Data driving signal processing method and system based on improved flower pollination

Abstract

The invention discloses a data driving signal processing method and system based on improved flower pollination, and the method comprises the steps: constructing an input / output model according to a target single-input single-output discrete time nonlinear system, and generating an iterative input model according to the input / output model and a projection algorithm; target parameters of the iteration input model are obtained, iteration is carried out on the target parameters based on an improved flower pollination algorithm, a plurality of updating schemes are adopted to optimize a local pollination process and a global pollination process in the flower pollination algorithm, and the improved flower pollination algorithm is obtained; and when the number of iterations reaches the maximum number of iterations, obtaining an updated target parameter according to the last iteration, and generating a data driving signal processing model based on model-free adaptive control according to the updated target parameter and the iteration input model. According to the method, parameters of the data driving signal processing model based on model-free adaptive control can be optimized, and good performance is achieved.

Description

Technical Field

[0001] The present invention relates to the field of parameter optimization, and in particular to a data-driven signal processing method and system based on improving flower pollination. Background Art

[0002] At present, when the data-driven signal processing method is implemented, whether it is a single-input single-output discrete-time nonlinear system or a multi-input multi-output discrete-time nonlinear system, the setting of the system parameter values ​​is mostly based on experience, a large number of experimental parameters, or a random selection of a fixed value, which has a certain degree of randomness. At the same time, a large number of experimental parameters or manual adjustments require considerable manpower and time costs, which is time-consuming and unintelligent. Due to the different control systems or expected outputs, this type of parameters often change accordingly. The randomness of the parameter setting of the data-driven signal processing system based on Model Free Adaptive Control (MFAC) makes the system tracking performance poor, thereby affecting the corresponding system performance.

[0003] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0004] The main purpose of the present invention is to provide a data-driven signal processing method, system, terminal and computer-readable storage medium based on improved flower pollination, aiming to solve the problem in the prior art that due to the randomness of parameter settings of data-driven signal processing systems based on model-free adaptive control in the face of different control systems or expected outputs, the system tracking performance is poor, thereby affecting the corresponding system performance.

[0005] To achieve the above object, the present invention provides a data-driven signal processing method based on improving flower pollination, and the data-driven signal processing method based on improving flower pollination comprises the following steps:

[0006] According to the target single-input single-output discrete-time nonlinear system, an input-output model is constructed, and according to the input-output model and the projection algorithm, an iterative input model is generated;

[0007] Obtaining target parameters of the iterative input model, and iterating the target parameters based on an improved flower pollination algorithm, wherein a plurality of updating schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm;

[0008] When the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model.

[0009] Optionally, constructing an input-output model according to a target single-input single-output discrete-time nonlinear system, and generating an iterative input model according to the input-output model and a projection algorithm specifically includes:

[0010] Obtain a target single-input single-output discrete-time nonlinear system, and construct an input-output model according to the target single-input single-output discrete-time nonlinear system;

[0011] According to the theorem satisfied by the target single-input single-output discrete-time nonlinear system, the input-output model is converted into an equivalent dynamic linearized data model;

[0012] Generate a criterion function of a pseudo partial derivative in the equivalent dynamic linearization data model according to a projection algorithm, solve an optimal condition according to the criterion function, and generate a pseudo partial derivative formula according to the optimal condition and the criterion function;

[0013] Obtain an expected system output, and generate the iterative input model according to the expected system output and the pseudo partial derivative formula.

[0014] Optionally, the obtaining of the expected system output and generating the iterative input model according to the expected system output and the pseudo partial derivative formula specifically includes:

[0015] Obtaining an expected system output, and generating a criterion function according to the expected system output;

[0016] The equivalent dynamic linearization data model is updated according to the pseudo partial derivative formula to obtain an updated equivalent dynamic linearization data model, the updated equivalent dynamic linearization data model is substituted into the criterion function, and the partial derivative of the input is calculated to obtain the current optimal condition;

[0017] The iterative input model is generated according to the current optimal condition and the pseudo partial derivative formula.

