Power plant electrical automation control system based on MPC technology

Abstract

The invention discloses a power plant electrical automation control system based on an MPC technology, and relates to the technical field of electrical control systems, the power plant electrical automation control system comprises a short-term prediction core, a medium-term optimization core and a long-term scheduling core, the short-term prediction core, the medium-term optimization core and the long-term scheduling core are respectively connected with a shared memory area in a wired manner, the output end of the medium-term optimization core is connected with a control port of a boiler fan frequency converter through a PROFIBUS bus; the output end of the short-term prediction core is connected with a steam turbine vacuum pump driver. Through the short-term prediction kernel, the medium-term optimization kernel and the long-term scheduling kernel, layered optimization control of the power plant electrical system at different time scales is realized, the problem that the control system is difficult to give consideration to real-time adjustment and global optimization is solved, and the control precision and the system operation efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical control system technology, specifically to a power plant electrical automation control system based on MPC technology. Background Technology

[0002] The large-scale integration of new energy power generation has led to increased fluctuations in grid load. Most power plants still rely on traditional PID control algorithms for electrical automation control, which are difficult to apply precisely in complex systems with multivariable coupling and significant nonlinear characteristics. Parameters such as fuel quantity, air volume, and steam pressure are interconnected, making traditional control methods prone to overshoot and response lag.

[0003] Patent CN108270291B discloses a multifunctional power automation control system. This patent ensures that each electrical automation control cabinet has a unique IP code, facilitating identification by monitoring devices and users. Upon alarm occurrence, the system immediately identifies the alarm information and its correspondence with the electrical automation control cabinet. The threshold identification module determines which alarms are critical and require priority processing, and which can be deferred. This improves the timeliness of fault and accident handling and ensures stable system operation.

[0004] The aforementioned patent achieves targeted data monitoring by setting preset thresholds for each electrical automation control cabinet based on the site environment and conditions. With a unique IP code, identification and management are facilitated, and the threshold-exceeding identification module can effectively distinguish the priority of alarm information. While primarily focusing on control cabinet monitoring and alarm management, it fails to address the complex operating conditions caused by frequent parameter changes and strong coupling between equipment in power plant operations, making precise and efficient control difficult.

[0005] Therefore, this application proposes a power plant electrical automation control system based on MPC technology that can realize multi-timescale collaborative control of short-term prediction kernel, medium-term optimization kernel and long-term scheduling kernel. Summary of the Invention

[0006] The purpose of this invention is to provide a power plant electrical automation control system based on MPC technology to solve the technical problems mentioned in the background art, such as the mutual influence of power plant parameters, the tendency of traditional control methods to overshoot and lag, which lead to increased energy consumption and decreased efficiency.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a power plant electrical automation control system based on MPC technology, comprising a short-term prediction core, a medium-term optimization core, and a long-term scheduling core, wherein the short-term prediction core, the medium-term optimization core, and the long-term scheduling core are respectively wired to a shared memory area, and the short-term prediction core is wired to the medium-term optimization core;

[0008] The output of the intermediate optimization core is connected to the control port of the boiler fan frequency converter via a PROFIBUS bus.

[0009] The output of the short-term prediction kernel is connected to the turbine vacuum pump driver.

[0010] Preferably, the intermediate optimization core generates fuel flow control quantities at a 5-minute cycle;

[0011] The short-term prediction kernel receives water level sensor data every 10 seconds and calls the turbine vacuum dynamic model to convert it into a condenser water level value h. w The vacuum regulating valve opening command is calculated in 10-second cycles. The short-term prediction of the core embedded condenser pressure constraint is used. When the predicted value exceeds the constraint, the penalty term coefficient λ2 is increased to 3 times the original value.

[0012] Mid-term optimization of the core-connected temperature sensor and boiler combustion state-space model; optimization of fuel / air mixture ratio with a 5-minute cycle; rolling solution; generation of fuel command u f And write it to the shared memory area;

[0013] The long-term dispatch core receives current transformer data and couples it with a line loss calculation model. It outputs reactive power compensation commands in a 1-hour cycle. The output of the long-term dispatch core is connected to the IGBT trigger module of the reactive power compensation device via a relay circuit.

