Gas turbine primary frequency modulation optimization method and system

Abstract

The invention provides a gas turbine primary frequency modulation optimization method and system, and the method comprises the steps: carrying out the measurement and estimation of a current state, and obtaining a current state x (k) of a system through a sensor or a state observer; applying a prediction model; wherein the state of future Np steps is predicted based on a system dynamic model (a state space model); rolling optimization is carried out by using an optimizer, an optimal control input sequence {u (k), u (k + 1),..., u (k + N)} is solved in a limited time domain, an objective function is minimized, and input and state constraints are considered at the same time; the control input is applied, the optimized first control quantity u (k) is applied, the subsequent control input is ignored, and the rolling optimization operation is completed; and performing feedback correction, comparing the actual output y (k) with the predicted output, and correcting a prediction error, such as updating an initial state or adjusting model parameters. According to the method, the technical problems of dynamic response delay, insufficient constraint processing and prediction capability deficiency are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of combustion engine control, in particular to a gas turbine primary frequency modulation optimization method and system. BACKGROUND

[0002] The frequency stability of a power system is a core indicator of power quality. According to the International Council for Large Electric Systems (CIGRE), when the frequency fluctuation exceeds ±0.5 Hz, the failure rate of modern power electronic equipment will increase by more than 40%. Gas turbines play an important role in the primary frequency modulation of the power grid due to their fast regulation characteristics (response speed up to 5-10% / min). The traditional gas turbine primary frequency modulation control strategy has significant limitations. Existing methods rely on fixed logic configuration and static parameter settings, making it difficult to adapt to the rapid frequency fluctuations of the power grid caused by high proportion of new energy grid connection. Thermal power units often face response delays (some exceeding 3 seconds), regulation dead zones (±0.033 Hz), and coupling interference with automatic generation control (AGC). In addition, complex constraints such as power compensation amplitude limits and multivariable coupling further reduce the real-time and economic performance of frequency modulation, necessitating dynamic optimization methods to improve regulation accuracy and robustness.

[0003] In recent years, MPC has made significant progress in the study of frequency control in power systems. In the context of wind power frequency modulation, MPC can predict the trend of power grid frequency changes in real time and adjust the output of wind farms, quickly responding to frequency deviation requirements. In the context of wind storage combined frequency modulation systems, MPC coordinates the power output of energy storage and wind power by predicting frequency changes, significantly improving the response speed and accuracy of frequency modulation. For photovoltaic grid-connected frequency modulation scenarios, finite control set model predictive control (FCS-MPC) is widely used in inverter control. In the context of ultra-supercritical coal-fired units, MPC is used to optimize the load coordination control loop to enhance the primary frequency modulation fast response capability. The weighted coefficients are used to balance different control requirements.

[0004] However, the multi-constraint collaborative optimization of model predictive control (MPC) in gas turbine primary frequency modulation remains to be further explored.

[0005] Drawbacks of the prior art: Gas turbine primary frequency modulation refers to the process of adjusting the output of a gas turbine to quickly respond to frequency changes when the frequency of the power grid deviates from the rated value, in order to maintain the stability of the system frequency. The characteristics of gas turbine primary frequency modulation include response speed, regulation accuracy, stability, and energy consumption. Among them, response speed and regulation accuracy are important indicators of frequency modulation effect, while stability and energy consumption are key factors in evaluating the effectiveness of frequency modulation strategies.

[0006] 1. Classical control method Traditional strategies mainly employ PID control and fuzzy logic control: (1) PID control: adjusts the gas turbine output through proportional-integral-derivative links, but suffers from defects such as large overshoot (typical value 15-20%) and poor disturbance rejection; (2) Fuzzy control: based on expert experience rule base, although it can handle nonlinear problems, parameter tuning depends on human experience and lacks real-time performance. Traditional control methods have three significant defects: (1) lag problem; (2) multi-objective conflict; (3) model mismatch risk.

[0007] 2. Major Technical Bottlenecks (1) Dynamic response delay: Actual data of a 330MW gas turbine shows that the frequency recovery time is 12-18 seconds when the load changes stepwise; feedback control based on historical data is difficult to respond to sudden changes in the output of new energy sources in a timely manner.

[0008] (2) Insufficient constraint handling: Physical constraints such as valve opening and thermal stress are often simplified to threshold triggering mechanisms, resulting in loss of control margin; insufficient synergistic optimization of frequency regulation requirements with constraints such as unit economy and equipment life.