[0018] Optionally, the step of obtaining a target parameter of the iterative input model and iterating the target parameter based on an improved flower pollination algorithm specifically includes:

[0019] Obtaining target parameters of the iterative input model, initializing the number of individuals in the population according to the target parameters, and initializing the switching probability, the maximum number of iterations, and the dimensions of the individuals in the population;

[0020] In each iteration, determine whether the current number of iterations is within the maximum number of iterations. When the current number of iterations is within the maximum number of iterations, generate a random number and determine whether the random number is greater than the switching probability. If it is greater, the global pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. When it is not greater, the local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. The objective function value of the next-moment iterative solution of the current solution of each individual is calculated, and the global optimal solution is updated according to all objective function values.

[0021] Optionally, the step of using the global pollination process to calculate the next-moment iterative solution of the current solution of each individual specifically includes:

[0022] The next-moment iterative solution of the current solution of each individual is calculated using the global pollination process, wherein the global pollination process is expressed as the following formula:

[0023]

[0024] Among them, x i,g+1 represents the next iteration solution of the ith solution, x i,g represents the current position of the i-th solution, ξ is the scaling factor of the guide step, i is a positive integer in the set {1,2,...,N}, r is a random number generated in [0,1], t is the current iteration number, t max is the maximum number of iterations, L represents the pollination intensity, gbest g is the current optimal solution, x r1,g and x r2,g are two solutions randomly selected from the population when the number of iterations is t, v is a random number generated by a uniform distribution in the interval [0,1], r 1 and r 2 It is a randomly generated number between 0 and 1.

[0025] Optionally, the method of using the local pollination process to calculate the next-moment iterative solution of the current solution of each individual specifically includes:

[0026] The local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual, where the local pollination process is expressed as the following formula:

[0027]

[0028] Among them, x m1,g x m2,g x m3,g x m4,g are different solutions randomly selected from the population when the number of iterations is t, ζ, ζ 1 , 2 is the step size scaling factor.

[0029] Optionally, when the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model, specifically including:

[0030] When the number of iterations reaches the maximum number of iterations, the updated target parameters are obtained based on the global optimal solution obtained in the last iteration;

[0031] The updated target parameters are used to replace corresponding parameters in the iterative input model to generate a data-driven signal processing model based on model-free adaptive control.

[0032] In addition, to achieve the above-mentioned purpose, the present invention also provides a data-driven signal processing system based on improved flower pollination, wherein the data-driven signal processing system based on improved flower pollination comprises:

[0033] A model generation module, used to construct an input-output model according to a target single-input single-output discrete-time nonlinear system, and to generate an iterative input model according to the input-output model and a projection algorithm;

[0034] An iteration module, used for obtaining target parameters of the iterative input model, and iterating the target parameters based on an improved flower pollination algorithm, wherein a plurality of update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm;

[0035] The result generation module is used to obtain the updated target parameters according to the last iteration when the number of iterations reaches the maximum number of iterations, and to generate a data-driven signal processing model based on model-free adaptive control according to the updated target parameters and the iterative input model.

[0036] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a data-driven signal processing program based on improved flower pollination stored in the memory and executable on the processor, wherein the data-driven signal processing program based on improved flower pollination, when executed by the processor, implements the steps of the data-driven signal processing method based on improved flower pollination as described above.

[0037] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a data-driven signal processing program based on improved flower pollination, and when the data-driven signal processing program based on improved flower pollination is executed by a processor, the steps of the data-driven signal processing method based on improved flower pollination as described above are implemented.

[0038] In the present invention, an input-output model is constructed according to a target single-input single-output discrete-time nonlinear system, and an iterative input model is generated according to the input-output model and a projection algorithm; the target parameters of the iterative input model are obtained, and the target parameters are iterated based on an improved flower pollination algorithm, wherein multiple update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm; when the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model. The present invention improves the flower pollination algorithm (FPA) on the basis of the FPA, and further proposes an improved flower pollination algorithm with global optimization capability, and applies the improved flower pollination algorithm to the parameter optimization of the data-driven signal processing model based on model-free adaptive control, avoiding the problem that the system performance is affected by random parameters when the system is applied to the wrong parameters, thereby further maintaining the performance of the data-driven signal processing model based on model-free adaptive control in a good state. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flow chart of a preferred embodiment of the data-driven signal processing method for improving flower pollination according to the present invention;

