Gas turbine adaptive PID control method for load rejection condition

Abstract

The embodiment of the application provides a gas turbine adaptive PID control method for load rejection conditions, which relates to the technical field of gas turbine control and comprises the following steps: determining gain optimization variables based on a gas turbine model, constructing an optimization objective function and determining input parameters; iteratively processing the gain optimization variables through an intelligent optimization algorithm, constructing an initialization wolf group of the intelligent optimization algorithm through a chaotic mapping method based on a search space, a wolf group size and an optimization dimension, calculating an optimal solution of the gain optimization variables based on the initialization wolf group, ending the iteration and outputting the optimal solution if the number of iterations reaches a maximum number of iterations, and continuing the iteration until the maximum number of iterations is reached if the maximum number of iterations is not reached; and taking the optimal solution of the gain optimization variables as parameters of a cascade PID controller and controlling the gas turbine in the load rejection condition. The scheme enables the gas turbine to rapidly adjust parameters and maintain stable output in the load rejection condition.

Description

Technical Field

[0001] This invention relates to the field of gas turbine control technology, and in particular to an adaptive PID control method for gas turbines under load shedding conditions. Background Technology

[0002] Gas turbines are the core power units of modern energy conversion and propulsion systems, occupying a crucial position in power generation and industrial drives. Their control systems, acting as the "brain" of the unit, regulate key parameters such as main fuel flow, inlet guide vane angle, and exhaust valve opening in real time, enabling the gas turbine to maintain optimal operating performance under load shedding conditions. This optimizes performance indicators such as output power, efficiency, and emissions. However, gas turbines exhibit strong nonlinearity, significant load disturbances, and drastic time-varying parameters. Existing control strategies are prone to problems such as speed overshoot, exhaust temperature exceeding limits, and sudden reduction in surge margin during dynamic process regulation, making it difficult to meet the stability and adaptability requirements under load shedding conditions and limiting further improvements in gas turbine performance.

[0003] To address this challenge, various improved PID control algorithms have been systematically introduced and developed to achieve precise regulation and optimized control of the gas turbine's operating state. Some scholars have proposed a PID controller design method based on a zero-pole placement strategy, introducing an online augmented matrix identification algorithm to estimate the gas turbine model state in real time, thereby achieving adaptive updating of the PID gain parameters and continuous optimization of closed-loop performance. Other scholars have incorporated an active disturbance rejection control architecture into the gas turbine control system, constructing an extended state observer to achieve real-time estimation of total load disturbances, followed by dynamic compensation through a feedforward loop, introducing a nonlinear combination of errors to suppress deviations, thus improving the system's robustness and regulation quality under various operating conditions. Still others have used genetic algorithms to optimize the PI controller parameter space, obtaining the optimal combination that meets the requirements of robustness and speed, improving the system's control performance. However, some intelligent optimization algorithms require repeated iterations of the gas turbine model to search for the optimal control quantity, leading to a significant increase in computational complexity and excessively long solution times, making it difficult to meet the requirements of real-time control. Meanwhile, while the recently proposed Dhole Optimization Algorithm (DOA) achieves a good balance between exploration and exploitation by simulating the cooperative hunting behavior of a jackal pack, it still faces problems such as local optima stagnation and insufficient convergence speed within a finite number of iterations, similar to traditional metaheuristic algorithms. Therefore, it is necessary to improve the optimization algorithm for the controller to enhance its global search capability and convergence efficiency, thereby better meeting the gas turbine's requirements for optimal performance and real-time operation under load shedding conditions. Summary of the Invention

[0004] In view of this, the embodiments of this specification provide an adaptive PID control method for gas turbines under load shedding conditions. This method introduces the Cauchy inverse cumulative distribution function to generate disturbances, enhancing the ability to escape local optima and enabling the gas turbine to quickly approach the potential optimal region in the global stage, thereby obtaining controller parameters that conform to the current state. The aim is to enable the gas turbine to operate near the optimal efficiency point under load shedding conditions, reducing fuel consumption. Under load shedding conditions, parameters such as fuel quantity and intake air quantity are rapidly adjusted to maintain stable output, ensuring that the gas turbine operates within a safe range and preventing dangerous conditions such as surge, overheating, and overspeed.

