Intelligent fuzzy self-tuning PID optimization control method for gas turbine speed

Abstract

The application relates to the technical field of gas turbine control, and discloses an intelligent fuzzy self-tuning PID optimization control method for the rotating speed of a gas turbine, which adopts an intelligent fuzzy self-tuning PID optimization control method for the rotating speed of the gas turbine based on an improved cloud drift optimization algorithm, is used for self-adaptingly setting the control parameters of the fuzzy PID, can improve the fuzzy PID performance, improves the control effect of the gas turbine under complex conditions, and improves the precision and robustness of the gas turbine control. Compared with the existing classical intelligent optimization algorithm, the method has the advantages of fast convergence speed and good optimization effect, effectively improves the precision of the rotating speed control of the gas turbine power turbine, reduces the adjusting time, and reduces the overshoot.

Description

Technical Field

[0001] This invention relates to the field of gas turbine control technology, and discloses an intelligent fuzzy self-tuning PID optimization control method for gas turbine speed. Background Technology

[0002] As a highly complex control object, the application fields of gas turbines are constantly expanding. In traditional applications, the proportional-integral-derivative (PID) control scheme is commonly used for gas turbine speed control, which can achieve relatively ideal control results when the load disturbance is small. However, in application scenarios with large load disturbances, such as in industrial power generation, the conventional control methods are difficult to achieve ideal control results due to the significant increase in load disturbances.

[0003] To address this issue, researchers have proposed a strategy of incorporating fuzzy control algorithms into PID control methods, aiming to improve the dynamic performance of gas turbines under heavy load disturbances. Fuzzy control algorithms are a type of nonlinear control strategy; combining these two control methods to form a fuzzy PID (FLC-PID) controller not only retains the advantages of PID control but also compensates for its shortcomings in nonlinear system control. However, a major drawback of the PID-FLC controller structure is the difficulty in selecting its controller parameters. Furthermore, there is currently no clear and systematic method to guide the selection of controller parameters, and this tuning problem becomes more subtle and difficult as the complexity of gas turbines increases. Therefore, finding a tuning strategy for fuzzy PID control parameters in gas turbines under complex operating conditions is of great significance. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent fuzzy self-tuning PID optimization control method for gas turbine speed, which can improve the performance of fuzzy PID, enhance the control effect of gas turbine under complex conditions, and improve the accuracy and robustness of gas turbine control.

[0005] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows:

[0006] The intelligent fuzzy self-tuning PID optimization control method for gas turbine speed includes:

[0007] Obtain the expected and actual turbine speeds of the gas turbine at the current moment;

[0008] A fuzzy PID controller is constructed, which includes a fuzzy controller and a PID controller. The fuzzy controller takes the error between the expected value of the power turbine speed of the gas turbine and the actual value of the power turbine speed, as well as the rate of change of the error, as input, and outputs the PID parameter increment, which includes the proportional coefficient increment, integral coefficient increment, and derivative coefficient increment.

[0009] Based on the PID parameter setpoints input to the PID controller at the previous moment, the PID parameter increments at the current moment, and the weights of the PID parameter increments at the current moment, a function for the PID parameter setpoints at the current moment is constructed. The PID parameter setpoints include the proportional coefficient setpoints, integral coefficient setpoints, and derivative coefficient setpoints; the weights of the PID parameter increments include the weights of the proportional coefficient increments, integral coefficient increments, and derivative coefficient increments.

[0010] An improved cloud drift optimization algorithm is used to analyze the weight of the PID parameter increment at the current moment, obtain the optimal weight of the PID parameter increment at the current moment, and obtain the PID parameter tuning value at the current moment through the PID parameter tuning value function analysis.

[0011] The PID controller takes the current PID parameter setpoint as input to control the speed of the gas turbine power plant.

[0012] Furthermore, the functional expression for the PID parameter tuning value at the current moment is:

[0013] ;

[0014] in, The scaling factor is the set value for the current moment; This is the setpoint value for the integral coefficient at the current moment; This is the setpoint value for the differential coefficients at the current moment; The scaling factor setting value from the previous moment; This is the setpoint value of the integral coefficient from the previous time step; The value of the differential coefficient set at the previous moment; , , These are the weights of the proportional coefficient increment, integral coefficient increment, and differential coefficient increment at the current time, respectively. This represents the increment of the proportional coefficient at the current moment; This represents the increment of the integral coefficient at the current moment; This represents the increment of the differential coefficient at the current moment.