[0040] Figure 2 It is a schematic diagram of system control input of a simulation process of an embodiment of a data-driven signal processing method for improving flower pollination according to the present invention;

[0041] Figure 3 It is a numerical schematic diagram of system pseudo-partial derivatives of a simulation process of an embodiment of the data-driven signal processing method for improving flower pollination according to the present invention;

[0042] Figure 4 It is a schematic diagram of system output tracking performance of a simulation process of an embodiment of the data-driven signal processing method for improving flower pollination according to the present invention;

[0043] Figure 5 is the SISO system output y of the bad parameters in the data-driven signal processing method for improving flower pollination in the present invention. * Tracking performance diagram;

[0044] Figure 6 It is a schematic diagram of the algorithm flow in the data-driven signal processing method for improving flower pollination of the present invention;

[0045] Figure 7 It is a flow chart of a preferred embodiment of the data-driven signal processing method for improving flower pollination according to the present invention;

[0046] Figure 8 It is a schematic diagram of the output tracking performance of the MFAC data-driven signal processing system based on the random selection in the data-driven signal processing method for improving flower pollination of the present invention;

[0047] Fig. 9 It is a schematic diagram comparing the convergence performance of the system after the parameters of the MFAC data-driven signal processing SISO system are optimized by the MFP A and three classic natural heuristic algorithms in the data-driven signal processing method for improving flower pollination of the present invention;

[0048] Fig.10 It is a structural diagram of a preferred embodiment of the data driven signal processing system based on improved flower pollination of the present invention;

[0049] Fig.11 It is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] At present, when the data-driven signal processing method is implemented, whether it is a single-input single-output discrete-time nonlinear system or a multi-input multi-output discrete-time nonlinear system, the setting of the system parameter values ​​is mostly based on experience, a large number of experimental parameters, or a random selection of a fixed value, which has a certain degree of randomness. At the same time, due to the large amount of experimental parameters or manual adjustments, it takes a considerable amount of manpower and time costs, which is time-consuming and unintelligent. Due to the different control systems or expected outputs, this type of parameters often change accordingly. The randomness of the parameter setting of the data-driven signal processing system based on model-free adaptive control greatly affects the system tracking performance, fails to meet the system expectations, and causes the system performance to be effectively guaranteed.

[0052] In response to one or more of the above problems, the present application constructs an input-output model according to a target single-input single-output discrete-time nonlinear system, generates an iterative input model according to the input-output model and a projection algorithm; obtains target parameters of the iterative input model, and iterates the target parameters based on an improved flower pollination algorithm, wherein multiple update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm; when the number of iterations reaches the maximum number of iterations, obtains an updated target parameter according to the last iteration, and generates a data-driven signal processing model based on model-free adaptive control according to the updated target parameter and the iterative input model.

[0053] The data-driven signal processing method based on improved flower pollination described in the preferred embodiment of the present invention is as follows: Figure 1 As shown, the data-driven signal processing method based on improved flower pollination includes the following steps:

[0054] Step S10: construct an input-output model according to the target single-input single-output discrete-time nonlinear system, and generate an iterative input model according to the input-output model and the projection algorithm.

[0055] Specifically, in the present application, for a target single-input single-output (SISO) discrete-time nonlinear system and a data-driven signal processing discrete-time SISO nonlinear system based on MFAC, corresponding input-output models are constructed, and an iterative input model is constructed based on the corresponding input-output models.

[0056] Furthermore, the input-output model is constructed according to the target single-input single-output discrete-time nonlinear system, and the iterative input model is generated according to the input-output model and the projection algorithm, which specifically includes:

[0057] Obtain a target single-input single-output discrete-time nonlinear system, and construct an input-output model according to the target single-input single-output discrete-time nonlinear system;

[0058] According to the theorem satisfied by the target single-input single-output discrete-time nonlinear system, the input-output model is converted into an equivalent dynamic linearized data model;

[0059] Generate a criterion function of a pseudo partial derivative in the equivalent dynamic linearization data model according to a projection algorithm, solve an optimal condition according to the criterion function, and generate a pseudo partial derivative formula according to the optimal condition and the criterion function;

[0060] Obtain an expected system output, and generate the iterative input model according to the expected system output and the pseudo partial derivative formula.