[0014] Preferably, the turbine vacuum dynamic model, the boiler combustion state space model, and the line loss calculation model are stored in the dynamic model library;

[0015] The dynamic model library is wired to a multi-source data acquisition layer, a short-term prediction kernel, a medium-term optimization kernel, and a long-term scheduling kernel;

[0016] The multi-source data acquisition layer includes a water level sensor for the turbine condenser, a temperature sensor for the boiler combustion chamber, and a current transformer for the power grid bus. Temperature sensor data is transmitted to the mid-term optimization core via an optical fiber network.

[0017] Preferably, the long-term dispatching core collects the three-phase current I of the current transformer. abc and electricity price C e (t) Calculate the real-time loss and send the reactive power constraint to the short-term prediction kernel via the TSN bus.

[0018] Preferably, the turbine vacuum dynamic model describes the nonlinear relationship between condenser water level, circulating water temperature and vacuum degree.

[0019] Preferably, the short-term predicted nuclear output vacuum pump speed increment Δu p ;

[0020] Mid-term optimized kernel receives short-term predicted kernel Δu pAs a feedforward for boiler feedwater;

[0021] Long-term scheduling kernel read u f Calculate carbon emission constraints.

[0022] Preferably, the dynamic model library receives infrared temperature data from the mobile inspection robot via a wireless network;

[0023] The dynamic model library incorporates a data fusion engine that performs weighted calculations on temperature sensor data and infrared data.

[0024] The shared memory area is divided into a real-time data segment, a model parameter segment, and an instruction cache segment. The real-time data segment stores the sensor acquisition values ​​of the most recent 10 minutes, the model parameter segment stores the current parameter matrix of the dynamic model library, and the instruction cache segment retains control instructions for 3 cycles.

[0025] The shared memory area is configured with a dual-port RAM chip, which supports parallel read and write operations for short-term prediction cores, medium-term optimization cores, and long-term scheduling cores, and each data block is equipped with a timestamp and checksum.

[0026] Preferably, the rolling solution constructs an objective function J for the short-term prediction kernel. s ;

[0027] The model predictive control algorithm is used to solve the mid-term optimization kernel.

[0028] Preferably, the multi-source data acquisition layer acquires the pressure signal transmitted by the piezoresistive sensor of the boiler's main steam pipeline in real time;

[0029] The pressure signal is converted into a 4-20mA current through the safety barrier and transmitted to the shared memory area marked P via the data bus. act ;

[0030] Mid-term optimized nucleogenesis steam pressure prediction sequence P for the next 30 minutes pred P pred Write to the shared memory area;

[0031] Extracting P from the dynamic model library act and P pred The comparison circuit in the dynamic model library calculates the power deviation. When the power deviation exceeds 5%, the controller is interrupted. The controller then updates the parameter matrix B of the boiler combustion state-space model using the recursive least squares method. b ;

[0032] When the controller is interrupted, the standby computing thread reads the P values ​​from the shared memory area for the last 5 cycles. act and P pred The parameter matrix B is updated using a recursive least squares method with a forgetting factor λ=0.95. bAfter the parameters are updated, they are verified using a validation function. If the validation function value is less than 10%, it is written to the dynamic model library; if the validation function value is greater than or equal to 10%, the original parameters are retained and an exception log is recorded.

[0033] Preferably, the boiler combustion state space model adopts a second-order discrete form.

[0034] Compared with the prior art, the beneficial effects of the present invention are:

[0035] 1. This invention achieves hierarchical optimization control of power plant electrical systems at different time scales through short-term prediction kernel, medium-term optimization kernel, and long-term scheduling kernel, solving the problem that control systems cannot simultaneously take into account real-time adjustment and global optimization, and improving control accuracy and system operating efficiency;

[0036] 2. This invention achieves accurate model prediction based on multi-dimensional data by constructing a dynamic model library, solving the problem of insufficient model accuracy caused by a single data source and improving the accuracy of system state prediction;

[0037] 3. This invention achieves efficient collaboration of multi-core parallel data reading and writing by utilizing the shared memory area of ​​a dual-port RAM chip, solving the problems of multiple data access conflicts and consistency, and improving system response speed;

[0038] 4. This invention achieves dynamic adaptive optimization of the control model by updating and verifying the model parameters online, thus solving the problem of control performance degradation caused by model parameter drift during long-term operation and maintaining the stability of system control accuracy. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;

[0040] Figure 2 This is a schematic diagram of the data generation and transmission path of the present invention;

[0041] Figure 3 This is a schematic diagram of the model update fault tolerance process of the present invention;

[0042] Figure 4 This is a schematic diagram illustrating the correlation between vacuum level and thermal efficiency in this invention. Detailed Implementation

[0043] 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.

