Cooperative control method suitable for thermoelectric unit

Abstract

The invention provides a cooperative control method suitable for a thermoelectric unit, and relates to the technical field of thermoelectric units, and the method comprises the steps: constructing a machine-furnace coupling nonlinear state space model, designing a dynamic heat storage state observer, calculating a heat storage state index in real time to quantify the transient energy profit and loss of the unit, and building a multi-objective optimization function. And a dynamic weighting factor based on the index is introduced, the factor guides the controller to automatically switch strategies under different working conditions according to the priority of the real-time energy state, the intelligent balancing load response speed and the main steam pressure stability, and finally, a nonlinear model predictive control algorithm is adopted to solve in a rolling time domain, so that the real-time energy state is obtained. According to the method, the optimal steam turbine control valve and fuel quantity control increment is obtained, decision-making level deep cooperation of a turbine-boiler system is effectively achieved, the problem of divergence control under deep peak regulation is solved, the frequency modulation potential of a unit is released to the maximum extent on the premise that absolute safety of pressure is guaranteed, and the response rate of a power grid is increased.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heat and power generating unit, in particular to a coordinated control method suitable for heat and power generating unit. BACKGROUND

[0002] According to the analysis method for intelligent energy consumption of a combined heat and power generating unit disclosed in Chinese patent No. "CN114626270A", the following steps are included: S1, using a 600MW thermal power generating unit to perform cylinder cutting and heat supply modification work, comprehensively improving all key point measuring points of the low pressure cylinder, and combining finite element numerical analysis simulation; S2, taking a 600MW combined heat and power generating unit as the object; the present application supports each other through the simulation modeling of the long heat supply process of the combined heat and power generating unit and its energy consumption characteristics, realizes the visualization of energy consumption analysis, and provides a basis for heat and power coordinated control, then explores the characteristics of the cylinder cutting operation of the 600MW unit from the short time and deep heat and power decoupling characteristics under long heat supply, provides a basis for the heat and power coordinated automatic optimization control scheme, deepens the automatic optimization control system based on energy saving, and develops a full intelligent heat and power coordinated control platform of the 600MW thermal power generating unit, so that the energy consumption can be quickly and effectively analyzed, the energy waste is avoided, the production cost is reduced, and the energy saving effect is improved.

[0003] According to the optimization income method for a combined heat and power generating unit disclosed in Chinese patent No. "CN114580304A", the following steps are included: S1, using a 600MW unit to perform cylinder cutting and heat supply modification, and then using a completely sealable hydraulic butterfly valve to cut off the original steam inlet pipeline of the low pressure cylinder; S2, adding an online monitoring system to safely monitor the blade operation state, with a monitoring interval of 5-10 minutes. The present application modifies the combined heat and power generating system by using the cylinder cutting and heat supply technology of the 600MW unit, then simulates and models the long heat supply process of the unit, then analyzes the characteristics of the combined heat and power energy consumption through an intelligent energy consumption calculation platform, then calculates the optimization control scheme of the heat and power coordinated system based on the energy saving of the unit, and then automatically controls the unit throughout the process through the intelligent heat and power coordinated control system of the thermal power generating unit according to the calculation results, so as to save the energy consumed by the unit in operation and improve the production income.

[0004] The above patent documents and prior art have the following technical problems in use: Problem one, the existing coordinated control is mostly based on static balance or simple deviation feedback, lacking quantitative evaluation of the real-time heat storage capacity of the boiler, resulting in overdraw or conservative use of boiler heat storage when the AGC (automatic generation control) instruction of the power grid changes dramatically, causing the main steam pressure to exceed the limit or the load response to lag; Question 2: There is a strong nonlinear coupling between the steam turbine control valve and the boiler combustion system of the thermal power unit. Traditional PID control is difficult to balance load tracking accuracy and parameter stability under deep peak shaving (such as 40% rated load) and rapid load change conditions. Thirdly, the dynamic characteristics of the unit vary greatly under different load conditions. Controllers with fixed parameters cannot adapt to full operating conditions, and existing technologies lack effective compensation mechanisms for unpredictable disturbances such as coal quality fluctuations. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a collaborative control method suitable for thermal power units, solving the technical problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a cooperative control method suitable for thermal power units, the cooperative control method comprising the following steps: Sp1: Real-time acquisition of operating status data of thermal power units; Sp2: Construct a boiler-machine coupled nonlinear state-space model to describe the relationship between boiler energy storage and turbine energy conversion, and design a dynamic thermal storage state observer based on the model. Utilize the collected operating status data to calculate the thermal storage state index, which characterizes the unit's transient energy throughput capacity, in real time. Sp3: Establish a multi-objective collaborative optimization objective function that includes load tracking error, main steam pressure deviation, and control quantity change rate, and introduce a dynamic weighting factor that is adjusted in real time based on the heat storage state index. The dynamic weighting factor is used to weigh the priority between load response speed and pressure stability. Sp4: Using a nonlinear model predictive control algorithm, the multi-objective collaborative optimization objective function is solved in the rolling optimization time domain to calculate the optimal turbine control valve increment and fuel quantity control increment at the current moment; Sp5: The calculated optimal control increment is superimposed on the current control command and sent to the turbine actuator and the boiler combustion actuator respectively to complete the coordinated control of the thermal power unit.

[0007] Preferably, the operating status data in Sp1 includes at least the actual power generation, main steam pressure, main steam temperature, turbine valve opening, fuel quantity command, and grid automatic power generation control load command. It also includes estimated data on the calorific value of coal entering the boiler. The method for obtaining this data is as follows: by using real-time collected flue gas temperature signal, flue gas oxygen signal, and total air volume signal entering the boiler, the real-time calorific value of coal is calculated in reverse using the heat balance principle. The calculated calorific value of coal is then input as a measurable disturbance variable into the nonlinear model predictive control algorithm of Sp4 to correct the gain of the fuel quantity control command in real time.

[0008] Preferably, the calculation method of the thermal storage state index in Sp2 is as follows: multiply the differential value of the main steam pressure by the preset boiler thermal storage coefficient to obtain the first component, and multiply the difference between the current estimated value of fuel heat release and the actual power generation by the energy conversion gain coefficient to obtain the second component. The thermal storage state index is the weighted sum of the first component and the second component, which is used to quantitatively characterize the degree of energy surplus or deficit in the boiler metal and working fluid that can be quickly released or needs to be replenished at the current moment.

