Intelligent control method, system, equipment and medium for thermal system of thermal power plant

Abstract

An intelligent control method, system, device and medium for a thermal power plant thermal system relate to the technical field of thermal power plant thermal control, and comprise the following steps: collecting operation parameters of each device of the thermal system in real time, and constructing a multi-source data set; based on a deep neural network model, establishing a dynamic model of the thermal system for learning and analyzing the multi-source data set; adjusting control parameters and control strategies of each device of the thermal system according to the prediction result of the dynamic model and the multi-source data set by adopting a self-adaptive control algorithm; operation characteristics of the thermal system are extracted, a multi-objective optimization function is constructed, and control parameters and control strategies are optimized; the problems that a traditional control method is low in control precision and low in operation stability are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal control of thermal power plants, and in particular to an intelligent control method, system, equipment and medium for a thermal system of a thermal power plant. Background Art

[0002] Overview of Thermal Power Plant Thermal Systems: Thermal power plants utilize fossil fuels (such as coal and natural gas) to generate heat, which is then converted into electricity. As the core of a thermal power plant, the thermal system encompasses boilers, steam turbines, generators, and numerous auxiliary equipment. These devices collaborate through complex process flows to achieve energy conversion and transmission.

[0003] Currently, most thermal power plant thermal systems primarily utilize traditional PID (proportional-integral-differential) control strategies. Under stable operating conditions, PID control can maintain basic system operation. However, faced with complex and changing operating conditions, such as large load fluctuations and fluctuations in fuel quality, traditional PID control exposes numerous problems. For example, when the load rises and falls rapidly, key parameters such as steam temperature and steam pressure fluctuate significantly, making it difficult to quickly stabilize at set values, affecting unit operating efficiency. When fuel quality changes, boiler combustion control struggles to respond quickly, resulting in incomplete combustion, increased pollutant emissions and energy consumption, and an inability to accurately control the correct operation of the thermal system.

[0004] In addition, PID control strategies generally rely on accurate mathematical models, while the thermal system of a thermal power plant has characteristics such as strong nonlinearity, large lag, and multi-variable coupling, making it difficult to establish an accurate mathematical model, resulting in poor control effect; there is a lack of comprehensive perception and intelligent analysis of the system's operating status, and the control strategy cannot be automatically adjusted according to real-time operating conditions, resulting in poor adaptability; there is insufficient coordinated control between different subsystems, and it is easy to lose sight of one thing while focusing on another, affecting the stability and economy of the entire thermal system.

[0005] Chinese patent publication number CN110134095A discloses a method and terminal device for optimizing a thermal analog control system in a thermal power plant. The method identifies an acquired analog process object model to obtain a mathematical model of the analog object; based on the mathematical model of the analog object, the parameters of the acquired analog control system controller model are optimized to obtain a target controller model; when the performance of the target controller model does not meet preset requirements, the analog process object model is reacquired and subsequent operations are performed until the performance of the target controller model meets the preset requirements, terminating the process. This allows the thermal analog control system of a thermal power plant to be optimized, resulting in a control system with high control accuracy, small overshoot, fast control speed, and strong anti-interference capability.

[0006] However, this method relies on transfer function identification to build the model, which has limited ability to fit nonlinear and time-varying systems. In addition, this method is designed for single-loop design and cannot work with multiple systems in a coordinated manner.

[0007] Therefore, we propose a control method that can improve the control accuracy and operation stability of the thermal system of a thermal power plant. Summary of the Invention

[0008] The purpose of the present invention is to provide an intelligent control method, system, equipment and medium for a thermal system of a thermal power plant, which solves the problems of low control accuracy and low operational stability of traditional control methods.

[0009] The present invention is achieved through the following technical solutions:

[0010] The intelligent control method of thermal system in thermal power plant includes:

[0011] Real-time collection of operating parameters of each device in the thermal system to build a multi-source data set;

[0012] Based on the deep neural network model, a dynamic model of the thermal system is established to learn and analyze multi-source data sets;

[0013] Adopting adaptive control algorithms to adjust the control parameters and control strategies of various devices in the thermal system based on the prediction results of the dynamic model and multi-source data sets;

[0014] Extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategies.

[0015] Furthermore, the multi-source data set includes operating parameter variables, operating state variables and disturbance variables, wherein the operating parameter variable u includes coal feed rate, water feed rate, main steam valve opening, each damper opening, water spray desuperheating valve opening and induced draft fan speed;

[0016] The operating state variables x include main steam pressure, main steam temperature, reheat steam temperature, drum water level, furnace negative pressure, flue gas oxygen content, wall temperature of each heating surface, unit load and coal mill current;

[0017] The disturbance variable d includes fuel calorific value, ambient temperature / humidity and grid frequency fluctuation.