[0005] The embodiments in this specification provide the following technical solutions:

[0006] An adaptive PID control method for gas turbines under load shedding conditions includes:

[0007] Based on the gas turbine model, the gain optimization variables are determined, the optimization objective function is constructed, and the input parameters of the intelligent optimization algorithm are determined. These input parameters include the upper bound of the search space. Lower bound of the search space Jackal pack size Optimization Dimensions and maximum number of iterations ;

[0008] The gain optimization variable is iterated through an intelligent optimization algorithm, based on an upper bound. and the lower realm Determine the search space based on the search space and the size of the jackal pack. and optimization dimensions An initial jackal swarm for the intelligent optimization algorithm is constructed using a chaotic mapping method. The optimal solution for the gain optimization variable is then calculated based on the initial jackal swarm. If the number of iterations reaches the maximum number of iterations... End the iteration and output the optimal solution for the gain optimization variable. If the maximum number of iterations has not been reached... Continue iterating until the maximum number of iterations is reached. ;

[0009] The optimal solution of the gain optimization variable is used as the parameter of the cascade PID controller, which controls the gas turbine under load shedding condition.

[0010] Furthermore, based on the gas turbine model, the gain optimization variables are determined, and the optimization objective function is constructed, including:

[0011] The outer-loop PID controller and the inner-loop PID controller are used as controllers for the gas turbine model;

[0012] Obtain the output of the outer-loop PID controller, where, , For the output of the outer loop controller, For the proportional gain of the outer loop PID controller, For the integral of the outer-loop PID controller, The derivative coefficients of the outer-loop PID controller are... To control the error of the system, The rate of change of the system error;

[0013] Obtain the output of the inner-loop PID controller, where, , For the output of the inner loop controller, For the proportional gain of the inner loop PID controller, For the integral of the inner loop PID controller, For the inner-loop PID controller, (These are the differential coefficients.) To control the error of the system, The rate of change of the system error;

[0014] based on , , , , and Constructing gain optimization variables ;

[0015] Construct the optimization objective function ,in, For time variables, For instantaneous system error, t This is the current time scale. J For fitness, d It is a differential.

[0016] Furthermore, based on the search space and jackal population size and optimization dimensions The initial jackal swarm for intelligent optimization algorithms is constructed using a chaotic mapping method, including:

[0017] Based on the current chaos value The chaos value of the next individual after the update is calculated. ,in, , From 0 to The sequence value;

[0018] Based on the current chaos value Upper bound of the search space and the lower realm The initial number was calculated. The first jackal Values ​​of optimization variables By optimizing variable values Construct an initial jackal swarm for the intelligent optimization algorithm, wherein, , From 0 to The sequence value.

[0019] Furthermore, the optimal solution for calculating the gain optimization variables based on the initialized jackal population includes:

[0020] Determine the number of members in a jackal pack. Potential prey locations of jackals ;

[0021] The sound signal strength is randomly generated, where the sound signal strength is a random number between 0 and 1;

[0022] If the intensity of the vocal signal is less than the vocal threshold, and the number of members in the jackal pack is... If the number of members is less than the threshold, the jackal pack's position will shift from the previous iteration's position to the potential prey's position. Direction updated;

[0023] If the intensity of the vocal signal is less than the vocal threshold, and the number of members in the jackal pack is... If the number of members is greater than or equal to a threshold, the jackal pack starts from the position of the previous iteration and surrounds the potential prey's location. ;

[0024] If the intensity of the emitted signal is greater than or equal to the emission threshold, the optimal attack strategy is determined, and the optimal solution for the gain optimization variable is calculated using the optimal attack strategy.

[0025] Furthermore, the number of members in the jackal pack was determined. Potential prey locations of jackals ,include:

[0026] Calculate the number of members in a jackal pack. ,in, , A random number between 0 and 1. This is a rounding function;

[0027] By the number of members in a jackal pack The potential prey locations of the jackal pack were calculated. ,in, , This represents the optimal position of the jackal pack in the current iteration. This is the optimal position recorded by the jackal pack during the historical search process.

[0028] Furthermore, the location of the jackal pack is shifted from the position of the previous iteration to the location of the potential prey. Direction updates, including:

[0029] According to the t Optimization variable values ​​in the next iteration and potential prey locations Update # Optimization variable values ​​in the next iteration ,in, , For the first The first jackal The optimization variable is at the _th ... t Optimization of variable values ​​in the next iteration For the updated version Optimization of variable values ​​in the next iteration The decreasing coefficient, prey j For the first Potential prey locations for each optimization variable.