[0009] Preferably, the adjustment mechanism of the dynamic weighting factor in Sp3 is as follows: a thermal storage safety threshold is set. When the absolute value of the thermal storage state index calculated in real time is less than the thermal storage safety threshold, the weight of the corresponding main steam pressure deviation item in the dynamic weighting factor is reduced, so that the controller can give priority to using the unit's thermal storage to quickly respond to load commands. When the absolute value of the thermal storage state index calculated in real time is greater than or equal to the thermal storage safety threshold, the weight of the corresponding main steam pressure deviation item in the dynamic weighting factor is increased in a non-linear trend, so as to forcibly suppress the turbine regulating valve action amplitude and give priority to adjusting the fuel quantity to restore the main steam pressure.

[0010] Preferably, the nonlinear model predictive control algorithm in Sp4 also incorporates a feedforward compensation mechanism based on the load command change rate: the load change rate is obtained by calculating the derivative of the automatic power generation control load command with respect to time. When the absolute value of the load change rate exceeds the preset dead zone, the inertial differential feedforward quantity is calculated based on the magnitude of the load change rate. The inertial differential feedforward quantity is directly superimposed on the fuel quantity control increment optimized by Sp4 to compensate for the inertial delay of the boiler combustion system in advance.

[0011] Preferably, the parameter update method of the turbine-boiler coupled nonlinear state-space model in Sp2 is as follows: using the online least squares method with a forgetting factor, the static gain parameter and inertia time constant parameter in the model are identified and updated in real time using the input and output data in the most recent time window, so as to adapt to the dynamic characteristic changes of the unit under different load conditions.

[0012] Preferably, Sp5 further includes an actuator constraint correction step before sending the control command: based on the current number of operating coal mills and the maximum output limit of a single coal mill, it is determined whether the fuel quantity command calculated by Sp4 exceeds the limit. If it does, discrete variables of coal mill start-up and shutdown operations are introduced in the rolling optimization time domain to predict fuel quantity disturbances during coal mill switching and to reverse-correct the turbine control valve increment to smooth out expected pressure fluctuations.

[0013] Preferably, the collaborative control method is configured with two operating modes according to the grid demand: the first is the pressure-mode-dominated mode, which locks the dynamic weighting factor to a larger value so that the control system prioritizes maintaining the main steam pressure near the rated value under any operating condition, and is suitable for the grid load stability period; the second is the source-mode-dominated mode, which activates the real-time adjustment function of the dynamic weighting factor, allowing the main steam pressure to deviate from the set value within a safe range, and uses heat storage to increase the unit's load change rate, and is suitable for the period of high grid frequency regulation demand.

[0014] Preferably, in Sp4, the solution process in the rolling optimization time domain adopts a sequential quadratic programming method: the nonlinear optimization problem is linearized at each sampling time, transformed into a quadratic programming subproblem for solution, and the optimal solution at the previous sampling time is used as the initial value for the current iteration to meet the computational speed requirements of real-time control.

[0015] The present invention has the following beneficial effects: This invention achieves real-time quantitative and digital control of the unit's thermal storage state by constructing a dynamic thermal storage state observer and introducing a multi-objective collaborative optimization mechanism with dynamic weighting factors. This mechanism can automatically adjust the weighting priority of load response and pressure stability at the microsecond level based on energy surplus / deficit status: when thermal storage is sufficient, the pressure stability weight is reduced, fully utilizing boiler thermal inertia to significantly open the control valves to meet the grid's urgent needs; when thermal storage is overdrawn, the pressure weight is forcibly increased, limiting control valve action and prioritizing supplementary combustion to prevent pressure exceeding limits. This dynamic trade-off mechanism solves the risk of unit safety shutdown caused by blind adjustments when AGC commands change drastically. While ensuring absolute safety of main steam pressure, it maximizes the unit's frequency regulation potential and improves the grid response rate.

[0016] This invention employs a nonlinear model predictive control algorithm based on sequential quadratic programming, establishing a nonlinear state-space model of turbine-boiler coupling. This transforms the complex, strongly coupled control problem into a multi-objective constrained optimization problem in the rolling time domain. The controller does not independently regulate the turbine or boiler; instead, in each calculation, it simultaneously solves for the optimal turbine control valve increment and fuel quantity control increment that balance the constraints of both. This achieves deep collaboration between the turbine and boiler decision-making levels, overcoming the industry challenge of control divergence caused by extremely strong nonlinear characteristics under deep peak shaving conditions (e.g., 40% load). Through model prediction, the system can anticipate the reverse impact of valve action on pressure and compensate in advance, effectively eliminating mutual interference and oscillations between control quantities. This enables the unit to maintain high-precision load tracking capability across the entire load range, not only solving the problem of low automatic control availability under low load conditions but also significantly reducing ineffective reciprocating movements of actuators and extending equipment lifespan.

[0017] 3. This invention integrates two major technical means: online model adaptation and disturbance source feedforward compensation. On the one hand, it uses recursive least squares method to update model parameters in real time to adapt to operating condition drift. On the other hand, it combines coal quality soft measurement technology to achieve advance compensation for fuel calorific value fluctuations. This realizes a qualitative change in the control system from passive feedback to active adaptation and advance intervention, completely breaking the limitation of fixed parameter controllers that cannot adapt to unit aging and large changes in operating conditions. It gives the control system extremely strong robustness and generalization ability. In particular, for the most troublesome coal quality change problem of coal-fired units, it transforms it into a measurable variable through soft measurement and performs advance compensation. The fuel quantity is corrected before the pressure and load deviate substantially, which greatly reduces the system fluctuation amplitude and ensures that the unit always maintains the optimal operating state in complex and ever-changing external environments. Attached Figure Description

[0018] Fig. 1 This is a diagram illustrating the method steps of the present invention; Fig. 2 This is a diagram showing the system functional modules of the present invention; Fig. 3 This is a hardware structure diagram of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Specific Implementation Example 1