[0015] Furthermore, the method of analyzing the weights of the PID parameter increments at the current moment using an improved cloud drift optimization algorithm includes:

[0016] Step M1: Determine the weights of the PID parameter increments at the current time. , , Construct optimization variables and define optimization variables. ;

[0017] Step M2: Construct a fitness function based on the control accuracy, overshoot, and settling time of the gas turbine power speed;

[0018] Step M3: Using cloud particles as optimization variables, a population of cloud particles is formed; initialize the settings to obtain the initial population size, maximum number of iterations, optimization dimension, preset convergence fitness value, upper and lower bounds of the search space for the improved cloud drift optimization algorithm, and generate an initial population containing initial cloud particles based on the initial population size, and generate the search agent position for each initial cloud particle based on the upper and lower bounds of the search space.

[0019] Step M4: Calculate the fitness value of all cloud particles in the current population using the fitness function, and sort them in ascending order of fitness value to obtain the cloud particle with the smallest fitness value in the current population.

[0020] Step M5: Determine whether to end the loop. If the fitness value of the cloud particle with the smallest fitness value in the current population is less than or equal to the preset convergence fitness value, or the current iteration count reaches the maximum iteration count, then end the optimization and output the cloud particle with the smallest fitness value in the current population as the global optimal solution; otherwise, determine the preset grouping threshold, divide the cloud particles in the current population cloud particle sequence with an index less than or equal to the preset grouping threshold into the dominant subgroup, and divide the remaining cloud particles into the subordinate subgroup; calculate the weight value of each dimension of each cloud particle in the dominant subgroup and the subordinate subgroup according to the preset load perturbation factor.

[0021] Step M6: Select any cloud particle in the current population as the target cloud particle for position update. Determine the relative difference factor of the target cloud particle based on its fitness value and the minimum fitness value in the current population. Generate a random number for each dimension of the target cloud particle and compare it with the relative difference factor. For dimensions where the random number is less than the relative difference factor, perform a development phase position update and generate a new search agent position for that dimension based on the local development factor and the weight value of that dimension. For dimensions where the random number is greater than or equal to the relative difference factor, perform an exploration phase position update and generate a new search agent position for that dimension based on the global exploration factor. Traverse each cloud particle in the current population and perform position updates to obtain the cloud particles with updated positions.

[0022] Step M7: Randomly reinitialize the cloud particles in the current population after their positions have been updated, and obtain randomly reinitialized cloud particles;

[0023] Step M8: Repeat steps M4 to M7 until the global optimal solution is output, and obtain the optimal weight of the PID parameter increment at the current moment.

[0024] Furthermore, the fitness function is constructed as follows:

[0025] ;

[0026] in: This is the fitness value; , , All are weighted coefficients; , For time variables, For systematic error, To take the absolute value; The overshoot of the gas turbine power turbine speed; This refers to the adjustment time of the turbine speed in a gas turbine power plant.

[0027] Furthermore, the weighting coefficients , , The results are obtained by calculating using the following formulas:

[0028] ;

[0029] ;

[0030] ;

[0031] in: This represents the output power of the gas turbine in its current steady state. The expected output power of the gas turbine. , These are the preset maximum output power and preset minimum output power of the gas turbine, respectively. It is the hyperbolic tangent function.

[0032] Furthermore, through Generate the search agent location for each cloud particle, where, For the first The search agent position of each cloud particle in the initial iteration phase. A randomly generated number between 0 and 1. The upper bound of the search space. This is the lower bound of the search space.

[0033] Furthermore, the preset load disturbance factor is obtained through... Calculated, where, To preset the load disturbance factor, This represents the output power of the gas turbine in its current steady state. The expected output power of the gas turbine. , These are the preset maximum output power and preset minimum output power of the gas turbine, respectively. It is the hyperbolic tangent function.