[0044] Please see Figure 1 , Figure 2 , Figure 3 and Figure 4 This invention provides an embodiment of a power plant electrical automation control system based on MPC technology. The short-term prediction core, medium-term optimization core, and long-term scheduling core are wired to a shared memory area. The short-term prediction core is wired to the medium-term optimization core. The output of the medium-term optimization core is connected to the control port of a boiler fan frequency converter via a PROFIBUS bus. The output of the short-term prediction core is connected to a turbine vacuum pump driver. The short-term prediction core outputs the vacuum pump speed increment Δu. p ;

[0045] The short-term prediction kernel receives water level sensor data every 10 seconds and calls the turbine vacuum dynamic model to convert it into a condenser water level value h. w The vacuum regulating valve opening command is calculated in 10-second cycles. The short-term prediction of the core embedded condenser pressure constraint is used. When the predicted value exceeds the constraint, the penalty term coefficient λ2 is increased to 3 times the original value.

[0046] Furthermore, the nonlinear relationship between condenser water level, circulating water temperature, and vacuum degree is as follows:

[0047]

[0048] P v T represents the vacuum level of the condenser. c dP represents the temperature of the circulating water. v Indicates a minute increment in the degree of vacuum;

[0049] Construct the objective function

[0050]

[0051] λ1 and λ2 are set based on power plant operating experience, with priority given to ensuring water level stability;

[0052] Introducing condenser pressure constraints and speed increment constraints |Δu p |≤5%, when the condenser pressure exceeds 8kPa, λ2 will be automatically adjusted to 0.6 (increasing the speed adjustment weight).

[0053] J is solved using a quadratic programming algorithm. s The minimum value is used to obtain the optimal speed increment, which is then converted into a vacuum regulating valve opening command. This command is sent to the turbine vacuum pump driver via the PROFIBUS bus, and the optimal speed increment is written into the instruction cache segment of the shared memory area, while also being timestamped.

[0054] The function can predict the trend of vacuum changes, and then the vacuum can be stabilized by adjusting the water level or circulating water flow, thus ensuring the efficient operation of the steam turbine.

[0055] Please see Figure 1 , Figure 3 and Figure 4 The present invention provides an embodiment of a power plant electrical automation control system based on MPC technology, wherein the boiler combustion state space model adopts a second-order discrete form; the dynamic model library receives infrared temperature data from a mobile inspection robot via a wireless network; and the data fusion engine embedded in the dynamic model library performs weighted calculations on temperature sensor data and infrared data.

[0056] The turbine vacuum dynamic model, boiler combustion state space model, and line loss calculation model are stored in the dynamic model library. The dynamic model library is wired to the multi-source data acquisition layer, short-term prediction kernel, medium-term optimization kernel, and long-term scheduling kernel. The multi-source data acquisition layer includes the water level sensor of the turbine condenser, the temperature sensor of the boiler combustion chamber, and the current transformer of the power grid bus. The temperature sensor data is transmitted to the medium-term optimization kernel through the optical fiber network.

[0057] Furthermore, the short-term prediction core uses an 8-core processor, with a control cycle set to 10 seconds;

[0058] Mid-term optimization configuration: 16-core processor, running LinuxRT_PREEMPT real-time kernel, with optimization cycle set to 5 minutes;

[0059] The long-term scheduling core uses an industrial-grade processor, and the scheduling cycle is set to 1 hour.

[0060] The shared memory area uses a dual-port RAM chip, which is divided into a real-time data segment, a model parameter segment, and an instruction cache segment. It supports 32-bit parallel data read and write, and each data block is appended with a 64-bit timestamp and a 32-bit CRC checksum.

[0061] The dynamic model library is deployed on an SSD hard drive and pre-stores dynamic models of steam turbine vacuum, boiler combustion state space models, line loss calculation models, etc. The model call response time is ≤10ms.

[0062] Boiler combustion kinetics model establishes fuel flow rate u f With steam pressure y p State-space equations;

[0063] The line loss model calculates resistance loss based on cable length, cross-sectional area, and load current.