[0020] like Figs. 1-3 As shown, a cooperative control method suitable for thermal power units includes the following steps: Sp1: Real-time acquisition of operating status data of thermal power units; The acquisition of operating status data for the thermal power unit relies on the plant's distributed control system (DCS). The DCS acquires analog signals through sensors located at various physical nodes of the unit. The specific acquisition process is as follows: a pressure transmitter located on the main steam pipeline converts the main steam pressure physical signal into a current signal of 4 to 20 mA; a power transmitter located at the generator outlet acquires the actual generated power signal; a linear displacement sensor located at the turbine high-pressure cylinder inlet acquires the turbine control valve opening signal; a weighing feeder sensor located at the coal mill inlet acquires the real-time fuel quantity signal; a zirconia probe and thermocouple located at the boiler tail flue acquire the exhaust oxygen quantity and exhaust temperature signals, respectively; and an air volume measurement device located at the blower inlet acquires the total air volume signal entering the furnace. All the above analog signals are converted from analog to digital by the input / output modules of the DCS and transmitted to the central processing unit of the coordinating controller via a data bus. At the same time, the central processing unit receives automatic generation control load commands from the power grid dispatch center through hardwiring or a communication interface. The collected operating status data includes the actual power generation and main steam pressure as controlled variables, the main steam temperature as an auxiliary monitoring variable, the turbine valve opening and fuel quantity command as control feedback variables, and the grid automatic generation control load command for environmental disturbance monitoring. In addition, since the fluctuation of coal quality entering the boiler is the largest unmeasurable disturbance, the real-time calorific value of coal quality needs to be obtained through "backward calculation". In order to realize the soft measurement of coal calorific value, that is, the estimation data of the calorific value of coal entering the boiler, the data also includes the flue gas temperature signal, flue gas oxygen signal and total air volume signal entering the boiler. The method of obtaining this data is as follows: by using the real-time collected flue gas temperature signal, flue gas oxygen signal and total air volume signal entering the boiler, the real-time calorific value of coal quality is back-calculated using the heat balance principle. The calculated calorific value of coal quality is then input as a measurable disturbance variable into the nonlinear model predictive control algorithm of Sp4 to correct the gain of the fuel quantity control command in real time. Before entering the control algorithm, the collected raw data needs to be processed by first-order inertial filtering to remove high-frequency noise interference. For signals with different sampling frequencies, timestamp alignment is required to ensure that all input data are synchronized in time and form the system state vector at the current moment.

[0021] Sp2: Construct a boiler-machine coupled nonlinear state-space model to describe the relationship between boiler energy storage and turbine energy conversion, and design a dynamic thermal storage state observer based on the model. Utilize the collected operating status data to calculate the thermal storage state index, which characterizes the unit's transient energy throughput capacity, in real time. Construction and Contents of the Coupling Nonlinear State-Space Model of Boiler and Turbine: This model aims to mathematically reproduce the physical processes of boiler energy storage and turbine energy conversion. It is a nonlinear state-space model, and its core components include state variables, input variables, and state transition equations. Model state variables: Main steam pressure is selected to represent the energy storage state on the boiler side, and actual power generation is selected to represent the energy conversion state on the turbine side; Model input variables: The turbine control valve opening is selected as the fast control variable for regulating power, and the fuel quantity command is selected as the slow control variable for regulating pressure. Model Content and Explanation: The model consists of two core differential equations. The first differential equation describes the rate of change of the main steam pressure, which is equal to the heat released by fuel combustion minus the steam energy consumed by the turbine. The energy consumed by the turbine is nonlinearly related to the product of the main steam pressure and the turbine valve opening. This nonlinear relationship conforms to the Bernoulli equation in fluid mechanics. The second differential equation describes the rate of change of the actual generated power, which depends on the rotational inertia of the turbine rotor and the enthalpy drop of the working steam. Specifically, the actual generated power has a first-order inertial hysteresis characteristic in response to the main steam pressure and valve opening. In addition, the model also includes a time-varying parameter of coal calorific value to correct the energy conversion efficiency of fuel quantity. Furthermore, the parameter update method of the coupled nonlinear state-space model of the boiler and turbine is as follows: an online least squares method with a forgetting factor is used to identify and update the static gain parameter and inertial time constant parameter in the model in real time using the input and output data within the most recent time window to adapt to the dynamic characteristics of the unit under different load conditions.

[0022] Design of a dynamic thermal storage state observer: Since sensors alone cannot directly measure the energy surplus or deficit within the unit, a dynamic thermal storage state observer is designed. This observer uses an extended Kalman filter algorithm, and the design process is as follows: First, a prediction step is established based on the above nonlinear model. The pressure and power at the current moment are predicted using the state estimate value and control quantity from the previous moment. Second, a correction step is established. The deviation between the actual main steam pressure and actual generated power at the current moment and the predicted value is used to correct the predicted state through the Kalman gain matrix. The core function of the observer is to output the optimal estimate value of the system state after noise stripping in real time, and at the same time estimate the unmeasurable coal calorific value disturbance.

[0023] The calculation and physical meaning of the thermal storage state index: The thermal storage state index is a scalar indicator that quantitatively characterizes the energy throughput capacity of the unit at the current moment. Its calculation utilizes data collected and corrected by the observer. The calculation method is as follows: First, the main steam pressure signal is differentiated to obtain the rate of pressure change. This rate is then multiplied by the preset boiler thermal storage coefficient to obtain the first component, which characterizes the release or absorption of stored energy by the boiler metal and working fluid. Second, the calorific value of the coal obtained by soft sensing is multiplied by the current fuel quantity to calculate the total heat input from fuel combustion. This total heat is then subtracted from the actual power output of the generator, and the difference is multiplied by the energy conversion gain coefficient to obtain the second component, which characterizes the degree of imbalance between input and output energy. Finally, the first and second components are weighted and summed to obtain the thermal storage state index. When the thermal storage state index is positive and the value is large, it indicates that the unit has sufficient energy reserves or is accumulating energy. When the index is negative and the absolute value is large, it indicates that the unit is overdrawing its stored energy to maintain the load, and the boiler energy is close to being exhausted.