[0034] Furthermore, the weight value of each dimension of the cloud particles in the dominant and disadvantaged subgroups is calculated using the following formula:

[0035] ;

[0036] in: The first cloud particle in the current population, sorted in ascending order of fitness value. A cloud particle; The first cloud particle in the current population, sorted in ascending order of fitness value. The first cloud particle The weight values ​​of the dimensions; This is the minimum fitness value in the current population; For current cloud particles fitness value; This represents the maximum fitness value in the current population. It is a preset minimum positive number; for Uniform distribution; The number of cloud particles in the current population

[0037] Furthermore, the relative difference factor of the target cloud particles is determined by... ,in, As a relative difference factor, It is the hyperbolic tangent function. To take the absolute value, For the current target cloud particles fitness value, This is the minimum fitness value in the current population.

[0038] Furthermore, the location updates during the development and exploration phases are performed using the following formulas:

[0039] ;

[0040] in: For the first The second iteration The first cloud particle VI's new search agent location; The cloud particle with the lowest global fitness at present. Dimension's search agent location, For local development factors; For the current population The first cloud particle The weight values ​​of the dimensions; For the first In the next iteration, cloud particles are randomly selected from the dominant subgroup, and these cloud particles are... Dimension's search agent location; For the first In the next iteration, a cloud particle is randomly selected from the current population, and the cloud particle is... Dimension's search agent location; As a global exploration factor; For the first The current population in the next iteration. The cloud particle at the _ ... Dimension's search agent location; The relative difference factor of the current target cloud particles; A random number is generated for each dimension of the current target cloud particle. .

[0041] Compared with the prior art, the beneficial effects of this invention are:

[0042] This invention employs an intelligent fuzzy self-tuning PID optimization control method for gas turbine speed based on an improved cloud drift optimization algorithm. This method adaptively tunes the control parameters of the fuzzy PID controller, improving its performance and enhancing the control effectiveness of the gas turbine under complex conditions, thereby increasing the accuracy and robustness of the gas turbine control. Compared to existing classical intelligent optimization algorithms, this invention's method offers faster convergence, better optimization results, and effectively improves the accuracy of gas turbine power turbine speed control, reducing settling time and overshoot. Attached Figure Description

[0043] Figure 1 This is a flowchart of the intelligent fuzzy self-tuning PID optimization control method for gas turbine speed in the embodiment;

[0044] Figure 2 This is a flowchart illustrating the method for analyzing the weight of the PID parameter increment at the current moment using an improved cloud drift optimization algorithm in this embodiment.

[0045] Figure 3 This is a diagram illustrating the optimized structure for gas turbine speed control in the embodiment.

[0046] Figure 4 This is a simplified flowchart of a method for analyzing the weight of the PID parameter increment at the current moment using an improved cloud drift optimization algorithm, as described in this embodiment. Detailed Implementation

[0047] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0048] Example

[0049] See Figures 1 to 4 The intelligent fuzzy self-tuning PID optimization control method for gas turbine speed includes:

[0050] Step S1: Obtain the expected value of the power turbine speed of the gas turbine at the current moment. Actual value of power turbine speed .

[0051] Step S2: Construct a fuzzy PID controller that includes a fuzzy controller and a PID controller. See [link to relevant documentation]. Figure 3 The fuzzy controller uses the expected value of the power turbine speed of the gas turbine. Actual value of power turbine speed error and the rate of change of the error As input, the output is the PID parameter increment, which includes the proportional coefficient increment. Increment of integral coefficient Increment of differential coefficients Among them, input error and the rate of change of error The fuzzy universes are selected as [-3, 3] and [-3, 3] respectively, and the output is... , , The domains of discourse are selected as [0, 2], [0, 2], and [0, 0.5], respectively.

[0052] The input and output are described using seven fuzzy subsets: negative large, negative medium, negative small, zero, positive small, positive medium, and positive large, denoted as {NB, NM, NS, ZO, PS, PM, PB}. The membership functions for the input and output are selected as the leftmost "z"-shaped function, the rightmost "s"-shaped function, and the middle triangular function.

