A real-time control method and system for electron cyclotron resonance heating power

Abstract

This invention discloses a real-time control method for electron cyclotron resonance heating power based on model predictive control, belonging to the field of electron cyclotron resonance heating control technology. The method includes a piecewise modeling step, dividing the operating range of the electron cyclotron resonance heating system into multiple power intervals according to the output power, and establishing a linear system dynamic model describing the dynamic characteristics of the system within each interval; and a real-time control loop step, acquiring the real-time output power and state estimate in each control cycle, determining the target model to be used based on the real-time output power and a hysteresis switching strategy, constructing a quadratic programming problem based on the target model and a preset power reference trajectory, including an objective function and constraints that minimize the power tracking error, solving for the optimal control input and applying it to the system. This invention aims to solve the problem that existing control methods struggle to simultaneously consider adjustment speed and multivariable constraints, achieving rapid and accurate control of electron cyclotron resonance heating power, and improving the safety and stability of system operation.

Description

Technical Field

[0001] This invention relates to the field of electron cyclotron resonance heating control technology, and in particular to a real-time control method and system for electron cyclotron resonance heating power. Background Technology

[0002] Electron cyclotron resonant heating is an important auxiliary heating method in nuclear fusion devices. Its principle is to increase the plasma temperature and density by emitting microwave energy into the plasma. The power control of electron cyclotron resonant heating systems is characterized by strong nonlinearity, multivariable coupling, and rapid dynamic changes. Traditional proportional-integral-derivative (PID) control strategies are insufficient to meet the requirements for high precision and fast response.

[0003] In the actual operation of an electronic cyclotron resonant heating system, there are various physical constraints. For example, the anode voltage, cathode voltage, filament power, and magnet current of the cyclotron tube must be maintained within safe operating ranges, while the cathode temperature is also subject to strict safety limits. Existing proportional-integral-derivative (PID) control methods are difficult to effectively handle the above-mentioned multivariable coupling problems and strict physical constraints simultaneously. In practical applications, they often suffer from slow adjustment speed, large overshoot, and poor constraint handling capabilities, which can easily lead to equipment overvoltage, overcurrent, or frequent shutdown protection triggering due to touching safety boundaries.

[0004] Therefore, how to achieve rapid and precise control of electron cyclotron resonance heating power, and effectively handle multivariables and complex physical constraints while ensuring system safety, has become an urgent technical challenge. Summary of the Invention

[0005] The main objective of this invention is to provide a real-time control method and system for electron cyclotron resonance heating power, aiming to achieve rapid and accurate control of electron cyclotron resonance heating power, and effectively handle multiple variables and complex physical constraints while ensuring system safety.

[0006] To achieve the above objectives, this invention proposes a real-time control method for electron cyclotron resonance heating power, comprising the following steps:

[0007] Segmented modeling steps: Divide the working range of the electron cyclotron resonance heating system into multiple power intervals according to the output power, and for each power interval, establish a linear system dynamic model describing the dynamic characteristics of the system within that interval.

[0008] Real-time control cycle steps: Perform the following operations within each control cycle:

[0009] Obtain the real-time output power and state estimate of the electron cyclotron resonance heating system;

[0010] The target model used in the current control cycle is determined based on the real-time output power; the target model is selected from multiple linear system dynamic models established in the piecewise modeling step.

[0011] Based on the target model and the preset power reference trajectory, a quadratic programming problem for model predictive control is constructed. The quadratic programming problem includes an objective function that minimizes the power tracking error, as well as constraints on the control input and system state. The constraints include at least the magnitude and rate of change constraints on the control input.

[0012] The optimal control input is obtained by solving the quadratic programming problem and applied to the electron cyclotron resonance heating system.

[0013] Preferably, in the segmented modeling step, the division of the multiple power intervals is determined based on the nonlinear transition characteristics of the electron cyclotron resonance heating system; the multiple power intervals include at least a low power interval, a medium power interval, and a high power interval, and adjacent power intervals are divided by a preset power threshold; the linear system dynamic model is established by conducting system excitation experiments and collecting input and output data at the geometric center point of each power interval.

[0014] Preferably, the dynamic model of the linear system is a discrete-time state-space model, which takes the following form:

[0015]

[0016] in, For state vectors, For the input vector, For output; the state vector Includes cathode temperature, magnet power supply current, and gyrotron electron beam current; the input vector Includes filament power, anode voltage, and cathode voltage; the output This is the output power of the gyrotron; , , , These are the system matrix, input matrix, output matrix, and feedforward matrix for the corresponding power ranges.