[0018] Furthermore, the deep neural network model adopts LSTM, and the dynamic modeling model of LSTM is expressed as:

[0019]

[0020] Where k is the historical backtracking step, t is the time step, H is the prediction time domain, and θ is the trained LSTM network weight parameter.

[0021] Furthermore, the adaptive control algorithm is used to adjust the control parameters and control strategies of each device in the thermal system according to the prediction results of the dynamic model and the multi-source data set. The specific steps are as follows:

[0022] Set the expected value r(t:t+H) of the operating parameters in the multi-source dataset;

[0023] The prediction results of the dynamic model that meets the constraints and the multi-source data set are input into the adaptive control algorithm to calculate the optimal future control action sequence;

[0024] Adjust the control parameters and control strategies of each device in the thermal system according to the optimal future control action sequence.

[0025] Furthermore, the adaptive control algorithm is specifically expressed as:

[0026]

[0027] Where M is the future control time domain, and M≤H; represents the weighted Euclidean norm, Represents the sum of squares of changes in control actions based on weight R.

[0028] Furthermore, the operation characteristics of the thermal system are extracted and a multi-objective optimization function is constructed to optimize the control parameters and control strategy. The specific steps are as follows:

[0029] Extracting the operating characteristics of the thermal system based on the prediction results of the dynamic model, including environmental emission characteristics, equipment safety operation characteristics and power generation load instruction characteristics;

[0030] Define economic indicators based on the power generation load instruction characteristics, define environmental indicators based on environmental emission characteristics, and define equipment life indicators based on equipment safe operation characteristics;

[0031] Construct a multi-objective optimization function based on economic indicators, environmental indicators and equipment life indicators;

[0032] Calculate the multi-objective optimization function, obtain the optimized prediction results, input them into the adaptive control algorithm, replace the prediction results of the dynamic model as input, and finally obtain the optimized control parameters and control strategy.

[0033] Furthermore, the multi-objective optimization function J multi The expression is:

[0034] minJ multi =ω1J econ +ω2J emis +ω3J life

[0035] Where, J econ is the economic index, J emis is an environmental protection index, J life It is an indicator of equipment life.

[0036] The intelligent control system of the thermal system of a thermal power plant includes a data acquisition layer for real-time collection of operating parameters of various equipment in the thermal system and construction of a multi-source data set;

[0037] Data processing layer, used for preprocessing, fusion and storage of multi-source data sets;

[0038] The intelligent analysis and control layer is written with the dynamic model of the thermal system and the adaptive control algorithm. The dynamic model of the thermal system learns and analyzes the multi-source data set. The adaptive control algorithm adjusts the control parameters and control strategies of each device in the thermal system based on the prediction results analyzed by the dynamic model and the multi-source data set.

[0039] The control optimization layer is used to extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategies.

[0040] An electronic device, comprising:

[0041] processor;

[0042] a memory for storing instructions executable by the processor;

[0043] The processor is configured to execute to implement an intelligent control method for a thermal system of a thermal power plant.

[0044] A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to implement an intelligent control method for a thermal system of a thermal power plant.

[0045] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0046] The present invention discloses an intelligent control method, system, equipment and medium for the thermal system of a thermal power plant. By combining the powerful dynamic modeling and prediction capabilities of a deep neural network model with the forward-looking optimization control capabilities of an adaptive control algorithm and a multi-objective intelligent optimization algorithm, a data-driven, closed-loop intelligent control optimization system for the thermal system of a complex thermal power plant is constructed. This system breaks through the limitations of traditional PID and mechanism modeling, significantly improves the system's adaptability, control accuracy and stability under variable operating conditions, and explicitly and collaboratively optimizes the three key goals of economy, environmental protection and equipment reliability in the control closed loop, effectively solving many of the shortcomings of existing control strategies pointed out in the background technology and bringing significant practical benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic flow chart of a method of the present invention is shown;

[0048] Figure 2 A schematic diagram of the system structure of the present invention;

[0049] Figure 3 The figure is a schematic diagram of the structure of an electronic device. DETAILED DESCRIPTION

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0051] Example 1

[0052] like Figure 1 The intelligent control method for the thermal system of a thermal power plant shown in the figure specifically includes:

[0053] S1. Real-time collection of operating parameters of each device in the thermal system to construct a multi-source data set;

[0054] As needed, the multi-source data set includes operating parameter variables, operating state variables and disturbance variables, wherein the operating parameter variable u includes coal feed rate, water feed rate, main steam valve opening, each damper opening, water spray desuperheating valve opening and induced draft fan speed;

[0055] The operating state variables x include main steam pressure, main steam temperature, reheat steam temperature, drum water level, furnace negative pressure, flue gas oxygen content, wall temperature of each heating surface, unit load and coal mill current;

[0056] The disturbance variable d includes fuel calorific value, ambient temperature / humidity and grid frequency fluctuation.