[0061] Specifically, a target single-input single-output discrete-time nonlinear system is obtained, and an input-output model is constructed according to the target single-input single-output discrete-time nonlinear system, wherein the iterative input model is specifically shown in formula (1):

[0062] y(k)=f(y(k-1),...,y(kn p ),u(k),...,u(kn s )); (1)

[0063] Among them, f(.) is an unknown nonlinear function. y(k) and u(k) are the output and input of the system at the kth moment, n p and n s is of unknown order.

[0064] The iterative input model, that is, the dynamic linear representation of is based on the following two assumptions; Assumption 1: The function f(.) is conditioned on the input signals u(k),...,u(kn s ) is continuous; Assumption 2: The nonlinear system (1) satisfies the generalized Lipschitz condition, that is, the Lipschitz continuity condition, which is the following formula (2):

[0065] |y(k)-y(k-1)|≤b|u(k)-u(k-1)|; (2)

[0066] For any k, the following premise is satisfied:

[0067] Δu(k)=u(k)-u(k-1), Δu(k)≠0; (3)

[0068] Here, b is a positive constant.

[0069] For the iterative input model (1) that satisfies Assumptions 1 and 2 at any time k, there must be a pseudo-partial derivative (PPD) parameter so that the input-output model can be converted into the following equivalent dynamic linearized data model, that is, the following formula (4):

[0070]

[0071] in, is the pseudo partial derivative at the kth measurement, and

[0072] Since the pseudo partial derivative in formula (4) is unknown and has a time delay, the traditional projection algorithm and the least squares algorithm cannot track it well. The improved projection algorithm is used to estimate the PPD and obtain the criterion function of the pseudo partial derivative, which is specifically defined as the following formula (5):

[0073]

[0074] Among them, μ>0 is a weight factor, for The estimated value of . Partial derivative, and then solve the optimal condition, expressed as the following formula (6):

[0075]

[0076] Combining (5) and (6), that is, combining the optimal condition and the criterion function, the pseudo partial derivative formula can be generated, which is specifically the following formula (7):

[0077]

[0078] Wherein, η is a constant step size greater than 0. In order to accurately track the pseudo partial derivatives that change with iterations and ensure that formula (4) always holds, the present application also adopts the following reset scheme:

[0079] if

[0080] or |Δu(k)|≤σ or

[0081]

[0082] in, is the initial value of PPD, and σ is a small positive constant.

[0083] Further, the obtaining of the expected system output and generating the iterative input model according to the expected system output and the pseudo partial derivative formula specifically includes:

[0084] Obtaining an expected system output, and generating a criterion function according to the expected system output;

[0085] The equivalent dynamic linearization data model is updated according to the pseudo partial derivative formula to obtain an updated equivalent dynamic linearization data model, the updated equivalent dynamic linearization data model is substituted into the criterion function, and the partial derivative of the input is calculated to obtain the current optimal condition;

[0086] The iterative input model is generated according to the current optimal condition and the pseudo partial derivative formula.

[0087] Specifically, get the given expected system output y * , when you need to find the corresponding input u(k), you need to consider the following criterion function for the input u(k):

[0088] J(u(k))=|y *(k)-y(k)|+λ|u(k)-u(k-1)| 2 ; (9)

[0089] Among them, λ is a weight factor. Then, the equivalent dynamic linearized data model is updated according to the pseudo partial derivative formula, that is, (4) is rewritten as follows to update the equivalent dynamic linearized data model in combination with equation (7):

[0090]

[0091] Substitute the updated equivalent dynamic linearization data model into the criterion function, and find the partial derivative of the input, that is, substitute formula (10) into formula (9), find the partial derivative of u(k) in formula (9), and solve the optimal condition to obtain the current optimal condition, which is the following formula (11):

[0092]

[0093] Furthermore, according to the current optimal condition and the pseudo partial derivative formula, the iterative input of the control system is generated, that is, the iterative input model:

[0094]

[0095] Here, ρ is the step size constant.

[0096] Step S20, obtaining target parameters of the iterative input model, and iterating the target parameters based on the improved flower pollination algorithm, wherein multiple update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm.