[0009] (3) Lack of predictive ability: It is impossible to effectively predict the fluctuation trend of new energy output, and the adjustment action lags behind the frequency change; the fixed parameter control strategy is difficult to adapt to the dynamic characteristics of gas turbine.

[0010] In summary, existing technologies suffer from technical problems such as dynamic response delay, insufficient constraint handling, and lack of predictive capabilities. Summary of the Invention

[0011] The technical problem to be solved by this invention is: how to solve the technical problems of dynamic response delay, insufficient constraint processing and lack of prediction capability in the prior art.

[0012] This invention solves the above-mentioned technical problems by employing the following technical solution: A method for optimizing primary frequency regulation of a gas turbine includes: S1. Perform current state measurement and estimation, and obtain the current system state x(k) through sensors and state observers; S2. Analyze at least two past frequency regulation data of the gas turbine, establish a hybrid prediction model, and predict the future N based on the hybrid prediction model. p The states of each step are x(k+1), x(k+2), ..., x(k+N); S3. Use the optimizer to perform rolling optimization. Within the finite time domain, solve for the optimal control input sequence {u(k), u(k+1), ..., u(k+N)}, set the minimization objective function, and set input and state constraints. Apply the control input, apply the first optimized control quantity u(k) in the optimal control input sequence, ignore subsequent control inputs, complete the rolling optimization operation, and obtain the gas turbine output adjustment strategy. S4, feedback correction is performed, the actual output y(k) is compared with the predicted output, the prediction error is corrected, the gas turbine output adjustment strategy is applied to the control operation, and the prediction model is corrected according to the actual output, and closed-loop control is performed.

[0013] The MPC gas turbine primary frequency modulation strategy provided by the application significantly improves the frequency modulation response speed and control accuracy through the model prediction and rolling optimization mechanism, effectively overcomes the defects of hysteresis and rigid constraints of the traditional strategy, and provides theoretical support and engineering reference for frequency stability of a high-proportion new energy power grid.

[0014] In a more specific technical solution, in S2, a hybrid prediction model is established for the gas turbine by using the following logic:

[0015] In the formula, x t is a state variable; u t is an input variable; y t is an output variable; A is a state transition matrix; B is a control matrix; C is an observation matrix; v t represents process noise; and w t represents observation noise.

[0016] In a more specific technical solution, in S3, the minimization target function includes: tracking error, control amount weighted sum; and the state constraint includes: physical amplitude limiting.

[0017] In a more specific technical solution, in S3, a value function is designed according to the target and requirement of frequency modulation; in each control period, according to the current system frequency and the gas turbine state and the future state predicted by the hybrid prediction model, the value function optimization problem is solved to obtain the gas turbine output adjustment strategy.

[0018] The application performs economic optimization, introduces a fuel cost item into the target function, and realizes Pareto optimization of frequency modulation performance and economy. The core contradiction of Pareto optimization is that increasing the control action (fuel flow change) to improve the frequency modulation performance leads to an increase in fuel cost, and reducing the fuel cost requires suppressing the control action amplitude, which may affect the frequency tracking effect. Through the prediction model error compensation and dynamic weight adjustment strategy, dynamic performance is realized to quickly track the power grid frequency deviation (minimize tracking error) and reduce fuel consumption (minimize fuel cost).

[0019] In a more specific technical solution, the value function is expressed by using the following logic:

[0020] The constraint conditions of the value function, the upper and lower limits of the output, the ramp rate limit and the frequency deviation allowable range are output by using the following logic:

[0021]

[0022]

[0023] In the formula, Δf(k) represents the frequency deviation at time k: (f 实际 -f 额定 ); Q f represents the frequency deviation weight coefficient; Δu k represents the gas turbine output adjustment amount at time k: (P(k)-P(k-1)); Ru represents the control input weight coefficient; P(k) represents the actual output of the gas turbine at time k; P ref represents the reference output of the gas turbine in high efficiency operation; Q p represents the energy consumption optimization weight coefficient; Q term represents the terminal penalty weight coefficient; N p represents the prediction time domain length; ΔP max represents the maximum allowable change rate of the gas turbine output; Δf max represents the maximum allowable frequency deviation threshold.