[0030] Furthermore, the jackal pack starts from the position of the previous iteration and surrounds the potential prey location. ,include:

[0031] Through the first t Optimization variable values ​​in the next iteration and potential prey locations Update # Optimization variable values ​​in the next iteration ,in, , For the first The first jackal The optimization variable is at the _th ... t Optimization of variable values ​​in the next iteration For the updated version Optimization of variable values ​​in the next iteration For random individual indexes, prey j For the first Potential prey locations for each optimization variable.

[0032] Furthermore, the optimal attack strategy is determined, and the optimal solution for the gain optimization variables is calculated using the optimal attack strategy, including:

[0033] The fitness of the jackal was calculated by optimizing the objective function. ;

[0034] The fitness of the prey is calculated by optimizing the objective function. ;

[0035] Measure the size of the prey ,in, , For prey factor, For the jackal's fitness, For the prey's fitness;

[0036] Calculating the optimal hunting time for a jackal pack ,in, , For the size of prey under different conditions, The optimal number of group members required for a successful capture. This is the hunting efficiency coefficient. Environmental factors;

[0037] If measurement If the size exceeds the threshold, adjust the optimal step size based on the adjusted optimal step size and the best hunting time. The optimal solution for the gain optimization variable is calculated, if the metric is... Less than or equal to the size threshold, based on the optimal hunting time The optimal solution for the gain optimization variable is calculated.

[0038] Furthermore, if measurement If the size exceeds the threshold, adjust the optimal step size based on the adjusted optimal step size and the best hunting time. The optimal solution for the gain optimization variable is calculated, if the metric is... Less than or equal to the size threshold, based on the optimal hunting time The optimal solution for the gain optimization variables is calculated, including:

[0039] If measurement If the size exceeds a threshold, calculate the degree of prey weakening. The degree to which the prey is weakened As the adjusted step size for optimization, among which... This represents the optimal position of the jackal pack in the current iteration.

[0040] Through the first t The optimization variable values ​​in the next iteration and The degree of weakening of the prey Update # Optimization variable values ​​in the next iteration ,in, , For the first The first jackal The optimization variable is at the _th ... t Optimization of variable values ​​in the next iteration For the updated version Optimization of variable values ​​in the next iteration rand It is a random number;

[0041] If measurement Less than or equal to the body size threshold, through the first t Optimization variable values ​​in the next iteration Update # Optimization variable values ​​in the next iteration ,in, , This is the optimal position recorded by the jackal pack during the historical search process.

[0042] Furthermore, it also includes:

[0043] In passing the first t Optimization variable values ​​in the next iteration and the degree of weakening of the prey Update # Optimization variable values ​​in the next iteration At that time, a perturbation value is introduced to correct the degree of prey weakening. Disturbance value ,in, To define the location parameters of the distribution peak, To define the half-width scale parameter at half the maximum value, A random number that is uniformly distributed in the range of 0 to 1.

[0044] Compared with the prior art, the beneficial effects that at least one technical solution adopted in the embodiments of this specification can achieve include at least:

[0045] Based on the improved jackal optimization algorithm, the adaptive decision-making mechanism driven by simulated sound emission and the dynamic group construction strategy effectively coordinate the balance between global exploration and local development during the optimization process, enabling the gas turbine to quickly adjust parameters such as fuel quantity and intake air quantity under load shedding conditions and maintain stable output. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of an adaptive PID control method for gas turbines under load shedding conditions provided by an embodiment of the present invention;

[0048] Figure 2 This is a flowchart of the improved jackal optimization algorithm provided in an embodiment of the present invention. Detailed Implementation

[0049] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0050] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0051] In complex and highly volatile application scenarios, such as industrial power generation, gas turbines often face load shedding conditions, making conventional cascade control methods ineffective. This invention proposes a gas turbine cascade PID control method based on an improved jackal optimization algorithm. It introduces the Cauchy inverse cumulative distribution function to generate disturbances, enhancing the ability to escape local optima and enabling rapid approximation of the potential optimal region in the global phase. This yields controller parameters that conform to the current state, aiming to enable the gas turbine to operate near its optimal efficiency point under load shedding conditions, reducing fuel consumption and achieving high-precision online tuning of controller parameters. Under load shedding conditions, it rapidly adjusts parameters such as fuel quantity and intake air volume to maintain stable output, ensuring the gas turbine operates within a safe range and preventing dangerous conditions such as surge, overheating, and overspeed.

