A model prediction-based coordinated optimization control method for thermal power generating units

Abstract

The application discloses a model prediction-based coordinated optimization control method for a thermal power unit, and particularly relates to the field of coordinated optimization control of thermal power units, and comprises the following steps: S1, collecting real-time operation parameters of the thermal power unit through a sensor network, and establishing a unit operation state database after preprocessing the real-time operation parameters; S2, constructing a boiler-turbine dynamic prediction model based on the operation state database, adopting a state space equation to describe, predicting a parameter trend in the future 60s and verifying the accuracy; and S3, constructing an optimization objective function containing a multi-parameter deviation and an energy consumption item according to the prediction trend, and solving an optimal control amount every 10s, and the control amount is constrained by a rated parameter range; the application can realize multivariable collaborative control by constructing a boiler-turbine dynamic prediction model and combining a rolling optimization algorithm, and can predict a parameter change trend in advance, thereby reducing a parameter fluctuation range under a variable load condition and improving the operation stability of the unit.

Description

Technical Field

[0001] This invention relates to the field of coordinated optimization control technology for thermal power units, and more specifically, to a model-predictive-based coordinated optimization control method for thermal power units. Background Technology

[0002] During the operation of thermal power units, coordinated control is a core technology to ensure the safe, stable operation and economic performance of the units. It mainly achieves energy balance between the turbine and boiler by adjusting key parameters such as fuel supply, feedwater flow, and air volume to respond to grid load commands. Currently widely used coordinated control methods mainly include PID control, fuzzy control, and adaptive control. These methods can basically meet the unit's operating requirements under stable conditions.

[0003] Existing technologies can effectively maintain the stability of key parameters such as main steam pressure and drum water level during stable unit operation, and possess a certain load tracking capability. However, in actual operation, thermal power units often face complex and variable operating conditions, such as rapid load fluctuations, changes in coal composition, and equipment aging. Traditional control methods, relying on fixed control parameters and empirical models, struggle to adapt quickly to changes in operating conditions, leading to problems such as main steam pressure overshoot and large fluctuations in drum water level during load changes. This not only affects the stability of unit operation but also increases coal consumption. Furthermore, traditional methods are insufficient in handling multivariate coupling relationships; when multiple parameters fluctuate simultaneously, control strategies are prone to conflict, resulting in deteriorated regulation performance.

[0004] To address the problems of lag in dynamic response, poor parameter coupling, and high energy consumption in existing technologies under complex operating conditions, this invention proposes a model-based predictive coordinated optimization control method for thermal power units. This method constructs a dynamic prediction model of the boiler and turbine, combines it with real-time operating parameters for rolling optimization and feedback correction, and achieves multi-variable coordinated control to improve the stability and economy of the unit under varying loads and complex operating conditions. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, the present invention provides a model prediction-based coordinated optimization control method for thermal power units, which solves the problems mentioned in the background art through the following scheme.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a model-predictive-based coordinated optimization control method for thermal power units, comprising:

[0007] S1. Collect real-time operating parameters of thermal power units through a sensor network, and establish a unit operating status database after preprocessing the real-time operating parameters;

[0008] S2. Construct a dynamic prediction model for the boiler and turbine based on the operating status database, use state-space equations to describe it, predict the parameter trend for the next 60 seconds and verify the accuracy.

[0009] S3. Based on the predicted trend, construct an optimization objective function that includes multiple parameter deviations and energy consumption terms. Optimal control quantity is solved by rolling optimization every 10 seconds, and the control quantity is constrained by the rated parameter range.

[0010] S4. Input the optimal control quantity into the control execution unit, adjust the opening of the fuel valve, feedwater valve and air supply valve through the mapping relationship, and collect the unit parameters after control in real time.

[0011] S5. Calculate the state deviation between the actual parameters and the predicted parameters, and use the feedback correction mechanism to correct the model state matrix parameters to improve the model's adaptability.

[0012] S6. Execute the rolling optimization steps in a loop to achieve continuous optimization of the coordinated control of the thermal power unit until the unit is running stably or receives a new load command.

[0013] Preferably, the real-time operating parameters include main steam pressure P, drum water level H, furnace temperature T, turbine speed N, fuel quantity F, feedwater quantity W, air supply quantity Q, and load command L.

[0014] Preferably, the preprocessing includes outlier removal and normalization; the outlier removal uses the 3σ criterion to identify outliers, where σ is the standard deviation of the parameter; the normalization process involves applying the formula to the parameter after outlier removal. Normalization is performed, where Represents the original parameters. Represents the normalized parameters. This indicates the historical minimum value of the parameter. This indicates the historical maximum value of the parameter.