[0017] Preferably, determining the target model used in the current control cycle based on the real-time output power includes:

[0018] At the switching boundary between adjacent power ranges, a hysteresis switching strategy is used to determine the target model;

[0019] The specific process of the hysteresis switching strategy is as follows: set a switching threshold and a preset hysteresis bandwidth; when the real-time output power of the electron cyclotron resonance heating system crosses the switching threshold and exceeds the hysteresis bandwidth, update the target model; otherwise, keep the currently used target model.

[0020] Preferably, the mathematical expression of the objective function is:

[0021]

[0022] in, To predict the time domain, To control the time domain, For the predicted output power, For reference power trajectory, To control the rate of change of the input, For soft-constrained slack variables, and This is the weight matrix. This is the penalty coefficient.

[0023] Preferably, the constraints include:

[0024] Input amplitude constraints: Limit filament power, anode voltage, and cathode voltage to their respective preset allowable ranges;

[0025] Input rate of change constraint: Limits the rate of change of filament power, anode voltage, and cathode voltage to their respective preset maximum values;

[0026] State constraints: limit the cathode temperature to be less than or equal to the preset maximum temperature, limit the magnet power supply current to be less than or equal to the preset maximum current, and limit the gyrotron electron beam current to be less than or equal to the preset maximum electron beam current.

[0027] Preferably, the constraints further include output power constraints; the output power constraints introduce slack variables to achieve soft constraints, specifically including upper bound constraints and lower bound constraints:

[0028] The upper limit constraint is:

[0029]

[0030] The lower bound constraint is:

[0031]

[0032] in, This is a fundamental term for output power prediction. This is the coefficient representing the influence of the input rate of change on the output power. For the slack variable, and These are the preset maximum and minimum output power values, respectively. Indicates the first The rate of change of the control input at any given time.

[0033] Preferably, in the real-time control loop step, the state estimate is obtained through a Kalman filter algorithm; the Kalman filter algorithm includes a prediction step that uses the target model to predict the state at the next moment, and an update step that corrects the prediction result by combining the actual measurement value.

[0034] Preferably, the method further includes an online adaptive update step:

[0035] The input and output data of the electron cyclotron resonance heating system are recorded in real time to a ring buffer.

[0036] When the preset update trigger condition is met, the parameters of the linear system dynamic model are updated using the data in the circular buffer and the recursive least squares method.

[0037] The update triggering conditions include timed triggering, performance degradation triggering, or operation mode change triggering.

[0038] Preferably, the method further includes a safety rollback step:

[0039] After solving the quadratic programming problem, determine whether the solution was successful;

[0040] If the solution fails or times out, or if a system interlock signal is detected, a rollback control strategy is executed.

[0041] The fallback control strategy includes maintaining the previously effective control variable, reducing the output power at the maximum permissible slope, or switching to a standby control law.

[0042] This application also discloses a real-time control system for electron cyclotron resonance heating power, including:

[0043] The segmented modeling module is used to divide the working range of the electron cyclotron resonance heating system into multiple power intervals according to the output power, and to establish a linear system dynamic model describing the dynamic characteristics of the system in each power interval.

[0044] The state estimation module is used to obtain the real-time output power and state estimation value of the electron cyclotron resonance heating system;

[0045] The model selection module is used to determine the target model to be used in the current control cycle based on the real-time output power; the target model is selected from multiple linear system dynamic models established by the piecewise modeling module.

[0046] The optimization solution module is used to construct a quadratic programming problem for model predictive control based on the target model and a preset power reference trajectory, and solve it to obtain the optimal control input. The quadratic programming problem includes an objective function that minimizes the power tracking error, as well as constraints on the control input and system state. The constraints include at least magnitude constraints and rate of change constraints on the control input.

[0047] An execution interface module is used to apply the optimal control input to the electronic cyclotron resonance heating system.