[0024] In the primary frequency modulation process of the gas turbine, physical constraints are crucial. For example, the output limit of the gas turbine is determined by its design capacity and operation safety requirements, and the rate limit is related to the mechanical and thermal characteristics of the gas turbine, and the valve rate limit (≤3% / s) is directly included in the optimization problem. By processing multiple constraints, the present application ensures that the control strategy meets the performance requirements (such as frequency modulation accuracy) while complying with physical limits, thereby ensuring the safety and reliability of the frequency modulation process.

[0025] In a more specific technical solution, in S3, when the sudden increase of the power grid load causes the frequency to drop sharply, the value function is used to preferentially drive the gas turbine to rapidly increase the output according to the frequency deviation weight coefficient Q f , and the control adjustment amplitude R u is avoided.

[0026] In a more specific technical solution, in S3, when the frequency fluctuates slightly, the energy consumption optimization weight coefficient Q p is used to guide the gas turbine to return to the preset high efficiency point.

[0027] In a more specific technical solution, in S4, the key output parameters of the gas turbine are collected in real time by a sensor to obtain the actual output; the deviation is calculated; the actual output Yactual (k) with the model predicted value Y pred (k) and the error e(k) is obtained:

[0028] The error feedback operation is performed, and the error e(k) is introduced into the subsequent prediction in a weighted form to correct the hybrid prediction model.

[0029] .

[0030] The present application designs a feedforward controller, which introduces accurate wind power / photovoltaic power prediction data. According to the real-time updated wind power / photovoltaic power prediction data, the controller can adjust the operating state of the gas turbine in advance to cope with possible power fluctuations. The use of feedforward disturbance enables the system to respond faster when facing uncertainty, thereby enhancing the robustness of the system.

[0031] The predicted wind power / photovoltaic power fluctuation amount is input into the feedforward channel to generate a gas turbine reference power adjustment amount:

[0032] where Kf is the feedforward gain, and Td is the differential time constant, which is used to match the inertia delay of the gas turbine.

[0033] In a more specific technical solution, in S4, when the hybrid prediction model is in a parameterized form, the parameters of the hybrid prediction model are updated in real time by using the recursive least square method RLS and Kalman filtering.

[0034] In a more specific technical solution, a gas turbine primary frequency modulation optimization system includes: A current state acquisition module is configured to measure and estimate the current state, and acquire the current state x(k) of the system through a sensor and a state observer. A prediction model construction module is configured to analyze past frequency modulation data of the gas turbine for not less than two times, construct a hybrid prediction model, and predict future states x(k+1), x(k+2),..., x(k+N) of the gas turbine based on the hybrid prediction model. p The current state acquisition module is connected to the prediction model construction module. A rolling optimization module is configured to perform rolling optimization by using an optimizer, solve an optimal control input sequence {u(k), u(k+1),..., u(k+N)} in a limited time domain, set a minimum objective function, set input and state constraints, apply the first control amount u(k) in the optimal control input sequence, ignore subsequent control inputs, complete the rolling optimization operation, and obtain a gas turbine output adjustment strategy. The rolling optimization module is connected to the prediction model construction module. A feedback correction module is configured to compare the actual output y(k) with the predicted output, correct the prediction error, apply the gas turbine output adjustment strategy in the control operation, and correct the prediction model according to the actual output to perform closed-loop control. The feedback correction module is connected with the rolling optimization module.

[0035] Compared with the prior art, the present application has the following advantages: The MPC gas turbine primary frequency modulation strategy provided by the present application significantly improves the frequency modulation response speed and control accuracy through the model prediction and rolling optimization mechanism, effectively overcomes the hysteresis and rigid constraint defects of the traditional strategy, and provides theoretical support and engineering reference for the frequency stability of the high-proportion new energy power grid.

[0036] The present application performs economic optimization, introduces a fuel cost term in the objective function, and realizes Pareto optimization of frequency modulation performance and economy. The core contradiction of Pareto optimization is that increasing the control action (fuel flow change) to improve the frequency modulation performance leads to an increase in fuel cost, and reducing the fuel cost requires suppressing the control action amplitude, which may affect the frequency tracking effect. Through the prediction model error compensation and dynamic weight adjustment strategy, dynamic performance is realized to quickly track the power grid frequency deviation (minimum tracking error) and reduce fuel consumption (minimum fuel cost).

[0037] The present application processes multiple constraints, uses MPC to integrate these constraints into the optimization problem, ensures that the control strategy meets the performance requirements (such as frequency modulation accuracy) while complying with physical limitations, and thus guarantees the safety and reliability of the frequency modulation process.