[0052] like Figure 1 As shown, an adaptive PID control method for gas turbines under load shedding conditions includes the following steps:

[0053] Step 1: For the gas turbine model studied in this invention, a cascade control method is adopted. Unlike a simple control system, it is dominated by an outer loop controller. Based on the deviation of the outer loop dominant variable and the disturbance, the set value of the inner loop controller is continuously corrected so that it can better adapt to the dynamic changes of the system and improve the control quality.

[0054] The output of the outer loop PID controller is:

[0055] (1)

[0056] The output of the inner loop PID controller is:

[0057] (2)

[0058] in, For the outer loop controller output, For the inner loop controller output, , , , , and The proportional, integral, and derivative coefficients of the outer and inner loop controllers, and To control the error of the system, and This represents the rate of change of the system error.

[0059] Step 2: Combining the improved jackal optimizer with a cascade PID controller enables self-tuning. During controller operation, the cascade PID controller parameters are automatically adjusted to achieve optimal control performance. The selectable optimization variables are:

[0060] (3)

[0061] Then, minimizing the integral of time-weighted absolute error (ITAE) is taken as the optimization objective function, and its mathematical expression is as follows:

[0062] (4)

[0063] In the formula, For time variables, For instantaneous system error, t This is the current time scale. J For fitness, d The function is the derivative. It applies time weighting to the error throughout the response process, thereby penalizing errors that remain unresolved for extended periods and encouraging rapid and smooth convergence of the system.

[0064] After determining the initial value of the optimization variable (3) through the above steps, it is loaded into the gas turbine control system, and the fitness is calculated using the objective function (4) as the evaluation index, thereby establishing the initial optimal controller parameters.

[0065] Step 3: Design of the improved jackal optimizer using the Cauchy inverse cumulative distribution function, the process is as follows: Figure 2 As shown in the diagram. In this design, the Cauchy inverse cumulative distribution function is introduced after the attack behavior, effectively preventing the optimizer from getting trapped in local optima too early, thereby significantly improving the stability and reliability of the algorithm. The optimization process mainly consists of the following five parts: jackal initialization, determination of the number of jackal members and the location of the prey, search behavior, encirclement behavior, and attack behavior.

[0066] Step 3.1, Jackal pack initialization.

[0067] The initial jackal swarm in the DOA algorithm is randomly generated, which limits early exploration capabilities. Therefore, to ensure good jackal swarm diversity in the early stages of iteration, the IDOA algorithm introduces a Logistic chaotic mapping, allowing the jackal swarm to cover the entire search space as evenly as possible, thereby improving the quality and applicability of the initial solution. Considering the upper bound of the search space... Lower Boundary Jackal pack size and optimization dimensions (Input parameters) yield the following initial positions:

[0068] (5)

[0069] (6)

[0070] In the formula, Indicates the current chaos value. The initial value is set randomly. This represents the chaos value of the next individual after the update. From 0 to sequence values, From 0 to sequence values, For the initial first The first jackal One optimization variable value, This constitutes the initial jackal swarm for the intelligent optimization algorithm.

[0071] Step 3.2: Determining the number of group members and the location of potential prey.

[0072] During hunting, jackal packs often exhibit highly structured and coordinated group behavior, typically consisting of 5 to 20 individuals forming temporary hunting units to employ various strategies to hunt different prey. This grouping mechanism can be simulated by a mathematical model (7), which helps to amplify the amount of search information subsequently, thereby simultaneously achieving global exploration and local development, effectively reducing the risk of getting trapped in local optima while improving convergence speed.

[0073] (7)

[0074] In the formula, The number of members in a jackal pack. A random number between 0 and 1. Ensure that the group size is an integer.

[0075] Meanwhile, the jackal pack needs to integrate the information of the current optimal position of the group with the historical global optimal position to estimate the coordinate information of potential prey, so as to generate a search benchmark that is both exploratory and directional. This cognitive process can be simulated by a mathematical model (8), whose output serves as the prey center for the subsequent encirclement and attack phases, effectively guiding the group to explore within the solution space.

[0076] (8)

[0077] In the formula, This represents the optimal position of the jackal pack in the current iteration. This represents the optimal location recorded by the jackal pack during the historical search process. Location of potential prey.