[0024] Sp3: Establish a multi-objective collaborative optimization objective function that includes load tracking error, main steam pressure deviation, and control quantity change rate, and introduce a dynamic weighting factor based on the thermal storage state index for real-time adjustment. The dynamic weighting factor is used to balance the priority between load response speed and pressure stability. Establishment and content of the objective function for multi-objective collaborative optimization: This objective function is a mathematical expression used to define the criteria for evaluating the performance of the control system. The specific content of the function consists of a weighted sum of squares of three parts: The first part is the load tracking error term: that is, the square of the difference between the predicted actual power generation and the grid load command, which is designed to ensure that the power generation meets the dispatch requirements. The second part is the main steam pressure deviation item: that is, the square of the difference between the predicted main steam pressure and the set pressure, which is designed to ensure the safe operation of the unit. The third part is the control variable change rate constraint: that is, the square of the change in the turbine control valve and fuel quantity at adjacent times, which is designed to prevent excessive actuator movement that could lead to equipment wear or system oscillation.

[0025] The dynamic weighting factor is a numerical coefficient that changes in real time. It is directly multiplied before the main steam pressure deviation term in the objective function. The purpose of introducing this factor is to change the focus of the optimization problem at the mathematical level. This factor is adjusted based on the heat storage state index calculated by Sp2. Speed-priority logic: When the absolute value of the thermal storage state index is less than the preset safety threshold, it indicates that the unit's energy reserves are in a safe range. At this time, the algorithm sets the dynamic weighting factor to a small value, which reduces the weight of the pressure deviation term in the objective function. The optimization solver will be more inclined to reduce the load tracking error when searching for the optimal solution, even if it causes a certain degree of fluctuation in the main steam pressure by a large operation of the turbine control valve, thereby achieving a fast load response speed. Stability-first logic: When the absolute value of the thermal storage state index is greater than or equal to the safety threshold, it indicates that the unit's energy has been severely overdrawn or overpressured. At this time, the algorithm rapidly increases the dynamic weighting factor according to the nonlinear trend, i.e., the exponential function law. This makes the weight of the pressure deviation term in the objective function dominate. In order to reduce the pressure deviation, the optimization solver will forcibly sacrifice the load tracking accuracy, i.e., limit the turbine control valve operation and significantly adjust the fuel quantity, thus establishing the highest priority of pressure stability.

[0026] Sp4: A nonlinear model predictive control algorithm is used to solve the multi-objective collaborative optimization objective function in the rolling optimization time domain, and to calculate the optimal turbine control valve increment and fuel quantity control increment at the current moment. Nonlinear model predictive control (MMC) is an advanced control strategy that predicts future dynamics based on a model and optimizes current control actions. The rolling optimization time domain is defined as a time window containing several future sampling moments, such as 30 to 60 seconds in the future, called the prediction time domain. At the same time, a shorter control time domain is defined, such as 3 to 5 seconds in the future. The algorithm logic is as follows: at each sampling moment, the algorithm starts from the current system state and simulates the future trajectory of the system within the prediction time domain. The solution process of the multi-objective collaborative optimization function: In order to obtain the optimal control increment, it is necessary to numerically solve the nonlinear optimization problem constructed above. Considering the real-time requirements, the sequential quadratic programming method is adopted. The specific steps are as follows: First, near the current operating point, the nonlinear model in Sp2 is locally linearized using Taylor series expansion. Second, the multi-objective objective function in Sp3 is transformed into a standard quadratic programming problem. Then, the quadratic programming problem is solved using the effective set method or interior point method to find a set of optimal control increment sequences in the control time domain that minimizes the value of the objective function. The output of the solution process is a set of future control increment sequences. However, according to the principle of model predictive control, only the first element in the sequence is selected, namely the optimal turbine valve control increment and fuel quantity control increment at the current moment. Because the thermal power unit is a large delay and strongly coupled system, the traditional PID controller cannot predict the impact of the control action on the future. By calculating this optimized increment, the controller actually makes the best decision at the current moment after weighing the energy changes, safety constraints and load demands in the next few tens of seconds. This ensures both fast response and avoids parameter over-limits in the future. The nonlinear model predictive control algorithm also incorporates a feedforward compensation mechanism based on the load command change rate: the load change rate is obtained by calculating the derivative of the automatic generation control load command of the power grid with respect to time. When the absolute value of the load change rate exceeds the preset dead zone, the inertial differential feedforward quantity is calculated according to the magnitude of the load change rate. The inertial differential feedforward quantity is directly superimposed on the fuel quantity control increment obtained by the Sp4 optimization solution to compensate for the inertial delay of the boiler combustion system in advance.

[0027] Sp5: The calculated optimal control increment is superimposed on the current control command and sent to the turbine actuator and the boiler combustion actuator respectively to complete the coordinated control of the thermal power unit.

[0028] Before sending control commands, there is also an actuator constraint correction step: based on the current number of coal mills in operation and the maximum output limit of a single coal mill, it is determined whether the fuel quantity command calculated by Sp4 exceeds the limit. If it exceeds the limit, discrete variables of coal mill start-up and shutdown operation are introduced in the rolling optimization time domain to predict the fuel quantity disturbance during the coal mill switching process, and the turbine control valve increment is corrected in reverse to smooth out the expected pressure fluctuation. The calculated optimal control increment is a relative change value. In order to generate the final execution command, this increment needs to be superimposed on the actual control command of the previous moment. For example, the current turbine valve command is equal to the valve command of the previous moment plus the calculated valve control increment; the current fuel quantity command is equal to the fuel quantity command of the previous moment plus the calculated fuel quantity control increment, and the feedforward compensation mentioned in Sp5 is superimposed on this basis. The generated final control command is converted into a physical signal through the output module of the distributed control system. The turbine valve command is sent to the servo card of the turbine digital electro-hydraulic control system to drive the hydraulic motor to change the valve opening. The fuel quantity command is sent to the coal feeder frequency converter to change the coal feeder speed to adjust the amount of coal fed into the furnace.