[0015] Preferably, the dynamic prediction model for the furnace and turbine is described using state-space equations: ; ;in Represents the vector of the rate of change of state at time k+1; This represents the output vector at time k+1; The state vector at time k includes the rate of change of main steam pressure ΔP, the rate of change of steam drum water level ΔH, and the rate of change of furnace temperature ΔT. The control vector includes fuel quantity adjustment ΔF, water supply quantity adjustment ΔW, and air supply quantity adjustment ΔQ. The disturbance vector is represented by ΔL, which includes the change in load command. A, B, C, D, and E represent the model parameter matrix, which is identified through historical data.

[0016] Preferably, the prediction accuracy of the model is verified by the root mean square error (RMSE); the root mean square error (RMSE) is specifically expressed as... ,in Indicates the actual value. Let n represent the predicted value, and n represent the number of samples; and when <5% of models passed validation.

[0017] Preferably, the optimization objective function is specifically expressed as: ,in This indicates the optimization of the objective function value; This indicates the deviation between the main steam pressure at time t and the set value; This indicates the deviation between the steam drum water level and the set value at time t; This indicates the deviation between the furnace temperature at time t and the set value; This represents the deviation between the turbine speed at time t and the set value; This represents the fuel quantity adjustment at time t; This represents the water supply adjustment amount at time t; This represents the air supply volume adjustment at time t; Indicates the fuel energy consumption coefficient. This represents the energy consumption coefficient of the water supply. The energy consumption coefficient for air supply is obtained through unit heat balance calculations.

[0018] Preferably, the parameter range constraint is: during the rolling optimization process, the minimum value of the objective function is solved once every 10 seconds; ,in Indicates the rated fuel quantity; ,in Indicates the rated water supply; ,in This indicates the rated air supply volume.

[0019] Preferably, the mapping relationship is as follows: ,in Indicates the fuel valve opening. Indicates the initial fuel valve opening. Indicates the fuel valve adjustment coefficient; ,in Indicates the opening degree of the water supply valve. Indicates the initial opening degree of the water supply valve. This indicates the adjustment coefficient of the water supply valve; ,in This indicates the opening degree of the air supply valve. This indicates the initial air supply valve opening. Indicates the air supply valve adjustment coefficient; and satisfies .

[0020] Preferably, the state deviation ,in Indicates the actual state. The state matrix represents the predicted state. ,in Let R represent the correction coefficient matrix, and let R represent the original state matrix.

[0021] The technical effects and advantages of this invention are as follows:

[0022] 1. This invention constructs a dynamic prediction model for the boiler and turbine and combines it with a rolling optimization algorithm to predict parameter change trends in advance to achieve multi-variable collaborative control. This solves the problems of lag in dynamic response, overshoot of main steam pressure and large fluctuations in steam drum water level caused by traditional control methods under complex operating conditions. It achieves the benefits of reducing parameter fluctuation range and improving unit operation stability under variable load conditions.

[0023] 2. This invention integrates energy consumption parameters by optimizing the objective function, and continuously optimizes the control strategy by using a real-time feedback correction mechanism to adapt to complex working conditions such as coal quality changes and equipment aging. It solves the problems of insufficient handling of multi-variable coupling, high energy consumption and poor adaptability of traditional control methods, and achieves the benefits of reducing coal consumption and improving the economic efficiency of unit operation during load changes.

[0024] 3. By clearly defining the parameter mapping relationship between the control quantity and the actuator and designing the constraint conditions, and by identifying and optimizing the model parameters through historical data identification and real-time correction, this invention solves the problems of difficult parameter tuning and easy disconnect between control strategy and actual operation in traditional control methods. It achieves the benefits of ensuring that control commands can be directly applied to the actuator and enhancing the practicality and operability of the technical solution. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall structure of the present invention. Detailed Implementation

[0026] 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.