[0048] The above technical solution has the following advantages:

[0049] This invention effectively reduces the impact of system nonlinearity on control accuracy by dividing the operating range of the electron cyclotron resonance heating system into multiple power intervals based on output power and establishing a linear system dynamic model for each interval. This ensures the predictive accuracy of the model across the entire operating domain. Simultaneously, an online adaptive model parameter update mechanism is introduced to compensate for equipment drift and aging, reducing the need for manual tuning and lowering the steady-state error after long-term system operation. During real-time control, the target model is dynamically selected based on the real-time output power, and a quadratic programming problem for model predictive control is constructed using a preset power reference trajectory. This method not only directly incorporates the input amplitude constraints, rate of change constraints, and state constraints of the controlled object into the optimization solution process, thereby maximizing control performance while ensuring safe system operation, but also obtains the optimal control input by solving the quadratic programming problem, achieving rapid response and accurate tracking of output power. This significantly outperforms traditional control strategies in handling multivariable coupling and hard constraints. Attached Figure Description

[0050] The present invention will now be described in detail with reference to specific embodiments and accompanying drawings, wherein:

[0051] Figure 1 A schematic diagram of the system architecture for the real-time control method of electron cyclotron resonance heating power provided in an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of the working interval division of the piecewise linearization model provided in an embodiment of the present invention.

[0053] Figure 3 A functional block diagram of the real-time control system for electron cyclotron resonance heating power provided in an embodiment of the present invention.

[0054] Figure 4 The flowchart of real-time control loop and model switching provided in the embodiments of the present invention.

[0055] Figure 5 The flowchart illustrates the safety interlocking and solution failure rollback control process provided in this embodiment of the invention.

[0056] 101. Power Measurement Module; 102. State Estimation Module; 103. Model Selection Module; 104. MPC Optimization Solution Module; 105. Constraint and Safety Monitoring Module; 106. Backoff Control Module; 107. Execution Interface Module; 108. Electronic Cyclone Heating Power Execution Object; 109. Interlock Signal Input. Detailed Implementation

[0057] The technical solutions in the embodiments of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0058] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0059] Example 1

[0060] like Figures 1 to 5 As shown, this embodiment provides a real-time control method for electron cyclotron resonant heating power. This method primarily addresses the problem that traditional PID control strategies struggle to balance regulation speed and safety margin in existing electron cyclotron resonant heating systems under nonlinear, multivariable coupling, and stringent safety constraints. By introducing piecewise linearization modeling and model predictive control mechanisms, this embodiment can achieve precise tracking of output power within a millisecond timescale while strictly meeting the operational constraints of the physical equipment.

[0061] The real-time control method mainly relies on a real-time control system, which includes a power measurement module 101, a state estimation module 102, a model selection module 103, an MPC optimization solution module 104, a constraint and safety monitoring module 105, a backoff control module 106, an execution interface module 107, and an electronic gyratory heating power execution object 108 as the controlled object.

[0062] The specific execution flow of this method is as follows:

[0063] First, a segmented modeling step is performed. Because the input-output characteristics of the electron cyclotron resonant heating system exhibit significant nonlinearity across a wide power adjustment range, a single linear model is insufficient to maintain high-precision prediction capabilities across the entire operating domain. Therefore, this embodiment first divides the operating range of the electron cyclotron resonant heating system into multiple power intervals based on output power. Specifically, based on the nonlinear transition characteristics of the system at different power levels, the operating range is divided into at least three intervals: a low-power interval, a medium-power interval, and a high-power interval. In this embodiment, the low-power interval is set to 100kW to 300kW, the medium-power interval to 300kW to 600kW, and the high-power interval to 600kW to 1000kW.

[0064] For each of the aforementioned power ranges, its geometric center point was selected as the operating point for system excitation experiments. For example, 200kW, 450kW, and 800kW were selected as the center operating points for the low, medium, and high power ranges, respectively. During the experiment, a composite excitation signal was applied to the system, which included a step signal, a pseudo-random binary sequence (PRBS) signal, and a frequency sweep signal ranging from 0.1Hz to 10Hz. Simultaneously, the system's input and output data were acquired. Based on the acquired data, a linear system dynamic model describing the system's dynamic characteristics within each power range was established using system identification methods or Taylor expansion methods of physical equations. During this process, the goodness-of-fit index was calculated by comparing the model's predicted output with the actual measured data, ensuring that the normalized root mean square error (NRMSE) was not less than 85%.