[0053] Step S3: Based on the PID parameter setpoints input to the PID controller at the previous moment, the PID parameter increments at the current moment, and the weights of the current PID parameter increments, construct the PID parameter setpoint function for the current moment. The PID parameters include the proportional gain setpoint. Integral coefficient setting value Differential coefficient setting value The weights of the PID parameter increments include the weights of the proportional coefficient increments. Weight of integral coefficient increment Weights of the differential coefficient increments ;

[0054] Specifically, in this embodiment, the function expression for the PID parameter tuning value at the current moment is:

[0055] ;

[0056] in, The scaling factor is the set value for the current moment; This is the setpoint value for the integral coefficient at the current moment; This is the setpoint value for the differential coefficients at the current moment; The scaling factor setting value from the previous moment; This is the setpoint value of the integral coefficient from the previous time step; The value of the differential coefficient set at the previous moment; , , These are the weights of the proportional coefficient increment, integral coefficient increment, and differential coefficient increment at the current time, respectively. This represents the increment of the proportional coefficient at the current moment; This represents the increment of the integral coefficient at the current moment; This represents the increment of the differential coefficient at the current moment.

[0057] Existing fuzzy PID controllers suffer from drawbacks such as reliance on empirical rules, poor real-time performance and adaptability, and low portability. To address these issues, this invention introduces weights for the PID parameter increments. , , An improved cloud drift optimization algorithm was used to adjust the weights. , , Optimization is performed to improve the performance of fuzzy PID and enhance its control effect under complex conditions.

[0058] Step S4: Apply an improved cloud drift optimization algorithm to the weights of the PID parameter increments at the current time. , , The analysis is performed to obtain the optimal weight for the PID parameter increment at the current moment.

[0059] Specifically, this embodiment combines an improved cloud drift optimization algorithm with fuzzy PID control to achieve real-time tuning of PID parameters, thereby improving control performance. See also... Figure 2 and Figure 4 Methods for analyzing the weights of PID parameter increments at the current moment using an improved cloud drift optimization algorithm include:

[0060] Step M1: Determine the optimization variables as That is, the weight of the PID parameter increment at the current moment. , , Optimization variables .

[0061] Step M2: In the process of controlling the power turbine speed, it is necessary to consider the control accuracy, the overshoot of the power turbine speed, and the settling time. Furthermore, considering the varying degrees of load fluctuations, the requirements for control accuracy, overshoot, and settling time differ significantly. Therefore, this embodiment introduces weighted coefficients into the optimized fitness function. Specifically, based on the control accuracy, overshoot, and settling time of the gas turbine speed, the fitness function is constructed as follows:

[0062] ;

[0063] in: This is the fitness value; , , All are weighted coefficients; , For time variables, For systematic error, To take the absolute value; The overshoot of the gas turbine power turbine speed; This refers to the adjustment time of the turbine speed in a gas turbine power plant.

[0064] The weighting coefficients , , The results are obtained by calculating using the following formulas:

[0065] ;

[0066] ;

[0067] ;

[0068] in: This represents the output power of the gas turbine in its current steady state. The expected output power of the gas turbine. , These are the preset maximum output power and preset minimum output power of the gas turbine, respectively. It is the hyperbolic tangent function.

[0069] Step M3: Optimize variables Cloud particles are represented by a group of cloud particles, and a population of cloud particles is formed. Initialization settings are used to obtain the initial population size. Maximum number of iterations Optimization Dimensions Preset convergence fitness value Upper bound of the search space and limiting the lower bound And based on the initial population size Generate includes The initial population of initial cloud particles, constrained by the upper bound of the search space. and limiting the lower bound Generate the search agent location for each initial cloud particle, i.e., through... Generate the search agent location for each initial cloud particle, where, For the first The search agent position of each cloud particle in the initial iteration phase. This is a randomly generated number between 0 and 1, and the random number is a floating-point number.

[0070] Step M4: Calculate the fitness value of all cloud particles in the current population using the fitness function, and sort them in ascending order of fitness value to obtain the cloud particle with the smallest fitness value in the current population.

[0071] Step M5: Determine whether to end the loop. If the fitness value of the cloud particle with the smallest fitness value in the current population is less than or equal to the preset convergence fitness value. Or the current iteration count has reached the maximum iteration count. If the fitness value is low, the optimization ends, and the cloud particle with the lowest fitness value in the current population is output as the global optimal solution; otherwise, adaptive weight adjustment is performed.