[0064] The turbine condenser water level sensor uses a 4-20mA current output, a range of 0-5m, and a sampling frequency of 1kHz. It begins acquiring water level signals h after passing a self-test. w (t), a sample value is generated every 100ms, high-frequency noise is removed by a Kalman filter, and the processed data is stored in the real-time data segment;

[0065] The boiler combustion chamber temperature sensor transmits data to the mid-term optimization kernel via optical fiber, updates the data every 500ms, and uses a moving average filter to eliminate instantaneous fluctuations;

[0066] The power grid bus current transformer collects the three-phase current and outputs an effective value every 10ms. Then it calculates the three-phase unbalance and triggers a warning when the unbalance is greater than 5%.

[0067] Please see Figure 1 , Figure 2 and Figure 3 The present invention provides an embodiment of a power plant electrical automation control system based on MPC technology, wherein the long-term dispatch core receives current transformer data and couples it with a line loss calculation model, and outputs reactive power compensation instructions in a 1-hour cycle. The output terminal of the long-term dispatch core is connected to the IGBT trigger module of the reactive power compensation device via a relay circuit.

[0068] Long-term dispatching of the three-phase current I of the current transformer abc and electricity price C e (t), calculate real-time losses, and send reactive power constraints to the short-term prediction core via the TSN bus; the long-term scheduling core reads u f Calculate carbon emission constraints;

[0069] Furthermore, C e (t) Data from the electricity market interface is used to calculate the resistance of each phase based on the transmission line resistance and line length, and then the real-time line loss P is calculated. loss ;

[0070] Construct the reactive power compensation objective function

[0071]

[0072] Q c γ1 represents the reactive power compensation amount, γ2 represents the line loss weight, and γ2 represents the compensation cost weight.

[0073] Solving for the optimal Q c Verify constraint |Q c |≤0.3×S n (S) n =10000kvar, rated capacity), after the conditions are met, a trigger signal is generated, which drives the IGBT module (model: Infineon FF300R12KE4) of the reactive power compensation device through the relay circuit to switch the capacitor bank.

[0074] Please see Figure 1 , Figure 3 and Figure 4 One embodiment of the present invention provides a power plant electrical automation control system based on MPC technology, wherein the intermediate optimization core generates fuel flow control quantities at 5-minute intervals, and the intermediate optimization core receives Δu from the short-term prediction core.p As a feedforward for boiler feedwater; in the mid-term optimization, the core-connected temperature sensor and boiler combustion state-space model are optimized, and the fuel / air mixture ratio is optimized at a 5-minute cycle, with rolling solutions to generate fuel command u. f And write it to the shared memory area;

[0075] Furthermore, the intermediate-term optimization kernel reads the Δu of the short-term prediction kernel from the shared memory area. p Current parameters of the boiler combustion state-space model (from the model parameter segment), real-time furnace temperature T furnace Oxygen content (O2) in flue gas (from temperature and oxygen sensors);

[0076] Amplitude limiting processing Δu p If Δu p If the amplitude is 6%, then the amplitude will be 5% after the limiting process.

[0077] Constructing the boiler combustion optimization objective function

[0078] .

[0079] μ1, μ2, and μ3 are weighting coefficients;

[0080] Based on the rolling solution of the model predictive control algorithm, the prediction time domain N is obtained. p =6, (covering 30 minutes), controlling the time domain N c =3, the constraint is that the fuel flow rate 50 ≤ u f For fuel flow rates ≤150t / h and air-to-fuel ratios ≥1.2, the gradient descent method is used to solve the problem, and the optimal fuel command u is obtained after 100 iterations. f The data is written to the shared memory area and sent to the boiler fan frequency converter (model: ABBACS880) via the PROFIBUS bus to adjust the fuel valve opening.

[0081] Please see Figure 1 , Figure 3 and Figure 4 One embodiment of the present invention is a power plant electrical automation control system based on MPC technology, which extracts P from a dynamic model library. act and P pred The comparison circuit in the dynamic model library calculates the power deviation. When the power deviation exceeds 5%, the controller is interrupted. The controller then updates the parameter matrix B of the boiler combustion state-space model using the recursive least squares method. b ;

[0082] When the controller is interrupted, the standby computing thread reads the P values ​​from the shared memory area for the last 5 cycles. act and P pred The parameter matrix B is updated using a recursive least squares method with a forgetting factor λ=0.95. bAfter the parameters are updated, they are verified by a validation function. If the validation function value is less than 10%, it is written to the dynamic model library. If the validation function value is greater than or equal to 10%, the original parameters are retained and an exception log is recorded.