[0065] In this embodiment, the established linear system dynamic model is a discrete-time state-space model with a sampling period of 10ms, and its mathematical expression is as follows:

[0066]

[0067] in, These represent different power range numbers. For state vectors, For the input vector, For output. Specifically, the state vector. It includes three key physical quantities, namely cathode temperature. Magnet power supply current and gyrotron electron beam current Input vector It includes three control variables: filament power Anode voltage and cathode voltage Output Output power of the gyrotron This method of constructing a state-space model not only establishes a mapping relationship between control input and power output, but also incorporates key safety indicators such as cathode temperature and magnet current into state variables, thus providing a mathematical basis for directly handling these safety constraints in subsequent control algorithms.

[0068] After modeling is completed, the system enters the real-time control loop. This loop executes once every control cycle, i.e., 10ms.

[0069] Step one: Obtain the real-time output power and state estimate of the electron cyclotron resonant heating system. The power measurement module 101 acquires the output power of the electron cyclotron heating power execution object 108 in real time. Simultaneously, the state estimation module 102 uses a Kalman filter algorithm to obtain the state estimate. The Kalman filter algorithm includes a prediction step and an update step. The prediction step uses the currently used target model to predict the state at the next moment, while the update step corrects the prediction result based on the actual measured values. In this way, even when some state variables, such as the electron beam current, are difficult to measure directly with high precision, accurate state estimates can be obtained, thereby improving the quality of feedback control.

[0070] Step two: Determine the target model for the current control cycle based on the real-time output power. The model selection module 103 selects the best-matching linear system dynamic model from the multiple linear system dynamic models established in the piecewise modeling step, based on the current real-time output power. To avoid frequent model switching at the boundaries of adjacent power ranges due to measurement noise or minor fluctuations, this embodiment employs a hysteresis switching strategy at the switching boundaries of adjacent power ranges. Specifically, a switching threshold and a preset hysteresis bandwidth are set. For example, at the switching boundary of 600kW between the medium and high power ranges, a hysteresis bandwidth of ±50kW is set. The target model is updated to the high-power range model only when the real-time output power of the electron cyclotron resonant heating system crosses the switching threshold from low to high and exceeds the upper limit of the hysteresis bandwidth (650kW); conversely, it switches back to the medium-power range model only when the power crosses the switching threshold from high to low and falls below the lower limit of the hysteresis bandwidth (550kW). Otherwise, the currently used target model remains unchanged. This strategy significantly enhances the stability of the control system during operating condition transitions.

[0071] Step 3: Based on the target model and the preset power reference trajectory, a quadratic programming problem for model predictive control is constructed. The MPC optimization solution module 104 receives the state estimates from the state estimation module 102 and the model parameters from the model selection module 103, and combines them with the preset power reference trajectory to construct a quadratic programming problem aimed at minimizing the power tracking error. This quadratic programming problem includes an objective function and constraints on the control input and system state.

[0072] The mathematical expression for the objective function is designed as follows:

[0073]

[0074] In the formula, For the time domain prediction, the value is set to 10 in this embodiment; To control the time domain, the value is set to 3 in this embodiment; For the predicted output power; For reference power trajectory; To control the rate of change of the input; These are soft-constraint slack variables; and These are the output error weight matrix and the input rate of change weight matrix, respectively. The first term of the objective function guarantees that the system can quickly and accurately track the desired power trajectory. The second term limits drastic changes in the control input, thereby protecting the actuator and preventing system oscillations. The third term is used to handle soft constraints, allowing minor constraint violations under extreme conditions to ensure the feasibility of the solution.

[0075] Meanwhile, the constraint and safety monitoring module 105 defines strict constraints, which are transformed into linear inequality constraints in a quadratic programming problem. These constraints include input magnitude constraints, input rate of change constraints, and state constraints.

[0076] Input amplitude constraints limit the absolute range of each control variable. For example, the filament power is limited to 0W to 1200W, the anode voltage is limited to 18kV to 24kV, and the cathode voltage is limited to -50kV to -40kV.

[0077] Input rate-of-change constraints limit the maximum change of each control variable within a single control cycle. For example, the filament power rate of change is limited to within 0.5 W / ms, the anode voltage rate of change to within 0.1 kV / ms, and the cathode voltage rate of change to within 0.2 kV / ms. This limitation is crucial for protecting high-voltage power supplies and expensive gyrotron equipment, preventing the risk of breakdown due to voltage surges.