[0072] When performing adaptive weight adjustment, a preset grouping threshold is first determined. Cloud particles in the current population cloud particle sequence whose index is less than or equal to the preset grouping threshold are divided into dominant subgroups, and the remaining cloud particles are divided into inferior subgroups. According to the preset load perturbation factor, the weight value of each dimension of each cloud particle in the dominant subgroup and the inferior subgroup is calculated.

[0073] Specifically, the preset clustering threshold is determined as follows: The cloud particles in the current population whose index is less than or equal to the preset grouping threshold are classified as the dominant subgroup, and the remaining cloud particles are classified as the subgroups of inferior subgroups. That is, all cloud particles in the current population are arranged in ascending order of fitness value to form a cloud particle sequence. Let the end of the sequence with gradually decreasing fitness value be the front direction, and the end with gradually increasing fitness value be the back direction. The cloud particle with the smallest fitness value is the first cloud particle in the sequence, and the cloud particle with the largest fitness value is the last cloud particle in the sequence. Then, starting from the first cloud particle in the sequence, the cloud particles in the front of the sequence are classified as follows: The cloud particles are divided into a dominant subgroup and the remaining particles are divided into a subgroup to distinguish the importance of different candidate solutions. The current population size refers to the number of cloud particles in the current population. In the initial population, .

[0074] Based on the gas turbine's output power in the current steady state, the gas turbine's expected output power, the gas turbine's preset maximum output power, and the gas turbine's preset minimum output power, the preset load disturbance factor is determined, i.e., the preset load disturbance factor. , This represents the output power of the gas turbine in its current steady state. The expected output power of the gas turbine. , These are the preset maximum output power and preset minimum output power of the gas turbine, respectively. It is the hyperbolic tangent function.

[0075] According to the preset load disturbance factor The weight value of each cloud particle in each dimension of the dominant and disadvantaged subgroups is calculated. This embodiment introduces a load perturbation factor. The dominant subgroup is strengthened, while the weaker subgroup is weakened. Candidate solutions with better fitness values ​​have higher weights. The weight of cloud particles is adjusted based on their fitness and load perturbation factor. The weight calculation formula is as follows:

[0076] ;

[0077] in, The first cloud particle in the current population, sorted in ascending order of fitness value. A cloud particle; Indicates the first Each cloud particle belongs to the dominant subgroup; Indicates the first The cloud particle belongs to the inferior subgroup; The first cloud particle in the current population, sorted in ascending order of fitness value. The first cloud particle The weight values ​​of the dimensions; This is the minimum fitness value in the current population; For current cloud particles fitness value; This represents the maximum fitness value in the current population. The value is a preset, extremely small positive number to avoid a denominator of 0; it can be 10. -6 Or 10 -8 ; for Uniform distribution; This represents the current population size, specifically the number of cloud particles in the current population.

[0078] Step M6: Update position based on development and exploration. Select any cloud particle in the current population as the target cloud particle for position update. Determine the relative difference factor of the target cloud particle based on its fitness value and the minimum fitness value in the current population. Generate a random number for each dimension of the target cloud particle and compare it with the relative difference factor. For dimensions where the random number is less than the relative difference factor, perform a development phase position update, generating a new search proxy position for that dimension based on the local development factor and the dimension's weight value. For dimensions where the random number is greater than or equal to the relative difference factor, perform an exploration phase position update, generating a new search proxy position for that dimension based on the global exploration factor. Traverse each cloud particle in the current population and perform position updates to obtain the cloud particles with updated positions.

[0079] Specifically, a cloud particle is randomly selected from the current population as the target cloud particle for position update. Based on the fitness value of the target cloud particle and the minimum fitness value in the current population, the relative difference factor of the target cloud particle is determined, i.e., the relative difference factor of the current target cloud particle. ,in, It is the hyperbolic tangent function. To take the absolute value, For the current target cloud particles fitness value, This is the minimum fitness value in the current population.