[0027] As attached Figure 1 The method for coordinated optimization control of thermal power units based on model prediction, as shown, includes:

[0028] S1. Collect real-time operating parameters of thermal power units through a sensor network, and establish a unit operating status database after preprocessing the real-time operating parameters;

[0029] Specifically, the real-time operating parameters include main steam pressure P, drum water level H, furnace temperature T, turbine speed N, fuel quantity F, feedwater quantity W, air supply quantity Q, and load command L; the preprocessing includes outlier removal and normalization; outlier removal uses the 3σ criterion to identify outliers, where σ is the parameter standard deviation; the normalization process involves applying the formula to the parameters after outlier removal. Normalization is performed, where Represents the original parameters. Represents the normalized parameters. This indicates the historical minimum value of the parameter. This represents the historical maximum value of the parameter. Thermal power units are dynamic systems with multiple coupled variables. Their operating state is determined by key parameters such as main steam pressure, drum water level, and furnace temperature. These parameters directly reflect the unit's energy balance and equipment status, serving as the core basis for coordinated control. Real-time parameter acquisition is a prerequisite for dynamic control; only based on real operating data can subsequent model building and optimization have practical significance. The selected parameters are all core indicators of unit operation: main steam pressure relates to the energy matching between the boiler and turbine; drum water level affects boiler safety; furnace temperature determines combustion efficiency; turbine speed reflects power generation load; fuel quantity, feedwater quantity, and air quantity are direct control inputs; and load commands are external disturbance sources, covering the entire chain of unit input state and output disturbances. The parameter selection is comprehensive and crucial. In the preprocessing stage, the 3σ criterion is a mature method for identifying outliers in statistics. It can effectively eliminate extreme data caused by sensor failure or transient interference and prevent outliers from disrupting the data distribution pattern. Normalization process maps parameters to the [0,1] interval through linear transformation, eliminating the influence of different dimensions such as pressure and water level, and ensuring fair weighting of each parameter in the subsequent model.

[0030] S2. Construct a dynamic prediction model for the boiler and turbine based on the operating status database, use state-space equations to describe it, predict the parameter trend for the next 60 seconds and verify the accuracy.

[0031] Specifically, it should be noted that the aforementioned dynamic prediction model for the furnace and turbine is described using state-space equations. ; ;in Represents the vector of the rate of change of state at time k+1; This represents the output vector at time k+1; The state vector at time k includes the rate of change of main steam pressure ΔP, the rate of change of steam drum water level ΔH, and the rate of change of furnace temperature ΔT. The control vector includes fuel quantity adjustment ΔF, water supply quantity adjustment ΔW, and air supply quantity adjustment ΔQ. The disturbance vector is represented by ΔL, which includes the change in load command. A, B, C, D, and E represent the model parameter matrix, identified through historical data. The model prediction accuracy is verified by the root mean square error (RMSE). The RMSE is specifically expressed as... ,in Indicates the actual value. Let n represent the predicted value, and n represent the number of samples; and when <5% of the models passed validation. From a rationale perspective, thermal power units exhibit characteristics of high inertia, large lag, and strong coupling. For example, changes in fuel quantity require combustion and heat transfer processes to affect the main steam pressure; relying solely on real-time feedback control would lead to regulation lag. Constructing a predictive model and forecasting parameter trends 60 seconds in advance not only allows for adjustment time for optimized control but also avoids error accumulation due to excessively long prediction times, as 60 seconds matches the typical response cycle of the unit, thus balancing prediction accuracy and predictive capability.

[0032] The use of state-space equation modeling conforms to the description logic of complex dynamic systems: the state vector includes the rate of change of main steam pressure, steam drum water level, and furnace temperature, directly reflecting the dynamic trend of the core state of the unit; the control vector covers the regulation of fuel, feedwater, and air supply, corresponding to the actual operable control inputs; the disturbance vector focuses on load command changes, accurately capturing the most important external disturbances. The model structure fully covers the relationship between the system state, control input, and external disturbances, which conforms to the modeling specifications of multivariable systems in control theory.

[0033] The accuracy of the model was verified using root mean square error (RMSE), and a threshold of RMSE < 5% was set. The deviation between the model's predicted values ​​and the actual values ​​was objectively evaluated through quantitative indicators, avoiding subjective judgment and ensuring that the model's description error of the unit's dynamic characteristics was within an acceptable range. The verification method was scientific and quantifiable.

[0034] S3. Based on the predicted trend, construct an optimization objective function that includes multiple parameter deviations and energy consumption terms. Optimal control quantity is solved by rolling optimization every 10 seconds, and the control quantity is constrained by the rated parameter range.