[0029] This process completes the coordinated control, forming a complete closed loop: Sensing: Sp1 and Sp2 sense the system's real-time energy state and external demands; Decision: Based on the prediction of the future and the trade-off between the current energy state, Sp3 and Sp4 calculated the optimal control strategy that balances speed and safety, and resolved the supply and demand contradiction between the steam turbine and the boiler in the dynamic process. Execution: SP5 translates the strategy into physical actions, which are then applied to the aircraft. As the next sampling moment arrives, the entire process Sp1 to Sp5 will be repeated to continuously correct control deviations, thereby achieving deep collaborative optimization of the thermal power unit under all operating conditions. Specific Implementation Example 2

[0030] like Figs. 1-3 As shown, based on the content of the above specific embodiments, the following content is further disclosed: To more intuitively demonstrate how this collaborative control works in a real industrial setting, a specific example is provided below. This example describes the entire process of a 660 MW supercritical coal-fired unit responding to a large and rapid increase in grid load under deep peak shaving conditions. This process fully embodies the collaborative operation logic of Sp1 to Sp5. Initial operating scenario: The unit is currently in the deep peak shaving phase at 40% of rated load, with the actual power generation stable at 264 MW. At this time, the power grid dispatch center suddenly issues an instruction requiring the unit to increase the load to 50% of rated load, i.e. 330 MW, within 5 minutes. This is a large-scale load change condition. Traditional control methods are very likely to cause a "avalanche" drop in main steam pressure due to the steam valve being opened too quickly, thereby triggering the trip protection. Phase 1: Command mutation and feedforward compensation intervention (T=0 seconds to 10 seconds); Sp1 execution: The control system collects the real-time data of the grid automatic power generation control load command changing from 264 MW to 330 MW. At the same time, the system uses the soft measurement module to combine the current flue gas temperature and oxygen content to estimate that the calorific value of the current coal entering the furnace is slightly lower than the design value, which is low-quality coal. Sp4 feedforward action: Before the nonlinear model predictive control calculation, the logic judgment module first detects that the rate of change of the load command is extremely large. Based on the feedforward compensation mechanism in Sp4, the controller immediately calculates a huge inertial differential feedforward fuel amount according to the rate of change of the load. Sp5 execution: This feedforward quantity is directly superimposed on the current fuel command without optimization iteration, and the coal feeder speed increases instantly. The purpose of this action is to send a large amount of chemical energy into the furnace in advance during the 60 to 100-second delay before the boiler reacts, so as to prepare for subsequent energy demand. Phase 2: Thermal storage and rapid response (T=10 seconds to 60 seconds); Sp2 execution: The boiler-turbine coupled nonlinear model starts working. The state observer monitors the main steam pressure in real time and finds it to be near the rated value with a rate of change of zero. According to the calculation formula, the thermal storage state index is positive and at a high level at this time, indicating that the energy stored inside the unit is sufficient. Sp3 execution: Since the absolute value of the thermal storage state index is less than the set safety threshold, the dynamic weighting factor adjustment mechanism determines that "speed priority" is currently given, and the algorithm automatically reduces the weight of the main steam pressure deviation term in the objective function; Sp4 solution: The optimization solver performs calculations in the rolling time domain and concludes that in order to minimize the load error, the best strategy is to open the turbine control valve significantly, because the pressure weight is low at this time, and the controller considers sacrificing some pressure to be acceptable. Actual performance: The turbine control valve opens rapidly, utilizing the steam energy stored in the boiler metal and pipes, the actual power output increases rapidly within tens of seconds, closely following the grid command. At this time, the main steam pressure begins to show a downward trend. Phase 3: Energy Overdraft and Strategy Reversal (T = 60 to 150 seconds); Sp2 execution: As the regulating valve continues to open, a large amount of boiler energy is released. The dynamic heat storage status observer detects that the main steam pressure drops at an accelerated rate, and the heat released by fuel combustion has not yet been fully converted into steam energy (combustion lag). The calculated heat storage status index quickly turns from positive to negative, and the absolute value continues to increase, eventually exceeding the preset heat storage safety threshold. This indicates that the unit has entered the dangerous area of ​​"energy overdraft". Sp3 Execution: The system immediately triggers the adjustment logic of the dynamic weighting factor, which increases exponentially, meaning that the priority of "pressure stability" outweighs "load response speed". The shape of the objective function changes, and the main steam pressure deviation term becomes the most severely penalized part. Sp4 Solution: Under the guidance of the new objective function, the nonlinear model predictive control algorithm is re-solved. In order to avoid further deterioration of pressure leading to shutdown, the optimization results give new instructions: stop opening the turbine control valves significantly, or even slightly close the control valves to suppress pressure, while continuing to maintain a high fuel input. Actual performance: On-site observation showed that the turbine control valve movement slowed down significantly, and the rate of increase in actual power generation was temporarily limited. At this time, the load may lag behind the command for a short period of time, but this strategy successfully curbed the vicious decline in main steam pressure, prevented pressure over-limit accidents, and ensured the online safety of the unit. Phase 4: Start-up and shutdown of the coal mill and constraint correction (T=150 seconds to 300 seconds); Sp5 Execution (Constraint Correction): As the target load approaches, the calculated total fuel quantity command continues to increase and is about to exceed the sum of the maximum output of the three currently operating coal mills, triggering the constraint correction mechanism in Sp5. Logical action: The system automatically issues a command to start the fourth standby coal mill. At the same time, the algorithm predicts that cold air will enter the furnace during the initial startup of the new coal mill, causing a brief fluctuation in steam pressure. Therefore, in the rolling optimization time domain, the controller reverses the turbine regulating valve command and fine-tunes the valve in advance to smooth out the impending pressure fluctuation. Through the aforementioned coordinated control, the 660 MW unit completed the load ramp-up process from 264 MW to 330 MW within 5 minutes, compared to traditional PID control: Load response speed: Overall settling time was reduced by approximately 20% because the “aggressive” strategy was fully utilized during the sufficient heat storage phase; Pressure stability: The lowest point of the main steam pressure is 0.5 MPa higher than that of traditional control, and it has never reached the trip protection value because it switched to a "conservative" strategy in time during the heat storage overdraft stage. Synergy: It truly realizes the dynamic decoupling and matching of energy supply and demand between the steam turbine side (fast) and the boiler side (slow). Specific Implementation Example 3