[0038] The present application designs a feedforward controller, introduces accurate wind power / photovoltaic power prediction data, and the controller can adjust the operating state of the gas turbine in advance according to the real-time updated wind power / photovoltaic power prediction data to cope with possible power fluctuations. The use of feedforward disturbance enables the system to respond faster when facing uncertainties, thereby enhancing the robustness of the system.

[0039] The present application solves the technical problems of dynamic response delay, insufficient constraint processing, and lack of prediction ability in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The present application is a MPC rolling time domain optimization framework data flow processing schematic diagram for embodiment 1 of the present application; Figure 2 The present application is a gas turbine primary frequency modulation optimization method basic step schematic diagram for embodiment 1 of the present application; Figure 3 The present application is a MPC rolling time domain optimization framework schematic diagram for embodiment 2 of the present application; Figure 4 A typical scene test effect diagram for embodiment 2 of the present application; Figure 5 A typical scene test time diagram for embodiment 2 of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0042] Embodiment 1 The MPC used in the gas turbine primary frequency modulation optimization method provided by the present application is an algorithm for solving online optimization control problems. According to a relatively accurate system model, system state information, output information and system tracking trajectory setting information are collected at each sampling time, and the MPC controller is used to solve the optimal dynamic behavior of the system output variable in the prediction time domain online. The MPC solves the control signal of a multivariable and multi-constraint system online through rolling optimization, and has the ability to solve problems such as linear or nonlinear system model mismatch and control target change, and gradually develops into an important tool for studying power system stability.

[0043] As shown in Figure 1 , the MPC is based on a rolling time domain optimization framework, which dynamically adjusts the gas turbine output instruction by solving a multi-step prediction control sequence online, thereby realizing the advance compensation of the frequency deviation. The MPC uses a rolling time domain optimization framework, as shown in Figure 2 , the links in the rolling time domain optimization framework include but are not limited to: prediction model, rolling optimization, feedback correction.

[0044] As shown in Figure 2 , a gas turbine primary frequency modulation optimization method includes the following basic steps: S1, current state measurement and estimation, acquiring the current state x(k) of the system through a sensor or a state observer; S2, using a prediction model; wherein, based on the system dynamic model (state space model), the states x(k+1), x(k+2),..., x(k+N) of the future N p steps are predicted; In this embodiment, the system dynamic model can use, for example: a state space model; In the prediction model of the present embodiment, a prediction model of the gas turbine is established according to the dynamic characteristics (such as thermodynamic cycle process, component response characteristics, etc.) and historical operation data (including operation parameters under different loads, frequency response data, etc.) of the gas turbine. This model can predict the changes of the output of the gas turbine and the system frequency in a future period of time (for example, in the next few seconds to tens of seconds). For example, for a certain type of heavy-duty gas turbine, a hybrid prediction model based on physical principles and data driving is established by analyzing the data of the gas turbine in multiple frequency modulation processes.

[0045] Establishing a state space model of the gas turbine:

[0046] Wherein: State variable x t Describing the internal dynamics of the system, including turbine inlet temperature x1, compressor outlet pressure x2, rotor angular velocity x3, combustion chamber fuel residence amount x4, etc. Input variable u t Including fuel valve opening degree u1 (controlling fuel flow), grid frequency deviation u2 (frequency modulation demand signal), inlet guide vane (IGV) opening degree signal u3, etc. Output variable y t Including power generation power y1, exhaust gas temperature y2, NOx emission concentration y3, etc. State transition matrix A: reflecting the coupling relationship between state variables (such as the influence of temperature on pressure); Control matrix B: direct influence of input on state (such as the effect of fuel valve opening degree on combustion chamber fuel amount); Observation matrix C: mapping of state variables to output (such as the relationship between turbine temperature and power generation power); Noise term v t , w t : process noise (such as combustion instability) and observation noise (sensor error).

[0047] S3, using an optimizer to perform rolling optimization to solve the optimal control input sequence {u(k), u(k+1),..., u(k+N)} in a limited time domain, minimize the objective function, and consider input and state constraints; apply the control input, apply the first control quantity u(k) after optimization, ignore the subsequent control input, and complete the rolling optimization operation; In the present embodiment, the minimization objective function can be set to, for example: tracking error and control quantity weighted sum; the state constraint can be, for example: physical amplitude limiting; In the rolling optimization process of the present embodiment, a value function is designed according to the target (such as quickly recovering the system frequency, reducing the frequency deviation) and requirements (such as meeting the operation constraints of the gas turbine, optimizing energy consumption) of frequency modulation.