[0035] Specifically, it should be noted that the optimization objective function is expressed as follows: ,in This indicates the optimization of the objective function value; This indicates the deviation between the main steam pressure at time t and the set value; This indicates the deviation between the steam drum water level and the set value at time t; This indicates the deviation between the furnace temperature at time t and the set value; This represents the deviation between the turbine speed at time t and the set value; This represents the fuel quantity adjustment at time t; This represents the water supply adjustment amount at time t; This represents the air supply volume adjustment at time t; Indicates the fuel energy consumption coefficient. This represents the energy consumption coefficient of the water supply. The energy consumption coefficient for air supply is obtained through unit heat balance calculation; the parameter range constraint is that the minimum value of the objective function is solved every 10 seconds during the rolling optimization process. ,in Indicates the rated fuel quantity; ,in Indicates the rated water supply; ,in This represents the rated air supply volume. The core objective of coordinated control is to balance stability and economy: parameter deviations directly affect unit stability, and energy consumption is directly related to operating costs. The objective function integrates the sum of squares of both, minimizing fluctuations and energy consumption simultaneously during optimization, meeting the actual needs of prioritizing safety and economical operation of thermal power units. Rolling optimization is performed every 10 seconds. Since the response time of the unit's control actuators is approximately 5-15 seconds, the 10-second interval allows for timely tracking of operating condition changes while avoiding frequent actuator actions due to excessive optimization frequency, thus matching the dynamic characteristics of the equipment. The deviations of each parameter in the objective function are expressed in square form, amplifying the penalty weight for larger deviations and forcing the optimization results to converge towards the setpoint. The energy consumption term introduces α, β, and γ, weighting the adjustments of fuel, feedwater, and air supply according to actual energy consumption costs, ensuring that economic optimization has a physical basis. Constraints are set based on the equipment's rated values, reserving sufficient range for adjustments while avoiding exceeding the equipment's safe operating limits, complying with industrial equipment safety operating regulations, and ensuring the feasibility of control quantities.

[0036] S4. Input the optimal control quantity into the control execution unit, adjust the opening of the fuel valve, feedwater valve and air supply valve through the mapping relationship, and collect the unit parameters after control in real time.

[0037] Specifically, it should be noted that the mapping relationship is as follows: ,in Indicates the fuel valve opening. Indicates the initial fuel valve opening. Indicates the fuel valve adjustment coefficient; ,in Indicates the opening degree of the water supply valve. Indicates the initial opening degree of the water supply valve. This indicates the adjustment coefficient of the water supply valve; ,in This indicates the opening degree of the air supply valve. This indicates the initial air supply valve opening. Indicates the air supply valve adjustment coefficient; and satisfies The optimized control quantity is an abstract numerical value, which needs to be transformed into the specific action (opening degree) of the actuator to achieve physical adjustment. This mapping is a key and indispensable link in the actual operation of the control strategy.

[0038] The linear relationship between valve opening and adjustment amount conforms to the working characteristics of most industrial valves. It can accurately fit the adjustment amount and opening curve of the actual valve, ensuring that the mapping is without deviation. The opening is constrained to 0-100%, which conforms to the physical limits of the valve, avoids invalid adjustment commands, and ensures the safe operation of the actuator.

[0039] S5. Calculate the state deviation between the actual parameters and the predicted parameters, and use the feedback correction mechanism to correct the model state matrix parameters to improve the model's adaptability.

[0040] Specifically, it should be noted that the aforementioned state deviation... ,in Indicates the actual state. The state matrix represents the predicted state. ,in Let L represent the correction coefficient matrix and R represent the original state matrix. From a rationale perspective, thermal power units are subject to uncertainties such as coal quality fluctuations and equipment aging. The parameters of the prediction model may drift over time, leading to prediction bias. Real-time feedback correction can dynamically adjust the model, offsetting the effects of uncertainties and ensuring that the model always matches the actual unit characteristics. This is a necessary means to improve control robustness. Deviation calculation directly quantifies the difference between the actual and predicted states, providing a clear basis for correction. The correction formula adjusts the state matrix through the correction coefficient matrix L, which is determined by the pole placement method. This ensures rapid deviation convergence, conforming to the logic in adaptive control theory that the larger the deviation, the stronger the correction, ensuring that the model maintains high accuracy even when operating conditions change.

[0041] S6. The rolling optimization steps are executed cyclically to continuously optimize the coordinated control of the thermal power unit until the unit reaches stable operation or receives a new load command. Thermal power units are continuously operating dynamic systems, with constantly changing operating conditions such as load commands and coal quality. A single optimization result cannot adapt to the entire process. Rolling optimization, through a cycle of predictive optimization, correction, and re-prediction, tracks new operating conditions in real time and dynamically updates the control strategy, preventing fixed strategies from becoming ineffective after changes in operating conditions, thus meeting the control requirements of dynamic systems. Furthermore, rolling optimization is the core mechanism of model predictive control, and its effectiveness has been widely verified in industrial control. By repeating S3-S5, the control strategy is always based on the latest operating data and model state, ensuring that control decisions are synchronized with real-time operating conditions, ultimately achieving stable and economical operation of the unit under all operating conditions. The theoretical basis is mature and the practice is feasible.