[0078] Step 3.3, Search behavior (update position in the direction of prey).

[0079] After pinpointing the estimated location of prey, the jackal pack needs to assess the feasibility of the hunt based on its own strength. This depends on the size of the group. The number of members is less than the member number threshold (the member number threshold can be set to 10), and the sound signal strength is... If the prey's position is below the vocalization threshold (which can be set to 0.5), the individual needs to approach the prey in a low posture to confirm its status through close-range perception, thus avoiding losses caused by blind hunting. This behavior can be simulated by a mathematical model (9), which updates the position towards the prey, and helps in subsequent decisions on probing or re-exploring.

[0080] (9)

[0081] In the formula, Indicates the first The first jackal Each optimization variable is in t Optimization of variable values ​​in the next iteration For the updated version Optimization of variable values ​​in the next iteration As a decreasing coefficient, this helps enhance the algorithm's global exploration ability and accelerate convergence speed, where, The maximum number of iterations, prey j For the first Potential prey locations for each optimization variable.

[0082] Step 3.4: The jackal pack performs an encirclement maneuver.

[0083] After identifying potential prey locations and confirming hunting feasibility, the size of the jackal pack... Greater than 10 and the intensity of the sound signal If the prey's vocalization threshold is below the threshold, it will engage in encirclement behavior. During this stage, by coordinating the relative positions of different jackals while maintaining a safe distance, they will gradually reduce the encirclement radius, thereby effectively limiting the prey's activity range and increasing the success rate of capture. This behavior can be simulated by a mathematical model (10) to achieve refined development of the optimal solution.

[0084] (10)

[0085] In the formula, Index for random individuals.

[0086] Step 3.5, the jackal's aggressive behavior.

[0087] When the intensity of the group vocal signal When the threshold is exceeded, the jackal pack transitions to the attack phase. However, before launching an attack, it is necessary to comprehensively assess the comparative relationship between its own advantages and the prey in order to formulate the optimal attack strategy. This decision-making process can be characterized by a mathematical model (11) to ensure that the algorithm approximates the optimal solution in the solution space.

[0088] (11)

[0089] In the formula, A measure of prey size, used to quantify the impact of prey size on hunting difficulty and attack strategies. For prey factor, and These represent the adaptability of the jackal and its prey, respectively.

[0090] Meanwhile, during the hunting process of jackals, the active time of the prey is one of the key variables that determines the success or failure of the hunt. This variable can be quantified by modeling the natural behavior of the prey and simulated by a mathematical model (12), thereby providing a theoretical basis for improving the optimal time for hunting success.

[0091] (12)

[0092] In the formula, This is the best time for jackals to hunt. For the size of prey under different conditions, The optimal number of group members required for a successful capture. This is the hunting efficiency coefficient. Represented as environmental factors, this comprehensively quantifies the impact of external environmental disturbances and external factors on the probability of successful capture.

[0093] If the prey's size If the size threshold is exceeded (the size threshold can be set to 2), it indicates that the prey has strong resistance, and the jackal pack will find it difficult to capture the prey successfully in a single attempt. They need to weaken the prey by alternating attacks (the jackal pack performs a weakening kill). This coordinated weakening strategy can be simulated by mathematical models (13) and (14), in which the cosine function and the sine function are mixed to simulate the exchange technique, generating a periodically shrinking search trajectory in the solution space so that it tends to the optimal solution.

[0094] (13)

[0095] (14)

[0096] In the formula, The degree of weakening of the prey is used to adjust the attack intensity, i.e., the optimal step size.

[0097] Meanwhile, although the attack phase has concentrated the group in the candidate optimal neighborhood, the Cauchy inverse cumulative distribution function is used at this time, and a heavy-tailed perturbation is introduced to efficiently escape local optima (i.e., after the jackal group completes the attack, a perturbation operation is added to prevent it from falling into a local optimum). The perturbation can be expressed by the mathematical formula (16), which compares the perturbation value with the degree of weakening of the prey. Multiply.

[0098] (16)

[0099] In the formula, For disturbance values, To define the location parameters of the distribution peak, To define the half-width scale parameter at half the maximum value, control the "tail thickness". These are random numbers uniformly distributed within the range of 0 to 1. An adaptive scaling factor can be used to adjust the search step size, thereby accelerating the attainment of optimal fitness.

