop controller uses the feedback information to correct the initial state and constraint margin of its prediction model, making the rolling optimization closer to the real-time dynamics of the unit, thereby making more accurate long-term decisions.

Images