[0048] In each control cycle (e.g. every few seconds), the optimal gas turbine output adjustment strategy is obtained by solving the optimization problem of the value function, according to the current system frequency and gas turbine state, as well as the predicted future state by the prediction model. Various constraints of the gas turbine, such as maximum output limit, output change rate limit, etc., are considered in this process.

[0049] For example, the value function can be expressed as follows:

[0050] The constraints of the value function, output upper and lower limits, ramp rate limit, and frequency deviation allowed range, are expressed as follows:

[0051]

[0052]

[0053] wherein, Δf(k): frequency deviation (f 实际 -f 额定 ) at time k, unit Hz, reflecting the degree of system frequency deviation from the target value, which needs to converge to zero quickly; Q f : frequency deviation weight coefficient, controlling the punishment degree of the frequency deviation term, the larger the value, the more gentle the output change, but it may reduce the frequency regulation response speed, which needs to be set in combination with the dynamic characteristics of the gas turbine (such as ramp rate); Δu k : gas turbine output adjustment amount (P(k)-P(k-1)) at time k, unit MW, reflecting the change rate of the gas turbine output, which needs to be adjusted smoothly to avoid mechanical wear and thermal stress; Ru: control input weight coefficient, punishing the amplitude of output adjustment, the larger the value, the more gentle the output change, but it may reduce the frequency regulation response speed, which needs to be set in combination with the dynamic characteristics of the gas turbine (such as ramp rate); P(k): actual output of the gas turbine at time k, unit MW, directly related to the operating state of the gas turbine, which needs to meet the output upper and lower limits and the ramp rate constraint; P ref : reference output of the gas turbine in high efficiency operation (such as economic operating point), unit MW, guiding the gas turbine to operate as close to the high efficiency interval as possible during frequency regulation to reduce energy consumption, which needs to be set dynamically according to the heat efficiency curve of the gas turbine; Q p: Energy consumption optimization weight coefficient, penalizes the magnitude of output deviation from the high-efficiency benchmark. The larger the value, the higher the priority of energy consumption optimization, but it may sacrifice part of the frequency modulation performance. It needs to be set comprehensively according to the grid frequency modulation demand and gas turbine economy; Q term : Terminal penalty weight coefficient, ensures that the frequency deviation tends to zero at the end of the prediction time domain, avoids the accumulation of frequency deviation at the end of the prediction due to short-term optimization, and enhances the stability of the closed loop; N p : Prediction time domain length, unit time step, needs to cover the dynamic response time of the gas turbine (usually 10-30 seconds), too long increases the calculation complexity, and too short reduces the foresight; ΔP max : Maximum allowed change rate of gas turbine output, unit MW / s, determined by the mechanical characteristics of the gas turbine, to prevent sudden changes in output from causing equipment damage; Δf max : Maximum allowed frequency deviation threshold (e.g. ±0.2 Hz), unit Hz, set according to grid safety standards, exceeding the threshold triggers emergency control.

[0054] When the grid suddenly increases the load and causes the frequency to drop sharply, the value function prioritizes driving the gas turbine to quickly increase the output (high Q f ), while controlling the adjustment amplitude (R u ) to avoid overshoot; in the case of small frequency fluctuations, the gas turbine is guided to return to the high-efficiency point through Q p , reducing fuel consumption.

[0055] S4, feedback correction, compare actual output y(k) with predicted output, correct prediction error, for example: update initial state or adjust model parameters.

[0056] In the feedback correction operation of the present embodiment, the feedback correction link is the key to ensuring control accuracy and robustness in the model predictive control (MPC) based real-time optimization strategy for gas turbine primary frequency modulation, and its core objectives include: real-time correction of the cumulative error of the prediction model; enhance the adaptability of the system to operating condition fluctuations; improve the stability of the closed-loop control.

[0057] Apply the optimized gas turbine output adjustment strategy to actual control, and modify the prediction model through actual output (such as actual gas turbine output, actual system frequency change value) to achieve closed-loop control. For example, if the actual frequency recovery speed is slower than predicted, then adjust the relevant parameters in the prediction model so that more accurate control strategies can be obtained in the next optimization.

[0058] Here are the specific design and implementation methods for this link: (1) Actual output measurement and deviation calculation Data acquisition: Collect key output parameters of gas turbine (e.g. frequency, power, speed, etc.) in real time through sensors.