[0097] Specifically, the iterative input model includes four parameters, λ, ρ, μ, and η. The present application optimizes the selection of the initial values ​​of these four parameters of the SISO system of the data-driven signal processing algorithm based on MFAC. In one embodiment, the initial conditions of the system input and output are set to: u(1) = 0, u(2) = 0.5, y(1) = y(2) = 1; the corresponding parameter settings are shown in Table 1 below:

[0098] Table 1. Initial values ​​of parameters

[0099] parameter λ η ρ μ Numeric 1.2 2.5 3.2 1.6

[0100] For Table 1, a corresponding target single-input single-output discrete-time nonlinear system is selected as the MFAC data-driven signal processing system. The corresponding target single-input single-output discrete-time nonlinear system is shown in the following formula (13):

[0101]

[0102] Wherein, b(k) is a time-varying parameter, and its expression is shown in the following formula (14):

[0103] b(k)=0.1+0.1round(k / 100); (14)

[0104] The round function is used to round to the nearest integer.

[0105] Consider a type of step signal as the system tracking output for simulation. In this case, the expected signal y * Set as shown in the following formula (15):

[0106]

[0107] The simulation results of the step signal are as follows: Figure 2 , Figure 3 and Figure 4 As shown. Among them, Figure 2 This is the system control input diagram for the simulation process. Figure 3 It is a numerical diagram of pseudo partial derivatives of the system. Figure 4 Output the tracking performance diagram for the system. Figure 4 From the comparison between the system iterative control output and the expected output, it can be seen that the system control output has a large overshoot near the step jump of the step signal, and the tracking performance is relatively general, but the overall tracking performance error is in a relatively small range.

[0108] Taking the signal in formula (15) as an example, a set of different parameters are selected to show the difference in performance. Among them, a set of parameter values ​​different from those in Table 1 is selected to show the impact of randomly selected parameters on the performance of the algorithm.

[0109] Table 2. Random initial values ​​of parameters in formula 15

[0110] parameter λ η ρ μ value 0.8 4.8 4.7 2.4

[0111] According to the parameters in Table 2, the control output tracking performance of the target single-input single-output discrete-time nonlinear system is obtained as follows: Figure 5 shown. Figure 5 The system output tracking effect and Figure 4 The difference is very obvious; as the iteration continues, Figure 5 The control output in the subsequent overshoot continues to increase, and the final iteration result does not track the expected output very well. Therefore, the random selection of parameters may have a great impact on the algorithm tracking performance, making it difficult for the control output to perform well.

[0112] because Figure 4The result in is that the values ​​of parameters λ, ρ, μ, η are selected based on experience, and the optimal unknown parameters λ, ρ, μ, η are not fixed. In formula (7) and formula (12), the selection of the values ​​of unknown parameters λ, ρ, μ, η will affect the corresponding tracking performance. Figure 5 As shown in the figure, the uncertainty of unknown parameters may have adverse effects on the control system design. Although the parameter values ​​can be selected by manual experiment, in actual data-driven signal processing, such an approach is often blind and increases labor costs. In order to improve the system performance of the data-driven signal processing technology based on MFAC, this application uses an improved flower pollination algorithm to optimize the data parameters.

[0113] Furthermore, the step of obtaining the target parameter of the iterative input model and iterating the target parameter based on the improved flower pollination algorithm specifically includes:

[0114] Obtaining target parameters of the iterative input model, initializing the number of individuals in the population according to the target parameters, and initializing the switching probability, the maximum number of iterations, and the dimensions of the individuals in the population;

[0115] In each iteration, determine whether the current number of iterations is within the maximum number of iterations. When the current number of iterations is within the maximum number of iterations, generate a random number and determine whether the random number is greater than the switching probability. If it is greater, the global pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. When it is not greater, the local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. The objective function value of the next-moment iterative solution of the current solution of each individual is calculated, and the global optimal solution is updated according to all objective function values.