[0100] If the size of the prey Below the critical threshold, it indicates that the prey has weak resistance or has fallen into a weakened state, and the jackal pack can kill the prey instantly without the need for coordinated weakening (the jackal pack performs instantaneous killing behavior). This stage can be represented by a mathematical model (15), which corresponds to its rapid local depth development in the solution space, converging towards the optimal solution.

[0101] (15)

[0102] To verify the effectiveness of the proposed improved jackal optimizer in a cascaded PID controller (IDOA-PID), a simulation model of a gas turbine control system was built on the Simulink platform, and the optimization parameter vector was selected as follows:

[0103] ,

[0104] To achieve self-tuning of the gain coefficient, comparative simulation experiments were conducted with the IDOA-PID controller and the controller tuned by the original jackal optimizer (DOA-PID) to comprehensively evaluate its performance.

[0105] During the system's dynamic process, the IDOA-PID algorithm adjusts the system output to the target value within 5 seconds, rapidly and effectively correcting the initial deviation in a very short time, reducing the settling time by approximately 40% compared to DOA-PID. Furthermore, it can promptly terminate unnecessary dynamic adjustments when approaching the target value, keeping the overshoot below 7% and allowing the system to smoothly transition to a steady state. This is significantly superior to the 15% overshoot and 8-second settling time of traditional cascade PID control. Therefore, the system under IDOA-PID control exhibits excellent dynamic performance, effectively suppressing overshoot while ensuring speed.

[0106] Under load shedding conditions, the gas turbine's operating state may change drastically due to the sudden unloading of the load, placing high demands on the performance of the control system. To further verify the performance of the IDOA-PID method under load shedding conditions, the gas turbine, after reaching steady state under normal operating conditions, was suddenly put into load shedding operation. Simulation results show that although the instantaneous speed deviation of the gas turbine reached a peak of 15%, under the improved jackal optimizer, the cascade PID gain parameters were quickly tuned, and the turbine recovered to steady state within 4 seconds, a recovery time approximately 38% shorter than that of DOA-PID. Furthermore, throughout the entire disturbance response process, the gas turbine system under the action of the improved jackal optimizer did not exhibit significant oscillations, demonstrating good robustness.

[0107] Meanwhile, throughout the control process, the IDOA algorithm exhibits a faster convergence speed and a lower ITAE value than the DOA algorithm, which further proves the effectiveness of the Cauchy inverse cumulative distribution perturbation in enhancing global exploration capabilities and avoiding local extrema.

[0108] This invention proposes an Improved Dhole Optimization Algorithm (IDOA), which effectively coordinates the balance between global exploration and local exploitation during the optimization process by simulating a vocalization-driven adaptive decision-making mechanism and a dynamic swarm construction strategy. Furthermore, it employs a Logistic chaotic mapping to improve the initialization quality of the dhole swarm. Further, it introduces a Cauchy inverse cumulative distribution function for perturbation enhancement, improving convergence speed while strengthening search capabilities, thereby enhancing the ability to approach the global optimum.

[0109] A method combining the IDOA metaheuristic optimization algorithm with a cascaded PID controller is used to implement closed-loop control of a gas turbine to verify the feasibility of the proposed method. The inner and outer loop PID gains are selected. As an optimization variable, it is coordinated and tuned using the IDOA algorithm, enabling the controller to adaptively adjust parameters according to the gas turbine operating status, thereby achieving optimal dynamic performance under large load disturbances during load shedding.

[0110] The embodiments of the present invention achieve the following technical effects:

[0111] For the parameter tuning problem of the gas turbine cascade PID controller under load shedding conditions, the introduction of the Cauchy inverse cumulative distribution function in the improved Jackal optimizer provides a heavy-tailed perturbation mechanism for the position update of the population, effectively balancing the contradiction between global exploration and local exploitation. In the global exploration phase, the Cauchy perturbation expands the search range and enhances the diversity of the Jackal population. In the local exploitation phase, the Cauchy perturbation controls its intensity through an adaptive scaling factor, avoiding excessive jumps that could affect convergence accuracy, thus achieving a natural transition from wide-area exploration to fine-grained exploitation. These characteristics enable the optimizer to exhibit stronger adaptability and higher optimization efficiency in handling multi-extremum problems, significantly improving the parameter tuning accuracy of the cascade PID controller under load shedding conditions. By applying the improved Jackal optimizer to the gas turbine cascade PID control system, the system can still maintain good dynamic performance and steady-state accuracy under load shedding conditions, significantly reducing overshoot and settling time, and improving the system's anti-disturbance performance and operational stability. Simulation results show that, compared with traditional cascaded PID controllers and DOA-PID controllers, this algorithm exhibits significant advantages in convergence speed, control accuracy, and robustness, verifying its engineering feasibility and superiority in load shedding environments. Logistic chaotic mapping is used for jackal initialization, leveraging its ergodicity and pseudo-randomness to improve the uniformity of the initial solution coverage. Simultaneously, the Cauchy inverse cumulative distribution function is introduced to increase perturbation and avoid getting trapped in local optima, thus achieving synergistic optimization of exploration and development.