[0080] Simultaneously, a random number is generated for each dimension of the target cloud particle. , By comparing the random numbers in each dimension of the target cloud particles Relative difference factor with target cloud particles The location update phase is determined to improve search efficiency and accuracy.

[0081] For the target cloud particle, for the random number Less than the relative difference factor In the dimension described above, a development phase location update is performed, simulating the cloud's aggregation behavior towards a suitable region, i.e., candidate solutions move towards the optimal solution, generating a new search agent location for that dimension. The specific expression is:

[0082] ;

[0083] in: For the first The second iteration The first cloud particle VI's new search agent location; The cloud particle with the lowest global fitness at present. Dimension's search agent location; As a local development factor, ,express obey Uniform distribution It is the inverse function of the hyperbolic tangent function. The maximum number of iterations, For the first The next iteration; For the current population The first cloud particle The weight values ​​of the dimensions; For the first In the next iteration, cloud particles are randomly selected from the dominant subgroup, and these cloud particles are... Dimension's search agent location; For the first In the next iteration, a cloud particle is randomly selected from the current population, and the cloud particle is... The search agent position for the dimension. It should be noted that if the current population is the initial population, then it corresponds to the first iteration. , No. The second iteration corresponds to the 2nd iteration.

[0084] For the target cloud particle, for the random number Greater than or equal to the relative difference factor In the dimension of , the position is updated during the exploration phase, simulating the drift motion of cloud particles under disturbance, that is, the candidate solution moves randomly in the search space, avoiding the algorithm from getting trapped in local optima, and generating a new search agent position in the dimension, specifically expressed as:

[0085] ;

[0086] in, For the first The second iteration The first cloud particle VI's new search agent location; As a global exploration factor, ,express obey Uniform distribution It is the inverse function of the hyperbolic tangent function. The maximum number of iterations, For the first The next iteration; For the first The current population in the next iteration. The cloud particle at the _ ... The location of the search agent for the dimension.

[0087] Iterate through each cloud particle in the current population and update its position to obtain the cloud particles with updated positions.

[0088] Step M7: Randomly reinitialize the cloud particles in the current population after their positions have been updated, and obtain the randomly reinitialized cloud particles.

[0089] Specifically, the random reinitialization strategy is as follows: to simulate the disturbance of clouds by sudden weather, in each iteration, a random number in the range [0, 1] is generated for each cloud particle. If the triggering condition is met This will reinitialize the cloud particle, as shown in the following expression:

[0090] ;

[0091] in: For the current population After the cloud particle is reinitialized, at the _th ... VI's new search agent location; This serves as an upper bound for the search space. This serves as a lower bound for the search space. Let's say it's the probability of perturbation. .

[0092] The cloud particles whose positions have been updated in the current population are randomly reinitialized according to the above random reinitialization strategy to obtain randomly reinitialized cloud particles.

[0093] Step M8: Repeat steps M4 to M7 until the global optimal solution is output, obtaining the optimal weights for the PID parameter increments at the current moment. .

[0094] Step S5: Increment the PID parameters based on the current time. , , And the optimal weight of the PID parameter increment at the current moment. By analyzing the PID parameter tuning function at the current moment, the PID parameter tuning value at the current moment can be obtained. , , ;

[0095] Step S6: The PID controller is tuned to the PID parameters at the current time. , , The input is used to control the speed of the gas turbine.

[0096] This invention proposes a weighted processing method for objective functions that takes into account load changes, thereby achieving a trade-off between the priority and importance of multiple objectives. This enables the algorithm to achieve the optimal solution in terms of accuracy, overshoot, and settling time in gas turbine speed control, which meets actual requirements and enhances the algorithm's adaptability to complex operating conditions.

[0097] This invention proposes an improved cloud drift optimization algorithm. By adding a perturbation factor in the adaptive weight adjustment step, the algorithm's adaptability to load changes is improved, and the convergence speed is accelerated. At the same time, in the exploration phase of the algorithm's position update, the selection of individual cloud particles is adjusted, which improves the algorithm's efficiency while ensuring global exploration.