[0078] State constraints directly limit the safety boundaries of critical physical quantities within the system. For example, they limit the cathode temperature to less than or equal to 1200°C, the magnet power supply current to less than or equal to 90A, and the gyrotron electron beam current to less than or equal to 40A. By directly considering these state constraints during the optimization solution phase, the controller can predict and avoid exceeding these limits in the future, which is more advanced and safer than traditional alarm and shutdown mechanisms that trigger limits.

[0079] In addition, the constraints also include output power constraints. To improve the success rate of solving the problem, this embodiment introduces slack variables into the output power constraints to implement soft constraints, which include upper bound constraints and lower bound constraints.

[0080] The upper limit constraint is expressed as:

[0081]

[0082] The lower bound constraint is expressed as:

[0083]

[0084] in, and These are the preset maximum output power of 995kW and the preset minimum output power of 95kW, respectively. This is the coefficient representing the influence of the input rate of change on the output power. As a fundamental term for output power prediction, its calculation formula is as follows:

[0085]

[0086] In the formula, This is the state estimate. This is the control quantity from the previous moment.

[0087] Step four involves solving a quadratic programming problem to obtain the optimal control input, which is then applied to the electron cyclotron resonant heating system. The MPC optimization solution module 104 uses efficient solvers such as qpOASES to solve the constructed quadratic programming problem. From the resulting series of future control increment sequences, only the first component is selected as the optimal control input for the current moment. This optimal control input is then sent to the electron cyclotron heating power execution object 108 via the execution interface module 107, thereby completing the closed-loop control for one control cycle.

[0088] Through the above implementation process, this embodiment can achieve smooth adjustment of the electronic cyclotron resonance heating system across the entire power range, with steady-state error controlled within 1.5%, and effectively avoids faults such as overvoltage and overcurrent during rapid power rise and fall, significantly improving the system's operating efficiency and safety.

[0089] Example 2

[0090] Based on Example 1, this embodiment further elaborates on the specific implementation scheme of online adaptive update and safe rollback, addressing the model mismatch problem that may occur during the long-term operation of the electron cyclotron resonance heating system and the control safety problem under extreme conditions.

[0091] Because core components such as the gyrotron tube in the electron gyrotron resonance heating device undergo aging effects over time, deviations arise between the actual physical parameters of the system and the initially established linear system dynamic model. To overcome this problem, the method in this embodiment introduces an online adaptive update step.

[0092] First, the system records input and output data in real time to a circular buffer during operation. This circular buffer is designed with a fixed storage capacity, for example, capable of storing system operation data for the most recent 100 seconds. With a sampling period of 10ms, the buffer capacity is set to 10,000 data sets. Data writing employs a circular overwrite method, meaning new data automatically overwrites the oldest written data, ensuring that the buffer always contains data samples reflecting the latest dynamic characteristics of the system.

[0093] Secondly, when the preset update trigger conditions are met, the system uses data in the circular buffer to update the parameters of the linear system's dynamic model using the recursive least squares method. The design of the update trigger conditions comprehensively considers three dimensions: time period, performance indicators, and operating mode. Specifically, these include: timed triggering, i.e., setting to automatically perform parameter updates every 24 hours to cope with slow aging drift; performance degradation triggering, i.e., triggering an update immediately when the deviation between the model's predicted output and the actual measured value is detected to continuously exceed a preset threshold; and operating mode change triggering, i.e., triggering an update when the system switches from one operating mode to another significantly different operating mode. Through the recursive least squares method, the algorithm continuously corrects the element values ​​in the model parameter matrix using newly acquired data, while simultaneously updating the parameter covariance matrix, thereby enabling the control model to track the time-varying characteristics of the system online.

[0094] In addition, to ensure absolute safety in the event of control algorithm failure or system malfunction, the method in this embodiment also includes a safety rollback step.

[0095] After the MPC optimization solution module 104 solves the quadratic programming problem, the system immediately determines whether the solution was successful. Failure may occur if the algorithm fails to converge within the specified time or if the optimization problem itself has no solution under the current constraints. If the determination result is a solution failure or timeout, or if the constraint and safety monitoring module 105 detects that the system has triggered an interlocking signal, the rollback control strategy is immediately activated.