[0112] The above description is merely a specific embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any substitution of equivalent components or equivalent changes and modifications made within the scope of protection of this patent should still fall within the scope of this patent. Furthermore, the technical features, technical features and technical solutions, and technical solutions in this invention can be freely combined and used.

Claims

1. A self-adaptive PID control method for gas turbine in the condition of load rejection, characterized in that, Comprising: Determine a gain optimization variable based on a gas turbine model, construct an optimization objective function, and determine input parameters of an intelligent optimization algorithm, wherein the input parameters include an upper bound of a search space , a lower bound of the search space , a jackal group size , an optimization dimension , and a maximum number of iterations , wherein the intelligent optimization algorithm is an improved jackal optimization algorithm; iterating the gain optimization variable through the intelligent optimization algorithm, determining a search space based on the upper bound and the lower bound , constructing an initial jackal group of the intelligent optimization algorithm through a chaotic mapping method based on the search space, the jackal group size and the optimization dimension , calculating an optimal solution of the gain optimization variable based on the initial jackal group, ending iteration and outputting the optimal solution of the gain optimization variable if the number of iterations reaches the maximum number of iterations , and continuing iteration until the maximum number of iterations is reached if the maximum number of iterations is not reached . calculating the optimal solution of the gain optimization variable based on the initialized prey group, comprising: Determining the number of group members of a jackal group and potential prey locations for a jackal group ; randomly generating a vocalization signal strength, wherein the vocalization signal strength is a random number between 0 and 1; if the vocalization signal intensity is less than a vocalization threshold, and the number of group members of the jackal group is less than a member number threshold, updating the location of the jackal group from the location of the last iteration to the potential prey location direction;​ if the vocalization signal intensity is less than the vocalization threshold and the number of group members of the jackal group is greater than or equal to the member number threshold, causing the jackal group to move from a position of a previous iteration and encircle the potential prey location if the vocalization signal intensity is less than the vocalization threshold and the number of group members of the jackal group is greater than or equal to the member number threshold, causing the jackal group to move from a position of a previous iteration and encircle the potential prey location ; if the vocalization signal strength is greater than or equal to the vocalization threshold, determining an optimal attack strategy, and calculating the optimal solution of the gain optimization variable through the optimal attack strategy; using the optimal solution of the gain optimization variable as the parameters of a cascade PID controller, and controlling the gas turbine in the load rejection condition through the cascade PID controller.

2. The method of claim 1, wherein the PID control method is adapted for a load rejection condition. determining the gain optimization variable based on the gas turbine model, and constructing an optimization objective function, comprising: using the outer loop PID controller and the inner loop PID controller as the controller of the gas turbine model; obtaining an output of an outer loop PID controller, wherein , is an output of an outer loop controller, is a proportional of the outer loop PID controller, is an integral of the outer loop PID controller, is a differential coefficient of the outer loop PID controller, is an error of a control system, is a rate of change of the control system error; obtaining an output of an inner loop PID controller, wherein, , is an output of an inner loop controller, is a proportional of the inner loop PID controller, is an integral of the inner loop PID controller, is a differential coefficient of the inner loop PID controller, is an error of a control system, is a rate of change of the control system error; based on the , the , the , the , the , and the constructing a gain optimization variable ; Constructing an optimization objective function where, is a time variable, is an instantaneous system error, t is a current time scale, J is a fitness, d is a differential.