[0116] Specifically, the Flower pollination algorithm (FPA) process includes: cross-pollination and self-pollination. Correspondingly, cross-pollination can be defined as global pollination, which is used to explore the area of ​​the search space to find the optimal area; abiotic self-pollination can be described as a local optimal solution that uses the area around the current solution to accelerate convergence; flower constancy can be regarded as a reproduction ratio proportional to the similarity between two flowers; compared with global pollination, local pollination has a slight advantage due to physical and external natural factors. Specifically, local pollination and global pollination are controlled by a control variable p with a value between 0 and 1.

[0117] Correspondingly, the mathematical model steps of global pollination and floral constancy can be expressed as follows (16):

[0118] x i,g+1 =x i,g +ξL(gbest g -x i,g); (16)

[0119] Among them, g represents the current iteration, x i,g is the current position of the ith solution, x i,g+1 represents the next iterative solution of the ith solution, whose dimension is determined by the number of variables contained in the problem; ξ is the scaling factor of the guide step size, gbest g is the best solution, i is a positive integer in the set {1,2,...,N}, L is the intensity of pollination, which is generated according to the Levy distribution with β index, as shown in the following formula (17):

[0120]

[0121] Here, Γ(β) represents the standard gamma distribution.

[0122] The mathematical model of local pollination is shown in the following formula (18):

[0123] x i,g+1 =x i,g +ζ(x m1,g -x m2,g ); (18)

[0124] where m1 and m2 are different integers selected from the set {1,2,...,N}, and x m1,g and x m2,g are two solutions randomly selected from the current solution population, and ζ is a variable uniformly randomly selected from [0,1].

[0125] The pollination behavior is switched globally and locally according to a certain switching probability. The switching probability p is used as the switching probability for random selection. The entire FPA can be expressed as the following formula (19):

[0126]

[0127] Where r is a random number generated in [0,1].

[0128] FPA has the advantages of easy implementation, high optimization efficiency and strong robustness, but it should be pointed out that it also has the defects of weakened optimization performance in the later stage of algorithm iteration and difficulty in jumping out of local extreme values. In view of the defect that FPA is trapped in the local optimal solution, this application further optimizes it to improve its global search ability to avoid being trapped in the local minimum, so as to achieve better optimization results. This paper proposes a novel Modified Flower Pollination Algori thm (MFPA), which establishes an effective mathematical model based on the effective mixing of various update schemes, so that the proposed MFPA can be more comprehensive when looking for solutions to global optimization problems.

[0129] Specifically, in the improved flower pollination algorithm, the global pollination and local pollination around FPA are optimized, and formulas (16) and (18) are modified.

[0130] Furthermore, the global pollination process is used to calculate the next-moment iterative solution of the current solution of each individual, specifically including:

[0131] The next-moment iterative solution of the current solution of each individual is calculated using the global pollination process, wherein the global pollination process is expressed as the following formula:

[0132]

[0133] Among them, x i,g+ 1 represents the next iteration solution of the ith solution, x i,g represents the current position of the i-th solution, ξ is the scaling factor of the guide step, i is a positive integer in the set {1,2,...,N}, r is a random number generated in [0,1], t is the current iteration number, t max is the maximum number of iterations, L represents the pollination intensity, gbest g is the current optimal solution, x r1,g and x r2,g are two solutions randomly selected from the population when the number of iterations is t, v is a random number generated by a uniform distribution in the interval [0,1], r 1 and r 2 It is a randomly generated number between 0 and 1.

[0134] Specifically, in this application, for the global pollination in the flower pollination algorithm, the classic FPA designs a mathematical model for global pollination, which is based on the process of different insects spreading pollen between different plants; based on the reverse update of the current position to the best solution so far gbest g, to spread pollen to a greater distance. Since the main goal of this stage is to spread pollen to far places to pollinate other plants, there is no need to always move the current position in the opposite direction to the best position so far. Therefore, this application considers increasing the population diversity between individuals, and will use multiple schemes for updating to bring pollen to more different areas, so that it can explore more various pollination situations in the search space, thereby avoiding falling into local optimal solutions, which may significantly affect the optimization process. During the optimization process, the global pollination in the improved flower pollination algorithm effectively switches between three different update schemes, allowing pollen to reach multiple areas during the optimization process.