[0059] Deviation calculation: Compare actual output Y actual (k) with model predicted value Y pred (k) to get error e(k).

[0060]

[0061] (2) Prediction model correction Error feedback: Introduce error e(k) in a weighted form into subsequent prediction.

[0062]

[0063] Model parameter online update: If the model is in a parameterized form (such as the state-space model used in this paper), the parameters can be updated in real time through Recursive Least Squares (RLS) or Kalman filtering to improve prediction accuracy.

[0064] Example 2 In this example, a simulation platform is constructed; a gas turbine-grid co-simulation model is built based on simulation software, as shown in the framework Figure 3 .

[0065] MPC controller module. Receive grid frequency deviation signal, predict future load changes, optimize gas turbine power output. Define state-space model as prediction model. Set objective function: minimize frequency deviation + gas turbine power regulation cost. Constraints: upper limit of gas turbine output, ramp rate limit.

[0066] Gas turbine dynamic model. Use state equation to simulate gas turbine response (e.g. first-order inertia link: 1 / (Ts+1)). Power regulation system: includes fuel valve control logic and actuator delay model.

[0067] Grid model. Synchronous generator model, implemented using Synchronous Machine module (Simscape Electrical). Load disturbance model, simulate load mutation through Step or Random Number module.

[0068] Frequency calculation module. Calculate frequency deviation based on grid power balance.

[0069] Simulation platform parameter settings are as follows:

[0070] As Figure 4 and Figure 5As shown, in this embodiment, a typical scenario test is performed. The simulation simulates a 10% load surge of the power grid, and the frequency of the power grid decreases. Under the condition of a 10% load surge, the minimum point of the frequency of the power grid reaches 49.2 Hz, and the time for recovery to the initial frequency is 22 s; under the condition of a 10% load surge, the minimum point of the frequency of the power grid is 49.5 Hz, and the recovery time is 14 s (increased by 36%).

[0071] The MPC strategy is verified based on wind power fluctuation data of a provincial power grid in 2023, and a simulation is performed on the condition of wind power output fluctuation ±15% and power grid frequency fluctuation. The MPC controls the frequency deviation within ±0.13 Hz by adjusting the gas turbine output in advance, and reduces the overshoot by 45% compared with the PID control.

[0072] In summary, the MPC gas turbine primary frequency modulation strategy proposed in the application significantly improves the frequency modulation response speed and control accuracy through the model prediction and rolling optimization mechanism, effectively overcomes the defects of hysteresis and rigid constraints of the traditional strategy, and provides theoretical support and engineering reference for the frequency stability of the high-proportion new energy power grid.

[0073] The application performs economic optimization, introduces a fuel cost item into the objective function, and realizes Pareto optimization of the frequency modulation performance and economy. The core contradiction of Pareto optimization is that increasing the control action (fuel flow change) to improve the frequency modulation performance leads to an increase in fuel cost, and reducing the control action amplitude to reduce fuel cost may affect the frequency tracking effect. Through the prediction model error compensation and dynamic weight adjustment strategy, dynamic performance is realized to quickly track the power grid frequency deviation (minimum tracking error) and reduce fuel consumption (minimum fuel cost).

[0074] The application performs multi-constraint processing, and uses the MPC to integrate these constraints into the optimization problem, to ensure that the control strategy meets the performance requirements (such as frequency modulation accuracy) while complying with physical limitations, thereby ensuring the safety and reliability of the frequency modulation process.

[0075] The application designs a feedforward controller, introduces accurate wind power / photovoltaic power prediction data, and the controller can adjust the operating state of the gas turbine in advance according to the real-time updated wind power / photovoltaic power prediction data to cope with possible power fluctuations. The use of feedforward disturbance enables the system to respond faster when facing uncertainties, thereby enhancing the robustness of the system.

[0076] The application solves the technical problems of dynamic response delay, insufficient constraint processing and lack of prediction ability in the prior art.