[0057] S2. Build a dynamic model of the thermal system based on a deep neural network model to learn and analyze multi-source datasets;

[0058] In addition, the deep neural network model adopts LSTM (Long Short-Term Memory Network), and the dynamic modeling model of LSTM is expressed as:

[0059]

[0060] Where k is the historical backtracking step, t is the time step, H is the prediction time domain, and θ is the trained LSTM network weight parameter.

[0061] In addition, the LSTM output For key parameters that need to be predicted, such as main steam temperature, main steam pressure, unit efficiency, NOx (nitrogen oxide) emission concentration, etc. in the future period (prediction time domain), that is, to predict the future trend of key parameters in advance, especially when facing upcoming load command changes or known fuel disturbances.

[0062] By applying deep neural network models to the dynamic modeling of thermal systems in thermal power plants, the fundamental problem that traditional mechanism modeling is difficult to accurately characterize strong nonlinearity, large lag, and multivariable coupling characteristics has been effectively solved; and LSTM's powerful modeling ability for time series dependencies makes it particularly suitable for predicting the changing trends of key parameters with significant lags, such as steam temperature and steam pressure.

[0063] In addition, the deep neural network model is trained and learned entirely based on actual operating data, and does not rely on difficult-to-obtain precise physical equations and parameters, which reduces the difficulty and cost of modeling and improves the practicality and generalization ability of the model; and provides high-precision prediction capabilities for future operating conditions, which is the key foundation for achieving forward-looking and adaptive control.

[0064] S3. Adopt an adaptive control algorithm to adjust the control parameters and control strategies of each device in the thermal system based on the prediction results of the dynamic model and multi-source data sets;

[0065] Specifically, S31. setting the expected value r(t:t+H) of the operating parameter in the multi-source data set;

[0066] S32. Input the prediction results of the dynamic model and the multi-source data set that meet the constraints into the adaptive control algorithm to calculate the optimal future control action sequence; wherein the prediction results of the dynamic model and the constraints of the multi-source data set are:

[0067] Prediction Model:

[0068]

[0069] Control variable change rate constraint:

[0070] Δu min ≤Δu(t+j)≤Δu max

[0071] Δu min and Δu max To control the minimum and maximum limits of the input change rate;

[0072] Output Constraints:

[0073]

[0074] ymin and y max It is the safe operating limit of the output variable in the thermal system;

[0075] S33. Adjust the control parameters and control strategies of each device in the thermal system based on the optimal future control action sequence.

[0076] The adaptive control algorithm is specifically expressed as:

[0077]

[0078] Where M is the future control time domain, and M≤H; represents the weighted Euclidean norm, represents the sum of squares of the changes in the control actions based on the weight R, and j is the position in the control action sequence.

[0079] And the ultimate goal of this adaptive control algorithm is to find the optimal future control action sequence u * (t:t+M-1), and only the first control action u of the optimization sequence * (t) is actually applied to the controlled process, and then at the next sampling time t+1, the above steps are repeated;

[0080] The specific calculation process of this algorithm is:

[0081] At each moment, based on the latest state u(t+j) and prediction Resolve the optimization problem and automatically calculate the disturbance variable d and the optimal control action u that are suitable for the current working conditions * (t);

[0082] For example, when the load changes and the steam temperature / pressure is predicted to deviate from the set value, MPC will coordinate the changes in the boiler combustion rate (fuel / air volume) and the turbine throttle opening in advance to achieve a fast and smooth transition;

[0083] When fuel quality fluctuates: If the combustion effect or steam temperature is predicted to change, MPC will optimize the air-coal ratio (primary air, secondary air, and burnout air ratio) and coal feed in real time to ensure sufficient and stable combustion and maintain steam parameters.