[0135] Among them, the improved flower pollination algorithm contains three update schemes. The first update scheme is to associate the current position with each search agent through the current iteration to help the algorithm gradually explore various areas around the current solution in the search space until the end condition of the iteration is reached. In this case, the optimization process will focus on local search around this current solution, hoping to find a better solution. The specific first update scheme is shown in the following formula (20):

[0136]

[0137] Among them, g represents the current iteration, x i,g is the current position of the ith solution, x i,g+1 It represents the next iteration solution of the ith solution, and its dimension is determined by the number of variables contained in the problem. max is the maximum number of iterations, t is the current number of iterations, L is the distance control factor used to determine the distance around the current position to be detected, also called the intensity of pollination, ξ is the proportional factor of the guide step size, gbest g This is the best solution so far.

[0138] The second update scheme is based on two search steps to search around the current best solution; the first step is to search the optimal solution of the algorithm in reverse, while the other update scheme is committed to improving this direction to make it close to the best solution so far, so as to further improve the algorithm's exploration ability. The specific second update scheme is shown in the following formula (21):

[0139]

[0140] Among them, x r1,g and x r2,g are two solutions randomly selected from the population at iteration number t, and r is a random number generated in the interval [0,1] under uniform distribution.

[0141] The third update scheme is based on uniform distribution, exploring the area between the current optimal solution position and its reverse position to avoid falling into the local optimal value. The specific third update scheme is shown in the following formula (22):

[0142] x i,g+1 =gbest g .v; (22)

[0143] Where v is a random number generated by uniform distribution in the interval [0,1], as shown in formula (23):

[0144] v=U(0,1); (23)

[0145] The specific form of switching between these three update schemes, that is, the global pollination process, is shown in the following formula (24), which balances the implementation of one update scheme with the implementation of the other two update schemes, and then attempts to balance the exploration ability and optimization ability:

[0146]

[0147] Among them, r, r 1 and r 2 It is a randomly generated number between 0 and 1.

[0148] Furthermore, the method of using the local pollination process to calculate the next-moment iterative solution of the current solution of each individual specifically includes:

[0149] The local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual, where the local pollination process is expressed as the following formula:

[0150]

[0151] Among them, x m1,g x m2,g x m3,g x m4,g are different solutions randomly selected from the population when the number of iterations is t, ζ, ζ 1 , 2 is the step size scaling factor.

[0152] Specifically, in modifying the mathematical model of the local pollination process, this application is designed based on two different steps and uses a switching probability of 0.5 to switch so that a balance is achieved between them. The first step is to perform a scaled search around the current position based on the current iteration amount to improve the searchability of the algorithm in the search space and avoid falling into the local minimum. Specifically, formula (18) is modified to obtain the following formula (25):

[0153]

[0154] The second step is to search around the best solution so far and expand it according to the current iteration to improve the efficiency of finding the optimal solution, speed up the convergence speed, and make it move in the right direction close to the optimal solution. The specific formula is as follows:

[0155]

[0156] Combining the two equations, we get the local pollination process. The local pollination process of the improved flower pollination algorithm is expressed as shown in the following formula (27):

[0157]

[0158] Among them, x m1,g x m2,g x m3,g x m4,g are different solutions randomly selected from the population when the number of iterations is t, ζ, ζ 1 , 2 is the step scale factor, and r is a random number in [0,1].

[0159] Correspondingly, after obtaining the improved flower pollination algorithm, the target parameters of the iterative input model are obtained, and the target parameters are iterated based on the improved flower pollination algorithm.

[0160] Specifically, when implementing Figure 6 and Figure 7 As shown, first initialize: N, P, t max ,D; where N is the number of individuals in the population, p is the switching probability, g max is the maximum number of iterations, and D is the dimension of individuals in the population. When the algorithm is running, the above data is initialized first, and then the population x is initialized. Then it is determined whether the current number of iterations is less than the maximum number of iterations, that is, g<g max , when it meets, a random number r is created between 0 and 1. If r>p, the next moment iteration solution of the current solution of each individual is calculated by the global pollination process, that is, formula (24). When it is not greater than, the next moment iteration solution of the current solution of each individual is calculated by the local pollination process, that is, formula (27). Then, the objective function value of the next moment iteration solution of the current solution of each individual is calculated, and the global optimal solution is updated according to all objective function values. Among them, the target parameters of the iterative input model are the four parameters λ, ρ, μ, and η, but are not limited to these four. Other parameters can also be optimized. These parameters are used as the corresponding population, and the number of individuals in the population, the switching probability, the maximum number of iterations, and the dimension of the individuals in the population can be set by themselves. The corresponding switching probability is preferably set to 0.5, and the dimension of the individuals in the population is set to the number of parameters that need to be further optimized, that is, 4.