[0031] like Figs. 1-3 As shown, based on the content of the above specific embodiments, the following content is further disclosed: This method is applicable to the coordinated control system of thermal power units. Its internal logic structure mainly consists of four core modules: data acquisition and preprocessing module, boiler and turbine state observation and identification module, nonlinear predictive coordinated control module, and actuator constraint and output module. Data Acquisition and Preprocessing Module: This module is the system's sensing front end, primarily responsible for data acquisition and cleaning. It periodically reads real-time data from the distributed control system via a communication interface, including unit load, main steam pressure, temperature, valve opening, fuel quantity feedback, and flue gas parameters on the exhaust side. After data acquisition, this module integrates a signal filtering unit and a timing synchronization unit. The signal filtering unit uses a first-order low-pass filter algorithm to remove high-frequency noise generated by electromagnetic interference. The timing synchronization unit timestamps signals from different sampling periods, ensuring that the data input to subsequent algorithm models belongs to the same physical moment, thus avoiding control model mismatch caused by data delays. Furthermore, this module also embeds a coal quality soft measurement subunit, which uses the heat balance principle to calculate coal calorific value data in real time and transmits it to the next-level module.

[0032] Boiler and turbine condition observation and identification module: This module is one of the core computational components of the system, responsible for reconstructing the unmeasurable states within the system. Internally, this module runs a program based on the extended Kalman filter algorithm, receives cleaned data from the data acquisition module, and utilizes a built-in boiler-turbine coupled nonlinear state-space model to calculate intermediate state variables that cannot be directly measured in real time, including the effective heat storage inside the boiler and the dynamic energy of the turbine rotor. Simultaneously, this module includes a heat storage state index calculation unit, which calculates the heat storage state index based on the differential of the main steam pressure and the input-output energy difference, and uses this index as the basis for subsequent control strategy switching. Furthermore, this module also has an online parameter identification function, using the least squares method with a forgetting factor to correct the static gain and time constant in the model in real time, in order to address model errors caused by unit equipment aging or operating condition drift.

[0033] Nonlinear Predictive Cooperative Control Module: This module is the system's decision-making brain, responsible for generating optimal control commands. It integrates a multi-objective cooperative optimization objective function generator, a dynamic weight adjuster, and a sequential quadratic programming solver. The dynamic weight adjuster adjusts the weight coefficients of the pressure deviation and load tracking terms in the objective function in real time based on the thermal storage state index input from the previous module, achieving automatic and smooth switching between speed-priority and stability-priority modes. The solver performs optimization calculations in the rolling optimization time domain within each control cycle based on the nonlinear model and the adjusted objective function, deriving the optimal turbine control valve increment and fuel quantity control increment for the current moment. Feedforward Compensation Unit: This module also independently detects the rate of change of the grid load command. When the rate of change exceeds the dead zone, it directly generates an inertial differential feedforward signal and superimposes it onto the fuel quantity calculation channel to overcome the lag of pure feedback control.

[0034] Actuator Constraint and Output Module: This module serves as the system's safety barrier and execution endpoint. It primarily verifies the safety of calculated control commands and stores the physical constraint parameters of each actuator in the unit, such as the maximum output of the coal mill and the valve operation rate limit. When the command output by the predictive control module exceeds these physical boundaries, this module performs amplitude limiting. Specifically, this module includes discrete event processing logic for the coal mill. When a fuel command triggers the coal mill start-up / stop conditions, it automatically generates a corresponding warning signal and corrects the valve command. The final command, after successful verification, is converted into a standard communication protocol signal by this module and sent to the underlying actuator.

[0035] To support the real-time operation of the aforementioned complex algorithms, this system adopts a layered distributed hardware architecture, mainly consisting of a field device layer, a basic control layer, and a high-level computing layer: On-site equipment layer: This layer consists of physical equipment directly installed on the thermal power unit's production process. Sensor components include a high-precision pressure transmitter installed on the main steam pipeline, a power transmitter installed at the generator outlet, a zirconia analyzer and thermocouples installed in the flue, etc. These sensors are responsible for converting physical quantities into analog current signals of 4 to 20 mA. Actuator components: including the electro-hydraulic servo valve of the steam turbine electro-hydraulic control system, the coal feeder variable frequency motor, the blower moving blade adjustment mechanism, etc., are responsible for receiving electrical signals and converting them into mechanical actions such as valve opening degree or motor speed.

[0036] Basic control layer: This layer typically consists of the power plant's existing distributed control system. Input / output module: responsible for receiving analog signals from field sensors and performing analog-to-digital conversion, while converting digital control signals into analog signals to drive actuators; Distributed processing unit (DCS controller): responsible for executing routine basic logic protection, such as trip protection and equipment start-stop interlocking. In this system, the basic control layer serves as the data aggregation point and the final instruction forwarding station, ensuring that even if the upper-level high-level algorithm fails, the underlying basic logic can still maintain the safe operation of the unit.

[0037] Advanced computing layer: This is the core hardware carrier of the collaborative control system, typically consisting of one or more industrial-grade high-performance servers or embedded controllers. Advanced process control station: Since nonlinear model predictive control involves a large number of matrix operations and iterative solutions, the computing power of conventional DCS controllers cannot meet the millisecond-level real-time requirements. Therefore, an independent advanced process control station is configured, which is equipped with a high-frequency multi-core central processing unit and no less than 16GB of running memory. Communication Gateway: The Advanced Process Control Station establishes a bidirectional connection with the Distributed Control System of the Basic Control Layer through redundant Industrial Ethernet or OPC communication protocols. It reads real-time data from the entire plant from the DCS, completes all complex calculations from Sp2 to Sp4 locally, and writes the calculated control increments back to the control memory area of ​​the DCS. Finally, the DCS drives the field devices.

[0038] Network communication layer: Redundant data bus: a data highway connecting the basic control layer and the advanced computing layer. It uses twisted-pair or fiber optic media and is configured with a dual-network redundancy architecture. When the main communication line fails, the hardware circuit can automatically switch to the backup line without disturbance, ensuring the continuity and reliability of control command transmission.