[0084] The adaptive control algorithm adopts model predictive control (MPC), and the trained LSTM deep neural network model is directly used as the core prediction model of model predictive control (MPC). This combination fully utilizes the advantages of deep learning in complex system modeling and prediction, as well as the advantages of MPC in processing multiple variables, constraints and rolling optimization. Each step of the adaptive control algorithm is based on the latest real-time measurement data and the prediction of the deep neural network model's response to future disturbances (such as fuel changes, load instructions), and re-optimizes and calculates the current optimal control action u* (t), which enables it to automatically, quickly and effectively adapt to various complex and changeable working conditions, overcoming the fatal weakness of traditional PID with fixed parameters and poor adaptability;

[0085] Furthermore, the MPC framework naturally supports the coordinated optimization of multi-input, multi-output systems. For example, when load changes, it can simultaneously optimize the fuel, air flow, and desuperheating water on the boiler side and the throttle opening on the turbine side, achieving coordinated and rapid stabilization of steam temperature and pressure, thus resolving the issue of "focusing on one thing and neglecting another" between subsystems.

[0086] S4. Extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategy;

[0087] The purpose of constructing a multi-objective optimization function is to coordinate and optimize the conflicting economic indicators (coal consumption), environmental indicators (emissions), and equipment life indicators (critical equipment stress, operational stability) while meeting power generation demand (load instructions) and basic safety constraints, so as to achieve global optimal operation.

[0088] Specifically, S41. Extract the operating characteristics of the thermal system based on the prediction results of the dynamic model, which include environmental emission characteristics NO x (t+i), equipment safety operation characteristics and power generation load instruction features P(t+i); these features are output based on LSTM Including: extracted from data such as main steam temperature, main steam pressure, unit efficiency, NOx (nitrogen oxides) emission concentration, etc.

[0089] S42 defines economic indicators based on the characteristics of power generation load instructions, defines environmental indicators based on environmental emission characteristics, and defines equipment life indicators based on equipment safety operation characteristics;

[0090] S43. Construct a multi-objective optimization function based on economic indicators, environmental indicators, and equipment life indicators;

[0091] S44. Calculate the multi-objective optimization function, obtain the optimized prediction results, input them into the adaptive control algorithm, replace the prediction results of the dynamic model as input, and finally obtain the optimized control parameters and control strategy.

[0092] The multi-objective optimization function J multi The expression is:

[0093] minJ multi =ω1J econ +ω2J emis +ω3J life

[0094] Where, J econis the economic index, J emis is an environmental protection index, J life is the equipment life index, ω1, ω2 and ω3 are weights.

[0095] By combining the three traditionally conflicting goals of economy (coal consumption), environmental protection (NOx emissions) and equipment life / operation stability (parameter fluctuations, equipment stress), the mathematical function J multi The system is explicitly and quantitatively expressed and incorporated into a closed-loop optimization framework. This goes beyond the traditional approach of focusing on the stability or local optimization of a single parameter (such as steam temperature).

[0096] In addition, intelligent optimization algorithms such as NSGA-II (Non-Dominated Sorting Genetic Algorithm II) or MOPSO (Multi-Objective Particle Swarm Optimization) are used to solve the multi-objective optimization function in real time or quasi-real time. The optimization results, that is, the optimal setpoints or directly act on the MPC objective function, are then fed back into the control loop to form a closed-loop global optimization decision-making mechanism. This enables the system to actively find and maintain the optimal or satisfactory operating point that takes into account efficiency, environmental protection, and reliability, while meeting power generation needs and strict constraints.

[0097] Moreover, whether it is outer steady-state optimization or integration into the objective function of MPC, multi-objective optimization fully utilizes the predictive ability of LSTM, enabling it to consider the impact of future operating condition changes on the target and make forward-looking optimization decisions.

[0098] Example 2

[0099] like Figure 2 The shown intelligent control system for the thermal system of a thermal power plant includes a data acquisition layer for collecting the operating parameters of each device in the thermal system in real time and building a multi-source data set;

[0100] Data processing layer, used for preprocessing, fusion and storage of multi-source data sets;

[0101] The intelligent analysis and control layer is written with the dynamic model of the thermal system and the adaptive control algorithm. The dynamic model of the thermal system learns and analyzes the multi-source data set. The adaptive control algorithm adjusts the control parameters and control strategies of each device in the thermal system based on the prediction results analyzed by the dynamic model and the multi-source data set.

[0102] The control optimization layer is used to extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategies.

[0103] Example 3

[0104] like Figure 3 An electronic device as shown includes:

[0105] processor;

[0106] a memory for storing instructions executable by the processor;

[0107] The processor is configured to execute to implement an intelligent control method for a thermal system of a thermal power plant.