[0077] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced by equivalent features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for gas turbine primary frequency regulation optimization, characterized in that, The method comprises: S1, performing current state measurement and estimation, acquiring the current state x(k) of the system through sensors and state observers; S2, analyze past no less than 2 frequency modulation data of the gas turbine, establish a hybrid prediction model, predict future N p steps of state x(k+1), x(k+2),..., x(k+N); S3, performing rolling optimization using an optimizer, solving an optimal control input sequence {u(k), u(k+1),..., u(k+N)} in a limited time domain, setting a minimization objective function, setting input and state constraints; applying the control input, applying the first optimized control quantity u(k) in the optimal control input sequence, ignoring the subsequent control input, completing the rolling optimization operation, and obtaining a gas turbine output adjustment strategy; S4, performing feedback correction, comparing the actual output y(k) with the predicted output, and correcting the prediction error; applying the gas turbine output adjustment strategy to the control operation, obtaining and correcting the prediction model according to the actual output, and performing closed-loop control.

2. The method for gas turbine primary frequency regulation optimization according to claim 1, characterized in that, In S2, the hybrid prediction model for the gas turbine is established using the following logic: where x t is the state variable; u t is the input variable; y t is the output variable; A is the state transition matrix; B is the control matrix; C is the observation matrix; v t represents the process noise; and w t represents the observation noise.

3. The method for gas turbine primary frequency regulation optimization according to claim 1, characterized in that, In S3, the minimization objective function includes tracking error and control quantity weighted sum; and the state constraint includes physical amplitude limiting.

4. The method of claim 1, wherein, In S3, a value function is designed according to the target and requirements of frequency modulation; in each control period, the current system frequency and the gas turbine state, and the future state predicted by the hybrid prediction model are used to obtain a gas turbine output adjustment strategy by solving the optimization problem of the value function.

5. The method of claim 4, wherein, The value function is expressed using the following logic: The constraint conditions of the value function, i.e., output upper and lower limits, ramp rate limit, and frequency deviation allowable range, are expressed using the following logic: wherein Δf(k) represents the frequency deviation at time k: (f 实际 -f 额定 ); Q f represents the frequency deviation weight coefficient; Δu k represents the gas turbine output adjustment at time k: (P(k)-P(k-1)); Ru represents the control input weight coefficient; P(k) represents the actual gas turbine output at time k; P ref represents the gas turbine high-efficiency operation reference output; Q p represents the energy consumption optimization weight coefficient; Q term represents the terminal penalty weight coefficient; N p represents the prediction time domain length; ΔP max represents the maximum allowable change rate of the gas turbine output; Δf max represents the maximum allowable frequency deviation threshold.

6. The method of claim 1, wherein, In the S3, when the grid sudden load causes the frequency to drop, the value function is used to drive the gas turbine to increase power output quickly according to the frequency deviation weight coefficient Q f , and the control adjustment range R u is adjusted to avoid overshoot.

7. The method of claim 1, wherein, In the S3, when the frequency fluctuates slightly, the energy consumption optimization weight coefficient Q is used p Direct the gas turbine to return to the preset high efficiency point.

8. The method of claim 1, wherein, In S4, the key output parameters of the gas turbine are collected in real time by sensors to obtain the actual output; Perform bias calculation; actual output Y actual (k) is compared with the model predicted value Y pred (k) to obtain error e(k): Error feedback operation is performed to introduce the error e(k) into subsequent prediction in a weighted manner to correct the hybrid prediction model. 。 9. The method of claim 1, wherein, In S4, when the hybrid prediction model is in a parameterized form, the parameters of the hybrid prediction model are updated in real time by recursive least squares (RLS) and Kalman filtering.

10. A gas turbine primary frequency regulation optimization system characterized by, The system comprises: A current state acquisition module is configured to perform current state measurement and estimation, and acquire the current state x(k) of the system through sensors and state observers; The prediction model construction module is configured to analyze no less than two frequency modulation data of the gas turbine in the past, construct a hybrid prediction model, and predict the future N p steps of the state x(k+1), x(k+2),..., x(k+N), and the current state acquisition module is connected with the prediction model construction module. A rolling optimization module is configured to perform rolling optimization using an optimizer, solve an optimal control input sequence {u(k), u(k+1),..., u(k+N)} in a limited time domain, set a minimization objective function, set input and state constraints, apply the control input, apply the first optimized control quantity u(k) in the optimal control input sequence, ignore the subsequent control input, complete the rolling optimization operation, and obtain a gas turbine output adjustment strategy; the rolling optimization module is connected with the prediction model construction module; A feedback correction module is configured to compare the actual output y(k) with the predicted output, correct the prediction error, apply the gas turbine output adjustment strategy to the control operation, obtain and correct the prediction model according to the actual output, and perform closed-loop control; the feedback correction module is connected with the rolling optimization module.

Images