[0096] The rollback control strategy comprises three levels of processing logic. The first level maintains the previously valid control input, suitable for occasional single-time calculation timeouts by the solver. In this case, the output from the previous control cycle is directly reused to maintain system stability. The second level reduces output power at the maximum permissible slope, suitable for consecutive solution failures or minor anomalies. The system smoothly reduces power to zero or a safe low power level at a preset safe slope. The third level switches to a backup control law, suitable for severe faults where the MPC algorithm completely fails. In this case, the system seamlessly switches to a pre-designed backup controller based on rules or simple PID to ensure the system does not lose control. Simultaneously, if a serious hardware interlock signal input is detected, such as water cooling failure or vacuum breach, the above strategies are bypassed, and an emergency shutdown procedure is executed.

[0097] Example 3

[0098] This embodiment provides a real-time control system for electron cyclotron resonance heating power. This system serves as the hardware and software platform for the methods described in Embodiments 1 and 2, aiming to achieve high real-time performance and high reliability for industrial-grade control.

[0099] like Figure 3 As shown, this real-time control system mainly includes a segmented modeling module, a state estimation module, a model selection module, an optimization solution module, and an execution interface module. These modules are physically implemented using a high-performance real-time controller, such as the NI cRIO-9049 platform, and incorporate FPGA technology to meet millisecond-level control cycle requirements.

[0100] The segmented modeling module is mainly used during the system initialization or offline phases to divide the operating range of the electron cyclotron resonance heating system into low-power, medium-power, and high-power regions based on output power. For each power region, a linear system dynamic model describing the system's dynamic characteristics is established. These model parameters are stored in the controller's non-volatile memory for later retrieval.

[0101] The state estimation module corresponds to Figure 3 The state estimation module 102 in the system is responsible for acquiring the real-time output power and state estimates of the electron cyclotron resonant heating system. In terms of hardware, this module acquires voltage, current, and power signals through a high-precision analog input module, such as the NI-9223. In terms of software algorithms, it incorporates a Kalman filter to reconstruct unmeasurable state quantities such as cathode temperature in real time using the acquired noise data.

[0102] The model selection module corresponds to Figure 3The model selection module 103 is used to determine the target model to be used in the current control cycle based on the real-time output power. Internally, this module runs a logic judgment program with hysteresis, which compares the current power value with a preset switching threshold in real time and retrieves the corresponding parameter matrix from the model library established by the segmented modeling module, sending it to the optimization layer.

[0103] The optimization solution module corresponds to Figure 3 The MPC optimization solution module 104 is the core computing unit of the system. Based on the selected target model and the preset power reference trajectory, it constructs a quadratic programming problem for model predictive control. This module utilizes the processor's floating-point computing power to run the qpOASES solver, calculating the optimal control input that minimizes the power tracking error and satisfies all constraints in each control cycle.

[0104] The execution interface module corresponds to Figure 3 The execution interface module 107 is used to apply the calculated optimal control input to the electron cyclotron resonance heating system. In hardware, this module converts the digital control quantity into an analog voltage signal via an analog output module, such as the NI-9269, to drive actuators such as the filament power supply and high-voltage power supply. Simultaneously, this module also integrates a hardware-level limiting protection circuit as a final safety line beyond software control.

[0105] In addition, the system integrates data management functions, recording operational data in real time to local storage or uploading it to a host computer for subsequent fault analysis and offline calibration of model parameters. The real-time control program, written in LabVIEW and running on the RTLinux operating system, ensures that the data flow and calculation tasks of all the above modules are completed strictly within a 10ms deterministic timeframe.

[0106] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0107] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for real-time control of electron cyclotron resonance heating power, characterized in that, Includes the following steps: Segmented modeling steps: Divide the working range of the electron cyclotron resonance heating system into multiple power intervals according to the output power, and for each power interval, establish a linear system dynamic model describing the dynamic characteristics of the system within that interval. Real-time control cycle steps: Perform the following operations within each control cycle: Obtain the real-time output power and state estimate of the electron cyclotron resonance heating system; The target model used in the current control cycle is determined based on the real-time output power; the target model is selected from multiple linear system dynamic models established in the piecewise modeling step. Based on the target model and the preset power reference trajectory, a quadratic programming problem for model predictive control is constructed. The quadratic programming problem includes an objective function that minimizes the power tracking error, as well as constraints on the control input and system state. The constraints include at least the magnitude and rate of change constraints on the control input. The optimal control input is obtained by solving the quadratic programming problem and applied to the electron cyclotron resonance heating system.