[0039] Through the logical coordination of the aforementioned software modules and the layered deployment of hardware devices, this system constructs a closed-loop, highly reliable intelligent control environment for thermal power units with deep learning and adaptive capabilities. Specific Implementation Example 4

[0040] like Figs. 1-3 As shown, based on the content of the above specific embodiments, the following content is further disclosed: To further verify the feasibility of the technical solution in this application, the following case study is provided: Case 1: Rapid ramp response under deep peak shaving conditions; A 600 MW supercritical coal-fired power unit was operating during a low-load period in the power grid. The unit was maintaining a load of 40% of its rated capacity, or 240 MW, for deep peak shaving. At this time, the power grid dispatch center issued an emergency load increase order to the unit due to a sudden drop in wind power output. The order required the load to be increased to 300 MW within 2 minutes, which is a change rate of 5% per minute. This is a huge challenge for coal-fired power units with large delays. Traditional control methods are very likely to cause the main steam pressure to drop instantly to the trip protection value. Sp1 (Sensing): The system collects in real time the automatic power generation control command jumps from 240 MW to 300 MW. At the same time, it collects the current main steam pressure as 10 MPa and the valve opening as 35%. Sp2 (Observation): The dynamic thermal storage state observer calculates the thermal storage state index at the initial moment based on the current stable pressure differential and power data. If the index is positive, it indicates that the boiler is currently in a state of energy storage equilibrium. Sp3 (Decision Weight): Since the initial thermal storage state index is within the safe threshold, the system determines that it is currently capable of utilizing thermal storage. The dynamic weighted factor regulator reduces the weight of the "main steam pressure deviation term" in the objective function and increases the weight of the "load tracking error term", thus establishing a speed-priority control strategy. Sp4 (Optimization Solution): The nonlinear model predictive control algorithm is calculated in the rolling time domain. In order to eliminate the 60 MW load deviation as quickly as possible, the algorithm calculates the control increment that significantly opens the turbine control valve. At the same time, the feedforward compensation mechanism detects that the load command change rate is extremely large and superimposes a large fuel feedforward increment. Sp5 (Execution and Dynamic Correction): First 30 seconds: The turbine control valve opens rapidly, and the actual power output surges from 240 MW to 280 MW. At this time, the main steam pressure begins to drop rapidly. From the 30th to the 60th second: The observer in Sp2 detects that the pressure drop rate is too fast, and the calculated thermal storage state index quickly turns from positive to negative, exceeding the set thermal storage safety threshold. Strategy reversal: Sp3 responds immediately, increasing the weight of the "pressure deviation term" exponentially in a nonlinear manner. Sp4 senses this weight change in the next calculation cycle. In order to avoid the objective function value being too large, the optimization result forcibly stops the valve from opening further and even issues a command to slightly close the valve, while continuing to maintain a high fuel quantity command. Result: The unit's actual power output responded quickly to the grid's demand in the early stages, and in the later stages, the unit successfully curbed the severe pressure decline by limiting the regulating valves, keeping the minimum pressure above the safety line and successfully completing the load-grabbing task.

[0041] Case Study 2: Adaptive Stability Control under Disturbance of Low-Quality Coal; The unit was operating stably at a 500 MW load when the type of coal in the raw coal bunker suddenly changed from high-calorific-value bituminous coal to low-calorific-value lignite. This sudden drop in calorific value is usually unpredictable. Traditional control often only increases fuel through PID feedback after the pressure drops significantly, resulting in severe regulation lag. Sp1 (Sensing and Soft Measurement): The fuel quantity command remains unchanged, but the flue gas oxygen quantity signal begins to rise, the flue gas temperature drops slightly, and the actual power output shows a slight decline. The coal quality soft measurement module embedded in the system uses the principle of heat balance to reverse-calculate that the calorific value of the coal entering the furnace is decreasing from 22 MJ / kg to 18 MJ / kg. Sp2 (Model Update): The online parameter identification algorithm described in Sp6 captures changes in the ratio between input fuel energy and output power, updates the fuel gain parameters in the nonlinear state-space model in real time, and reduces the model's expectation of heat production per unit of fuel. Sp3 (Decision Weights): Due to the reduction in input energy, the thermal state index calculated by the observer begins to shift negatively. Although it has not yet reached the danger threshold, the dynamic weighting factor has begun to be fine-tuned, and the weight of pressure stability has been moderately increased. Sp4 (Optimization Solution): The nonlinear model predictive control algorithm uses the updated model for prediction. The algorithm finds that if the current fuel quantity is maintained, the predicted future main steam pressure will continue to decrease. Therefore, before the pressure decreases significantly, the optimizer calculates a positive fuel quantity control increment, and the increment is larger than that under normal operating conditions to compensate for the gap caused by the decrease in calorific value. SP5 (Execution): The coal feeder speed is increased significantly ahead of schedule; Result: Before the main steam pressure had a chance to drop significantly, a large amount of supplementary fuel had already been sent into the furnace. The actual operating curve showed that the main steam pressure only fluctuated by 0.2 MPa before returning to normal, achieving "imperceptible" suppression of coal quality disturbance.

[0042] Case 3: Coal mill switching constraint control under full-load impact; The unit is currently operating at a load of 550 MW, with 4 coal mills in operation. The output of a single coal mill has reached 90%. The power grid requires the load to be increased to the full load of 660 MW, which will inevitably trigger the start-up of the 5th standby coal mill. The start-up of the coal mill will blow a large amount of cold air into the furnace, causing a sudden drop in steam pressure. Subsequently, the combustion of pulverized coal will cause a sudden rise in steam pressure, which can easily cause system oscillation. Sp1 (Sensing): Received a load increase command; Sp4 (Prediction and Constraints): The control algorithm calculates the required fuel quantity and finds that the total required fuel exceeds the maximum output limit of the current four coal mills; Sp5 (Constraint Correction and Event Triggering): Decision: The system triggers the discrete event flow of "starting the E-mill"; Disturbance Prediction: Within the rolling optimization time domain of Sp4, the model introduces a preset disturbance sequence: the pressure tends to decrease from the 10th to the 60th second in the future (cold air enters), and the pressure tends to increase after the 120th second in the future (new coal combustion); Reverse Correction: Based on the above predictions, the nonlinear model predictive control algorithm calculates a special control sequence. At the same time as the mill start command is issued, the calculated turbine valve control increment is negative (i.e., slightly closing the valve). Execution process: The E coal mill starts up, and a large amount of cold primary air rushes into the furnace. Almost at the same time, the turbine control valve actively closes by 2%. This action increases the pressure in front of the turbine, which just offsets the pressure drop caused by the cold air. Two minutes later, the coal powder in the new coal mill begins to burn violently, and the pressure tends to rise. At this time, the control valve command calculated by the algorithm turns to open in the positive direction, which not only releases the excess pressure, but also just meets the load demand of 660 MW. Results: Throughout the load increase and start-up process, the main steam pressure curve was smooth, without the large fluctuations of "low first and then high" common under conventional control, and the unit smoothly transitioned to full-load operation.