[0083] Furthermore, after receiving the instruction, the backup prediction core reads the calculation data of the last 10 cycles of the main core from the shared memory area, loads the same turbine vacuum dynamic model as the main core, and sends a heartbeat signal (once every 100ms) to the main core to confirm that the data synchronization is complete.

[0084] After the switchover is complete, the standby core takes over control and outputs the first set of instructions Δu. p =3%, and mark the shared memory area as "standby core running".

[0085] The boiler combustion chamber temperature sensor outputs abnormal values ​​due to electromagnetic interference for 5 consecutive sampling cycles;

[0086] An anomaly detector in the multi-source data acquisition layer triggers an alert, invokes the data fusion engine, receives infrared temperature data from the mobile inspection robot, reads the historical standard deviation of the temperature sensor and the standard deviation of the infrared data, calculates a weighted coefficient, and fuses the temperature. The mid-term optimization kernel uses the fused temperature as input to adjust the fuel command u. f To avoid the furnace temperature from being too low.

[0087] Working principle: The multi-source data acquisition layer includes turbine condenser water level sensors, boiler combustion chamber temperature sensors, and power grid bus current transformers, which collect data such as water level, temperature, and current. Temperature sensor data is transmitted via optical fiber, and infrared temperature data from the mobile inspection robot is transmitted to the dynamic model library via wireless network. The data fusion engine of the dynamic model library will weight and process the temperature sensor data and infrared data to achieve multi-source data fusion. The processed data is transmitted to the dynamic model library and each core.

[0088] Then, each core performs hierarchical prediction and optimization control based on the model in the dynamic model library, and exchanges data through the shared memory area. The dynamic model library stores the turbine vacuum dynamic model, the boiler combustion state space model, and the line loss calculation model. The shared memory area is divided into real-time data segment, model parameter segment, and instruction cache segment, which supports parallel reading and writing of short-term prediction core, medium-term optimization core, and long-term scheduling core to achieve data sharing.

[0089] The short-term prediction kernel receives water level sensor data every 10 seconds, calls the turbine vacuum dynamic model, and combines condenser pressure constraints and speed increment constraints to construct an objective function J. s The optimal speed increment is solved using quadratic programming, which is then converted into a vacuum regulating valve opening command to control the turbine vacuum pump, and the result is written to the shared memory area.

[0090] The mid-term optimization kernel is based on the boiler combustion state-space model. It receives the speed increment of the short-term prediction kernel as a feedforward quantity, combines it with the fused temperature data, optimizes the fuel / air mixture ratio through model predictive control algorithm, generates fuel commands under constraints such as fuel flow rate, controls the boiler fan frequency converter via PROFIBUS bus, and writes the data to the shared memory area.

[0091] The long-term dispatching core receives current transformer data, couples it to a line loss calculation model, and constructs a reactive power compensation objective function J1 based on electricity price data. It then solves for the optimal reactive power compensation amount and controls the IGBT module of the reactive power compensation device via a relay circuit to achieve line loss optimization.

[0092] Each core will embed constraints during operation, and the weight coefficients will be automatically adjusted when the constraints are exceeded; the dynamic model library updates the model parameters through power deviation detection, and the standby prediction core will take over control when the main core is abnormal.

[0093] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A power plant electrical automation control system based on MPC technology, characterized in that: It includes a short-term prediction core, a medium-term optimization core, and a long-term scheduling core, which are each wired to a shared memory area. The short-term prediction core is wired to the medium-term optimization core. The output of the intermediate optimization core is connected to the control port of the boiler fan frequency converter via a PROFIBUS bus. The output of the short-term prediction kernel is connected to the turbine vacuum pump driver.