[0042] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0043] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A model-predictive-based coordinated optimization control method for thermal power units, characterized in that, include: S1. Collect real-time operating parameters of thermal power units through a sensor network, and establish a unit operating status database after preprocessing the real-time operating parameters; S2. Construct a dynamic prediction model for the boiler and turbine based on the operating status database, use state-space equations to describe it, predict the parameter trend for the next 60 seconds and verify the accuracy. S3. Based on the predicted trend, construct an optimization objective function that includes multiple parameter deviations and energy consumption terms. Optimal control quantity is solved by rolling optimization every 10 seconds, and the control quantity is constrained by the rated parameter range. The optimization objective function is specifically expressed as follows: ,in This indicates the optimization of the objective function value; This indicates the deviation between the main steam pressure at time t and the set value; This indicates the deviation between the steam drum water level and the set value at time t; This indicates the deviation between the furnace temperature at time t and the set value; This represents the deviation between the turbine speed at time t and the set value; This represents the fuel quantity adjustment at time t; This represents the water supply adjustment amount at time t; This represents the air supply volume adjustment at time t; Indicates the fuel energy consumption coefficient. This represents the energy consumption coefficient of the water supply. The energy consumption coefficient of the air supply is obtained through unit heat balance calculations; S4. Input the optimal control quantity into the control execution unit, adjust the opening of the fuel valve, feedwater valve and air supply valve through the mapping relationship, and collect the unit parameters after control in real time. S5. Calculate the state deviation between the actual parameters and the predicted parameters, and use the feedback correction mechanism to correct the model state matrix parameters to improve the model's adaptability. S6. Execute the rolling optimization steps in a loop to achieve continuous optimization of the coordinated control of the thermal power unit until the unit is running stably or receives a new load command.

2. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The real-time operating parameters include main steam pressure P, drum water level H, furnace temperature T, turbine speed N, fuel quantity F, feedwater quantity W, air supply quantity Q, and load command L.

3. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The preprocessing includes outlier removal and normalization. Outlier removal uses the 3σ criterion to identify outliers, where σ is the standard deviation of the parameter. The normalization process involves adjusting the parameters after outlier removal according to the formula... Normalization is performed, where Represents the original parameters. Represents the normalized parameters. This indicates the historical minimum value of the parameter. This indicates the historical maximum value of the parameter.

4. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The dynamic prediction model for the furnace and turbine is described using state-space equations: ; ;in Represents the vector of the rate of change of state at time k+1; This represents the output vector at time k+1; The state vector at time k includes the rate of change of main steam pressure ΔP, the rate of change of steam drum water level ΔH, and the rate of change of furnace temperature ΔT. The control vector includes fuel quantity adjustment ΔF, water supply quantity adjustment ΔW, and air supply quantity adjustment ΔQ. The disturbance vector is represented by ΔL, which includes the change in load command. A, B, C, D, and E represent the model parameter matrix, which is identified through historical data.

5. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The model's prediction accuracy is verified using the root mean square error (RMSE); the RMSE is specifically expressed as... ,in Indicates the actual value. Let n represent the predicted value, and n represent the number of samples; and when <5% of models passed validation.

6. The model-predictive-based coordinated optimization control method for thermal power units according to claim 2, characterized in that: The parameter range constraint is that the minimum value of the objective function is calculated every 10 seconds during the rolling optimization process. ,in Indicates the rated fuel quantity; ,in Indicates the rated water supply; ,in This indicates the rated air supply volume.

7. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The mapping relationship is as follows: ,in Indicates the fuel valve opening. Indicates the initial fuel valve opening. Indicates the fuel valve adjustment coefficient; ,in Indicates the opening degree of the water supply valve. Indicates the initial opening degree of the water supply valve. This indicates the adjustment coefficient of the water supply valve; ,in This indicates the opening degree of the air supply valve. This indicates the initial air supply valve opening. Indicates the air supply valve adjustment coefficient; and satisfies .

8. The model-predictive-based coordinated optimization control method for thermal power units according to claim 1, characterized in that: The state deviation ,in Indicates the actual state. The state matrix represents the predicted state. ,in Let R represent the correction coefficient matrix, and let R represent the original state matrix.

Images