[0043] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0044] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated control of a cogeneration unit, characterized by , comprising the following steps: Sp1: collecting real-time operation state data of the thermal power unit; Sp2: constructing a boiler-turbine coupling nonlinear state space model describing the relationship between energy storage of the boiler and energy conversion of the turbine, and designing a dynamic heat storage state observer based on the model, and using the collected operation state data to calculate a heat storage state index representing the instantaneous energy throughput capacity of the unit in real time; Sp3: establishing a multi-objective collaborative optimization objective function including load tracking error, main steam pressure deviation, and control variable change rate, and introducing a dynamic weighting factor based on the heat storage state index for real-time adjustment, the dynamic weighting factor being used to balance the priority between load response speed and pressure stability; Sp4: using a nonlinear model predictive control algorithm to solve the multi-objective collaborative optimization objective function in a rolling optimization time domain, and calculating the optimal turbine governing valve control increment and fuel quantity control increment at the current time; Sp5: superimposing the calculated optimal control increments on the current control instructions and sending them to the turbine actuator and the boiler combustion actuator respectively to complete the collaborative control of the thermal power unit.

2. A method for coordinated control of a thermoelectric unit according to claim 1, characterized in that: The operation state data in Sp1 includes at least real power output, main steam pressure, main steam temperature, turbine governing valve opening, fuel quantity instruction, and grid automatic generation control load instruction, and also includes estimated data of coal calorific value entering the boiler, which is obtained by inversely calculating the real-time coal calorific value using the heat balance principle based on the real-time collected flue gas temperature signal, flue gas oxygen content signal, and total air flow entering the boiler, and inputting the calculated coal calorific value as a measurable disturbance variable into the nonlinear model predictive control algorithm in Sp4 to real-time correct the gain of the fuel quantity control instruction.

3. A method for coordinated control of a thermoelectric unit according to claim 1, characterized in that: The calculation method of the heat storage state index in Sp2 is as follows: multiplying the differential value of the main steam pressure by a preset boiler heat storage coefficient to obtain a first component, multiplying the difference between the estimated value of the current fuel release heat and the real power output by an energy conversion gain coefficient to obtain a second component, and the heat storage state index being a weighted sum of the first component and the second component, and being used to quantitatively represent the degree of energy surplus or deficiency in the boiler metal and working medium that can be used for rapid release or needs to be supplemented at the current time.

4. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: The adjustment mechanism of the dynamic weighting factor in Sp3 is as follows: setting a heat storage safety threshold, when the absolute value of the real-time calculated heat storage state index is less than the heat storage safety threshold, reducing the weight of the main steam pressure deviation term in the dynamic weighting factor to make the controller preferentially use the unit heat storage to quickly respond to the load instruction, and when the absolute value of the real-time calculated heat storage state index is greater than or equal to the heat storage safety threshold, nonlinearly increasing the weight of the main steam pressure deviation term in the dynamic weighting factor to forcibly suppress the turbine governing valve action amplitude and preferentially adjust the fuel quantity to restore the main steam pressure.

5. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: The nonlinear model predictive control algorithm in the Sp4 also incorporates a feedforward compensation mechanism based on the rate of change of load command: the derivative of the grid automatic generation control load command with respect to time is calculated to obtain the load change rate, and when the absolute value of the load change rate exceeds a preset dead zone, an inertial derivative feedforward quantity is calculated according to the size of the load change rate, and the inertial derivative feedforward quantity is directly superimposed on the fuel quantity control increment obtained by solving the Sp4 optimization to compensate for the inertial delay of the boiler combustion system in advance.

6. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: The parameter updating method of the machine-furnace coupled nonlinear state space model in the Sp2 is as follows: an online least square method with a forgetting factor is used to identify and update the static gain parameters and inertial time constant parameters in the model in real time by using the input and output data in a recent time window, so as to adapt to the dynamic characteristic changes of the unit under different load conditions.

7. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: In the Sp5, a mechanism constraint correction link is further included before the control command is sent: according to the current number of operating coal mills and the maximum output limit of a single coal mill, it is judged whether the fuel quantity command calculated by the Sp4 is out of limit, if it is out of limit, the discrete variables of the start-stop operation of the coal mill are introduced in the rolling optimization time domain, the fuel quantity disturbance in the coal mill switching process is predicted, and the turbine governing valve control increment is corrected in reverse to suppress the expected pressure fluctuation.

8. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: The collaborative control method has two operation modes according to the grid demand configuration: the first mode is a pressure mode dominant mode, the dynamic weighting factor is locked to a larger value, so that the control system preferentially maintains the main steam pressure around the rated value under any operating condition, and is suitable for the grid load stable period; the second mode is a source mode dominant mode, the real-time adjustment function of the dynamic weighting factor is activated, the main steam pressure is allowed to deviate from the set value within a safe range, the variable load rate of the regenerative unit is utilized, and is suitable for the period when the grid frequency modulation demand is high.

9. A method for coordinated control of a thermoelectric generating unit as claimed in claim 1, wherein: In the Sp4, the solving process in the rolling optimization time domain adopts a sequential quadratic programming method: the nonlinear optimization problem is linearized at each sampling time, converted into a quadratic programming sub-problem for solving, and the optimal solution at the previous sampling time is taken as the initial value of iteration at the current time, so as to meet the calculation speed requirement of real-time control.

Images