[0098] This invention proposes an intelligent fuzzy self-tuning PID optimization control method for gas turbine speed based on an improved cloud drift optimization algorithm. This method adaptively tunes the control parameters of the fuzzy PID controller, aiming to improve the accuracy and robustness of gas turbine control. Compared with existing classical intelligent optimization algorithms, this invention's method has faster convergence speed, better optimization effect, effectively improves the accuracy of gas turbine power turbine speed control, reduces settling time, and lowers overshoot.

[0099] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent fuzzy self-tuning PID optimization control method for the rotating speed of a combustion engine, characterized in that, The method comprises the following steps: acquiring a power turbine speed expected value and a power turbine speed actual value of a gas turbine at a current moment; constructing a fuzzy PID controller comprising a fuzzy controller and a PID controller, the fuzzy controller taking an error between the power turbine speed expected value and the power turbine speed actual value of the gas turbine and a change rate of the error as inputs, and outputting a PID parameter increment, the PID parameter increment comprising a proportional coefficient increment, an integral coefficient increment and a differential coefficient increment; constructing a PID parameter setting value function at the current moment according to a PID parameter setting value input into the PID controller at a previous moment, the PID parameter increment at the current moment and a weight of the PID parameter increment at the current moment, the PID parameter setting value comprising a proportional coefficient setting value, an integral coefficient setting value and a differential coefficient setting value; the weight of the PID parameter increment comprises a weight of the proportional coefficient increment, a weight of the integral coefficient increment and a weight of the differential coefficient increment; analyzing the weight of the PID parameter increment at the current moment by using an improved cloud drift optimization algorithm, acquiring an optimal weight of the PID parameter increment at the current moment, and obtaining the PID parameter setting value at the current moment through analysis of the PID parameter setting value function at the current moment; the PID controller takes the PID parameter setting value at the current moment as input to control the power turbine speed of the gas turbine; wherein the method for analyzing the weight of the PID parameter increment at the current moment by using the improved cloud drift optimization algorithm comprises the following steps: Step M1 : determining the weight of the PID parameter increment from the current time instant , , constitute the optimization variables and define the optimization variables ; step M2: constructing a fitness function according to a control accuracy, an overshoot and a regulation time of the power turbine speed of the gas turbine; the fitness function is constructed as follows: ； wherein: is a fitness value; , , are weighting factors; , is a time variable, is a system error, is an absolute value; is an overshoot of the gas turbine power turbine speed; is a regulation time of the gas turbine power turbine speed; the weighting coefficients , , are calculated by the following equations, respectively: ； ； ； wherein: is an output power of the combustion engine at a current steady state, is a desired output power of the combustion engine, , are a preset maximum output power and a preset minimum output power of the combustion engine, respectively, is a hyperbolic tangent function; step M3: taking an optimization variable as a cloud particle, and a plurality of cloud particles forming a population; initializing to obtain an initial population number, a maximum iteration number, an optimization dimension, a preset convergence fitness value, a limited upper bound and a limited lower bound of a search space of the improved cloud drift optimization algorithm, and generating an initial population containing initial cloud particles according to the initial population number, and generating a search agent position of each initial cloud particle according to the limited upper bound and the limited lower bound of the search space; step M4: calculating fitness values of all cloud particles in a current population by using the fitness function, and arranging the cloud particles in an ascending order of the fitness values to obtain a cloud particle sequence, and obtaining a cloud particle with the minimum fitness value in the current population; step M5: determining whether to end the loop, if the fitness value of the cloud particle with the minimum fitness value in the current population is less than or equal to the preset convergence fitness value, or the current iteration number reaches the maximum iteration number, then ending the optimization, and outputting the cloud particle with the minimum fitness value in the current population as a global optimal solution; otherwise, determining a preset population threshold, dividing cloud particles with a serial number less than or equal to the preset population threshold in the cloud particle sequence in the current population into a dominant subgroup, and dividing the remaining cloud particles into a subordinate subgroup; and calculating a weight value of each dimension of each cloud particle in the dominant subgroup and the subordinate subgroup according to a preset load disturbance factor. Step M6: optionally selecting one cloud particle in the current population as a target cloud particle for position updating, determining a relative difference factor of the target cloud particle according to a fitness value of the target cloud particle and a minimum fitness value in the current population; generating a random number for each dimension of the target cloud particle and comparing the random number with the relative difference factor, for the dimension with the random number smaller than the relative difference factor, performing a development stage position updating, generating a new search agent position of the dimension according to a local development factor and a weight value of the dimension; for the dimension with the random number greater than or equal to the relative difference factor, performing an exploration stage position updating, generating a new search agent position of the dimension according to a global exploration factor; traversing each cloud particle in the current population and performing position updating to obtain cloud particles after position updating; Step M7: randomly reinitializing the cloud particles after position updating in the current population to obtain cloud particles after random reinitialization; Step M8: repeatedly performing steps M4 to M7 until a global optimal solution is output, and obtaining an optimal weight of the PID parameter increment at the current time.