3. The gas turbine adaptive PID control method for load rejection operating condition according to claim 1, characterized in that, based on the search space, the jackal group size and the optimization dimension , the initialization jackal group of the intelligent optimization algorithm is constructed by a chaotic mapping method, comprising: by the current chaotic value the chaotic value of the next individual after updating is calculated wherein , is a sequence value from 0 to . by the current chaotic value , the upper bound of the search space and the lower bound , the initial first only the first optimization variable value of the jackal , the initialization jackal group of the intelligent optimization algorithm is constructed by the optimization variable value , wherein , is a sequence value from 0 to .

4. The method of claim 1, wherein the PID control method is adapted for a load rejection condition. determining the number of group members of a group of jackals and potential prey locations of the group of jackals comprising: Computing the number of group members of a jackal group wherein, , is a random number between 0 and 1, is a rounding function; by the number of population members of the wolf pack , the potential prey location of the wolf pack is calculated wherein, , is the optimal position of the wolf pack in the current iteration, is the optimal position recorded by the wolf pack in the historical search process.

5. The method of claim 1, wherein the PID control method is adapted for a load rejection condition. causing the position of the jackal population to move from the position of the last iteration towards the potential prey position directional updates, including: According to the first t optimization variable value under the second iteration and the potential prey position of the first optimization variable value under the second iteration optimization variable value under the second iteration wherein, , is the optimization variable value under the second iteration of the first optimization variable value under the second iteration of the first optimization variable value under the second iteration of the first t optimization variable value under the second iteration of the first is the updated optimization variable value under the second iteration of the first optimization variable value under the second iteration of the first is a decrement coefficient, prey j is the potential prey position of the first optimization variable.

6. The method of claim 1, wherein the PID control method is adapted for a load rejection condition. causing the jackals to encircle the potential prey location from the position of the last iteration of the jackals comprising: by the first t optimization variable value at the first iteration and the potential prey location of the optimization variable value at the first iteration , is the optimization variable value of the first optimization variable at the first iteration, t is the optimization variable value of the first optimization variable at the first iteration, is a random individual index, prey j is the potential prey location of the first optimization variable.

7. The method of claim 1, wherein the PID control method is adapted for a load rejection condition. determining an optimal attack strategy, and calculating the optimal solution of the gain optimization variable through the optimal attack strategy, comprising: a fitness of the dhole is calculated by the optimization objective function ; calculating a fitness of the prey by the optimized objective function ; Measures of calculating prey body size wherein, , is a prey factor, is a fitness of a jackal, is a fitness of a prey; Computing optimal hunting time for a jackal pack wherein, , is the size of the prey for different situations, is the optimal number of pack members required for a successful encirclement, is the hunting efficiency coefficient, is the environmental factor; if the metric is greater than a body size threshold, adjusting an optimization step size, calculating an optimal solution for the gain optimization variable based on the adjusted optimization step size and the optimal hunting time , if the metric is less than or equal to the body size threshold, calculating an optimal solution for the gain optimization variable based on the optimal hunting time .

8. The method of claim 7, wherein the PID control method is adapted for a load rejection condition. if the metric is greater than a body size threshold, adjusting an optimization step size, calculating an optimal solution of the gain optimization variable based on the adjusted optimization step size and the optimal hunting time , if the metric is less than or equal to the body size threshold, calculating an optimal solution of the gain optimization variable based on the optimal hunting time , comprises: if the metric greater than a body size threshold, calculating a degree of impairment of the prey the degree of impairment of the prey as an adjusted step size for the optimization, wherein is the optimal position of the jackal swarm in the current iteration; Through the first t The optimization variable values ​​in the next iteration and The degree of weakening of the prey Update # Optimization variable values ​​in the next iteration ,in, , For the first The first jackal The optimization variable is at the _th ... t Optimization of variable values ​​in the next iteration For the updated version Optimization of variable values ​​in the next iteration rand It is a random number; If the metric is less than or equal to the body size threshold, the first t iteration of the optimization variable values is updated by the second iteration of the optimization variable values wherein , is the optimal location recorded by the jackal group during the historical search process.

9. The adaptive PID control method for gas turbines under load shedding conditions as described in claim 8, characterized in that, Further comprising: In the first iteration, the optimization variable values are initialized t the optimization variable values in the first iteration and the degree of weakening of the prey In the first iteration, the optimization variable values are updated the optimization variable values in the first iteration when the degree of weakening of the prey is corrected by introducing a perturbation value the perturbation value wherein, is a location parameter defining the peak of the distribution, is a scale parameter defining the half-width at half-maximum, is a random number uniformly distributed in the range from 0 to 1.

Images