[0108] A non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to implement an intelligent control method for a thermal system of a thermal power plant.

[0109] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An intelligent control method for thermal system of thermal power plant, characterized in that: This includes real-time collection of operating parameters of various devices in the thermal system and construction of multi-source data sets; Based on the deep neural network model, a dynamic model of the thermal system is established to learn and analyze multi-source data sets; Adopting adaptive control algorithms to adjust the control parameters and control strategies of various devices in the thermal system based on the prediction results of the dynamic model and multi-source data sets; Extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategies.

2. The intelligent control method for thermal system of thermal power plant according to claim 1, characterized in that: The multi-source data set includes operating parameter variables, operating state variables and disturbance variables, wherein the operating parameter variable u includes coal feed rate, water feed rate, main steam valve opening, each damper opening, water spray desuperheating valve opening and induced draft fan speed; The operating state variables x include main steam pressure, main steam temperature, reheat steam temperature, drum water level, furnace negative pressure, flue gas oxygen content, wall temperature of each heating surface, unit load and coal mill current; The disturbance variable d includes fuel calorific value, ambient temperature / humidity and grid frequency fluctuation.

3. The intelligent control method for thermal system of thermal power plant according to claim 2, characterized in that: The deep neural network model adopts LSTM, and the dynamic modeling model of LSTM is expressed as: Where k is the historical backtracking step, t is the time step, H is the prediction time domain, and θ is the trained LSTM network weight parameter.

4. The intelligent control method for thermal system of thermal power plant according to claim 3, characterized in that: The adaptive control algorithm is used to adjust the control parameters and control strategies of each device in the thermal system based on the prediction results of the dynamic model and the multi-source data set. The specific steps are as follows: Set the expected value r(t:t+H) of the operating parameters in the multi-source dataset; The prediction results of the dynamic model that meets the constraints and the multi-source data set are input into the adaptive control algorithm to calculate the optimal future control action sequence; Adjust the control parameters and control strategies of each device in the thermal system according to the optimal future control action sequence.

5. The intelligent control method for thermal system of thermal power plant according to claim 4, characterized in that: The adaptive control algorithm is specifically expressed as: Where M is the future control time domain, and M≤H; represents the weighted Euclidean norm, It represents the sum of squares of the changes in the control actions based on the weight R, i is each step in the prediction time domain, and j is the position in the control action sequence.

6. The intelligent control method for thermal system of thermal power plant according to claim 4, characterized in that: The specific steps of extracting the operating characteristics of the thermal system and constructing a multi-objective optimization function to optimize the control parameters and control strategy are as follows: Extracting the operating characteristics of the thermal system based on the prediction results of the dynamic model, including environmental emission characteristics, equipment safety operation characteristics and power generation load instruction characteristics; Define economic indicators based on the power generation load instruction characteristics, define environmental indicators based on environmental emission characteristics, and define equipment life indicators based on equipment safe operation characteristics; Construct a multi-objective optimization function based on economic indicators, environmental indicators and equipment life indicators; Calculate the multi-objective optimization function, obtain the optimized prediction results, input them into the adaptive control algorithm, replace the prediction results of the dynamic model as input, and finally obtain the optimized control parameters and control strategy.

7. The intelligent control method for thermal system of thermal power plant according to claim 6, characterized in that: The multi-objective optimization function J multi The expression is: mice multi =ω1J econ +ω2J emis +ω3J life Where, J econ is the economic index, J emis is an environmental protection index, J life It is an indicator of equipment life.

8. Intelligent control system for thermal system of thermal power plant, characterized by: It includes a data acquisition layer, which is used to collect the operating parameters of each device in the thermal system in real time and build a multi-source data set; Data processing layer, used for preprocessing, fusion and storage of multi-source data sets; The intelligent analysis and control layer is written with the dynamic model of the thermal system and the adaptive control algorithm. The dynamic model of the thermal system learns and analyzes the multi-source data set. The adaptive control algorithm adjusts the control parameters and control strategies of each device in the thermal system based on the prediction results analyzed by the dynamic model and the multi-source data set. The control optimization layer is used to extract the operating characteristics of the thermal system and construct a multi-objective optimization function to optimize the control parameters and control strategies.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute to implement the intelligent control method for a thermal system of a thermal power plant according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement the intelligent control method for a thermal system of a thermal power plant according to any one of claims 1 to 7.

Images