2. The intelligent fuzzy self-turning PID optimization control method for the rotating speed of the combustion engine according to claim 1, characterized in that, The PID parameter setting value function expression at the current time is: ； wherein, is a proportional coefficient setting value at a current time; is an integral coefficient setting value at a current time; is a differential coefficient setting value at a current time; is a proportional coefficient setting value at a previous time; is an integral coefficient setting value at a previous time; is a differential coefficient setting value at a previous time; , , are a weight of a proportional coefficient increment, a weight of an integral coefficient increment, and a weight of a differential coefficient increment, respectively, at a current time; is a proportional coefficient increment at a current time; is an integral coefficient increment at a current time; is a differential coefficient increment at a current time.

3. The intelligent fuzzy self-turning PID optimization control method for the rotating speed of the combustion engine according to claim 1, characterized in that, By generating a search agent position for each cloud particle, wherein, is a randomly generated random number between 0 and 1, is the search agent position for the first cloud particle in the initial iteration phase, is a randomly generated random number between 0 and 1, is the upper limit of the search space, is the lower limit of the search space.

4. The intelligent fuzzy self-turning PID optimization control method for the rotating speed of the combustion engine according to claim 1, characterized in that, The preset load disturbance factor is obtained by calculation, wherein, is a preset load disturbance factor, is an output power of the gas turbine at a current steady state, is an expected output power of the gas turbine, , respectively, a preset maximum output power and a preset minimum output power of the gas turbine.

5. The intelligent fuzzy self-turning PID optimization control method for the rotating speed of the combustion engine according to claim 4, characterized in that, The weight value of each dimension of the cloud particles in the dominant subpopulation and the inferior subpopulation is obtained by calculation according to the following formula: ； wherein: is the th cloud particle in the current population sorted in ascending order of fitness value; is the th cloud particle in the current population sorted in ascending order of fitness value; is the th weight value of the th cloud particle in the current population sorted in ascending order of fitness value; is the minimum fitness value in the current population; is the fitness value of the current cloud particle is the maximum fitness value in the current population; is the minimum fitness value in the current population; is the fitness value of the current cloud particle is the maximum fitness value in the current population; is a preset minimum positive number; is a uniform distribution of is a uniform distribution of is the number of cloud particles in the current population.

6. The intelligent fuzzy self-turning PID optimization control method for the rotating speed of the combustion engine according to claim 5, characterized in that, The relative difference factor of the target cloud particle is determined by wherein, is the relative difference factor, is the absolute value.

7. The intelligent fuzzy self-turning PID optimization control method of the rotating speed of the combustion engine according to claim 6, characterized in that, The development stage position updating and the exploration stage position updating are performed by position updating according to the following formula: ； wherein: is the number of iterations, is the number of dimensions, is the number of iterations, is the new search agent position in the d-th dimension of the i-th cloud particle, is the search agent position in the d-th dimension of the current global minimum cloud particle, is the local development factor, is the weight value in the d-th dimension of the i-th cloud particle of the current population, is the cloud particle randomly selected in the elite subpopulation in the i-th iteration, said cloud particle having a search agent position in the d-th dimension, is the cloud particle randomly selected in the current population in the i-th iteration, said cloud particle having a search agent position in the d-th dimension, is the global exploration factor, is the search agent position in the d-th dimension of the i-th cloud particle of the current population in the i-th iteration, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle, is a random number generated for each dimension of the current target cloud particle.

Images