2. The method according to claim 1, characterized in that, In the segmented modeling step, the division of the multiple power intervals is determined based on the nonlinear transition characteristics of the electron cyclotron resonance heating system; the multiple power intervals include at least a low power interval, a medium power interval, and a high power interval, and adjacent power intervals are divided by a preset power threshold; the linear system dynamic model is established by conducting system excitation experiments and collecting input and output data at the geometric center point of each power interval.

3. The method according to claim 1, characterized in that, The dynamic model of the linear system is a discrete-time state-space model, and its form is as follows: in, For state vectors, For the input vector, For output; the state vector Includes cathode temperature, magnet power supply current, and gyrotron electron beam current; the input vector Includes filament power, anode voltage, and cathode voltage; the output This is the output power of the gyrotron; , , , These are the system matrix, input matrix, output matrix, and feedforward matrix for the corresponding power ranges.

4. The method according to claim 1, characterized in that, The step of determining the target model used in the current control cycle based on the real-time output power includes: At the switching boundary between adjacent power ranges, a hysteresis switching strategy is used to determine the target model; The specific process of the hysteresis switching strategy is as follows: set a switching threshold and a preset hysteresis bandwidth; when the real-time output power of the electron cyclotron resonance heating system crosses the switching threshold and exceeds the hysteresis bandwidth, update the target model; otherwise, keep the currently used target model.

5. The method according to claim 1, characterized in that, The mathematical expression for the objective function is: in, To predict the time domain, To control the time domain, For the predicted output power, For reference power trajectory, To control the rate of change of the input, For soft-constrained slack variables, and This is the weight matrix. This is the penalty coefficient.

6. The method according to claim 1, characterized in that, The constraints include: Input amplitude constraints: Limit filament power, anode voltage, and cathode voltage to their respective preset allowable ranges; Input rate of change constraint: Limits the rate of change of filament power, anode voltage, and cathode voltage to their respective preset maximum values; State constraints: limit the cathode temperature to be less than or equal to the preset maximum temperature, limit the magnet power supply current to be less than or equal to the preset maximum current, and limit the gyrotron electron beam current to be less than or equal to the preset maximum electron beam current.

7. The method according to claim 6, characterized in that, The constraints also include output power constraints; these output power constraints introduce slack variables to achieve soft constraints, specifically including upper bound constraints and lower bound constraints. The upper limit constraint is: The lower bound constraint is: in, This is a fundamental term for output power prediction. This is the coefficient representing the influence of the input rate of change on the output power. For the slack variable, and These are the preset maximum and minimum output power values, respectively. Indicates the first The rate of change of the control input at any given time.

8. The method according to claim 1, characterized in that, In the real-time control loop step, the state estimate is obtained through a Kalman filter algorithm; the Kalman filter algorithm includes a prediction step that uses the target model to predict the state at the next moment, and an update step that combines the actual measurement value to correct the prediction result.

9. The method according to claim 1, characterized in that, The method also includes an online adaptive update step: The input and output data of the electron cyclotron resonance heating system are recorded in real time to a ring buffer. When the preset update trigger condition is met, the parameters of the linear system dynamic model are updated using the data in the circular buffer and the recursive least squares method. The update triggering conditions include timed triggering, performance degradation triggering, or operation mode change triggering.

10. The method according to claim 1, characterized in that, The method also includes a safety rollback step: After solving the quadratic programming problem, determine whether the solution was successful; If the solution fails or times out, or if a system interlock signal is detected, a rollback control strategy is executed. The fallback control strategy includes maintaining the previously effective control variable, reducing the output power at the maximum permissible slope, or switching to a standby control law.

11. A real-time control system for electron cyclotron resonance heating power, characterized in that, include: The segmented modeling module is used to divide the working range of the electron cyclotron resonance heating system into multiple power intervals according to the output power, and to establish a linear system dynamic model describing the dynamic characteristics of the system in each power interval. The state estimation module is used to obtain the real-time output power and state estimation value of the electron cyclotron resonance heating system; The model selection module is used to determine the target model to be used in the current control cycle based on the real-time output power; the target model is selected from multiple linear system dynamic models established by the piecewise modeling module. The optimization solution module is used to construct a quadratic programming problem for model predictive control based on the target model and a preset power reference trajectory, and solve it to obtain the optimal control input. The quadratic programming problem includes an objective function that minimizes the power tracking error, as well as constraints on the control input and system state. The constraints include at least magnitude constraints and rate of change constraints on the control input. An execution interface module is used to apply the optimal control input to the electronic cyclotron resonance heating system.

Images