2. The power plant electrical automation control system based on MPC technology according to claim 1, characterized in that: The intermediate optimization kernel generates fuel flow control quantities at a 5-minute cycle. The short-term prediction kernel receives water level sensor data every 10 seconds and calls the turbine vacuum dynamic model to convert it into a condenser water level value h. w The vacuum regulating valve opening command is calculated in 10-second cycles. The short-term prediction of the core embedded condenser pressure constraint is used. When the predicted value exceeds the constraint, the penalty term coefficient λ2 is increased to 3 times the original value. Mid-term optimization of the core-connected temperature sensor and boiler combustion state-space model; optimization of fuel / air mixture ratio with a 5-minute cycle; rolling solution; generation of fuel command u f And write it to the shared memory area; The long-term dispatch core receives current transformer data and couples it with a line loss calculation model. It outputs reactive power compensation commands in a 1-hour cycle. The output of the long-term dispatch core is connected to the IGBT trigger module of the reactive power compensation device via a relay circuit.

3. The power plant electrical automation control system based on MPC technology according to claim 1, characterized in that: The turbine vacuum dynamic model, boiler combustion state space model, and line loss calculation model are stored in the dynamic model library. The dynamic model library is wired to a multi-source data acquisition layer, a short-term prediction kernel, a medium-term optimization kernel, and a long-term scheduling kernel; The multi-source data acquisition layer includes a water level sensor for the turbine condenser, a temperature sensor for the boiler combustion chamber, and a current transformer for the power grid bus. Temperature sensor data is transmitted to the mid-term optimization core via an optical fiber network.

4. A power plant electrical automation control system based on MPC technology according to claim 1, characterized in that: The three-phase current I of the long-term dispatching core acquisition current transformer abc and electricity price C e (t) Calculate the real-time loss and send the reactive power constraint to the short-term prediction kernel via the TSN bus.

5. A power plant electrical automation control system based on MPC technology according to claim 1, characterized in that: The turbine vacuum dynamic model describes the nonlinear relationship between condenser water level, circulating water temperature and vacuum degree.

6. A power plant electrical automation control system based on MPC technology according to claim 1, characterized in that: The short-term predicted nuclear output vacuum pump speed increment Δu p ; Mid-term optimized kernel receives short-term predicted kernel Δu p As a feedforward for boiler feedwater; Long-term scheduling kernel read u f Calculate carbon emission constraints.

7. A power plant electrical automation control system based on MPC technology according to claim 3, characterized in that: The dynamic model library receives infrared temperature data from the mobile inspection robot via a wireless network. The dynamic model library incorporates a data fusion engine that performs weighted calculations on temperature sensor data and infrared data. The shared memory area is divided into a real-time data segment, a model parameter segment, and an instruction cache segment. The real-time data segment stores the sensor acquisition values ​​of the most recent 10 minutes, the model parameter segment stores the current parameter matrix of the dynamic model library, and the instruction cache segment retains control instructions for 3 cycles. The shared memory area is configured with a dual-port RAM chip, which supports parallel read and write operations for short-term prediction cores, medium-term optimization cores, and long-term scheduling cores, and each data block is equipped with a timestamp and checksum.

8. A power plant electrical automation control system based on MPC technology according to claim 2, characterized in that: The rolling solution constructs an objective function J for the short-term prediction kernel. s ; The model predictive control algorithm is used to solve the mid-term optimization kernel.

9. A power plant electrical automation control system based on MPC technology according to claim 3, characterized in that: The multi-source data acquisition layer acquires pressure signals transmitted by the piezoresistive sensor in the main steam pipeline of the boiler in real time. The pressure signal is converted into a 4-20mA current through the safety barrier and transmitted to the shared memory area marked P via the data bus. act ; Mid-term optimized nucleogenesis steam pressure prediction sequence P for the next 30 minutes pred P pred Write to the shared memory area; Extracting P from the dynamic model library act and P pred The comparison circuit in the dynamic model library calculates the power deviation. When the power deviation exceeds 5%, the controller is interrupted. The controller then updates the parameter matrix B of the boiler combustion state-space model using the recursive least squares method. b ; When the controller is interrupted, the standby computing thread reads the P values ​​from the shared memory area for the last 5 cycles. act and P pred The parameter matrix B is updated using a recursive least squares method with a forgetting factor λ=0.

95. b After the parameters are updated, they are verified using a validation function. If the validation function value is less than 10%, it is written to the dynamic model library; if the validation function value is greater than or equal to 10%, the original parameters are retained and an exception log is recorded.

10. A power plant electrical automation control system based on MPC technology according to claim 2, characterized in that: The boiler combustion state space model adopts a second-order discrete form.

Images