[0161] In the process of iterating the target parameters based on the improved flower pollination algorithm, each time the global optimal solution is updated, the corresponding iterative input model is updated by using the corresponding parameters and formulas (1) to (12), and then its performance is judged accordingly to select the corresponding global optimal solution.

[0162] Step S30: When the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model.

[0163] Specifically, in the present application, when the number of iterations reaches the maximum number of iterations, the optimized updated target parameters can be obtained, and the corresponding model can be generated by updating the target parameters.

[0164] Furthermore, when the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model, which specifically includes:

[0165] When the number of iterations reaches the maximum number of iterations, the updated target parameters are obtained based on the global optimal solution obtained in the last iteration;

[0166] The updated target parameters are used to replace corresponding parameters in the iterative input model to generate a data-driven signal processing model based on model-free adaptive control.

[0167] Specifically, in the present application, when the number of iterations reaches the maximum number of iterations, the global optimal solution obtained in the last iteration is taken as the output of the corresponding improved flower pollination algorithm. The updated target parameters, that is, the values ​​of the updated parameters λ, ρ, μ, η, are obtained through the output. The updated target parameters are used to replace the corresponding parameters in the original iterative input model, and the obtained model is the optimized data-driven signal processing model based on model-free adaptive control.

[0168] Furthermore, the present application compares the optimization capabilities of the proposed MFPA with PSO, FPA, and GSA for the parameters of the MFAC-based data-driven signal processing system. The present application sets each variable dimension in the population set of MFPA, PSO, FPA, and GSA to the number of parameters that need to be further optimized, that is, the individual dimension of the population is the number of optimized parameters in the MFAC data-driven signal processing SISO system 4; the maximum number of iterative optimization times for optimizing the MFAC data-driven signal processing algorithm based on MFPA, PSO, FPA, and GSA is set to 30 times, that is, to reduce the error, each algorithm performs 30 independent repeated tests, and takes 30 average values ​​of the iterative optimization results for comparison. Other specific parameter values ​​are as follows: The initial values ​​of the parameters λ, ρ, μ, and η are randomly selected from the interval [0,5]; the population number N of the three heuristic algorithms PSO, FPA, and GSA is 50; in the PSO algorithm, the inertia weight is set to δ=0.8, and the recognition rate c 1 =2, social rate c 2 =2; in FPA, the switching probability is set to p = 0.25, the step size factor ξ = 0.8, and ζ and r are uniformly randomly selected from the interval [0,1] in each iteration, and the Levy distribution exponent β = 1.5; in GSA, the gravity constant The initial value of the gravitational constant G 0 =150 and attenuation rate α=20; MFPA switching probability is set to p=0.25, step size scaling factor ξ=0.8, v, ζ, ζ 1 , 2 , r, r 1 and r 2 is randomly generated between 0 and 1 in each iteration.

[0169] In order to compare the parameter optimization capabilities of the proposed MFPA with those of PSO, FPA and GSA for the MFAC-based data-driven signal processing system, the present invention performs parameter optimization on the MFAC-based data-driven signal processing SISO system.

[0170] The present invention uses the system in formula (13) to perform simulation experiments, and the expected output y * (k) is set as follows:

[0171] y * (k) = 0.5 + 0.5 × (-1) round(k / 200) ; (28)

[0172] The round function is used to round to the nearest integer.

[0173] The performance of the MFAC data-driven signal processing algorithm based on MFPA optimization is compared with the performance of randomly selected parameters, and the iterative output signal tracking performance under different parameters is obtained as follows: Figure 8 As shown. Figure 8 It can be seen that the iterative control output of the system obtained by the MFAC data-driven signal processing algorithm using MFPA optimized parameters has smaller overshoot and better tracking performance, and the system control output gradually converges with the increase of the number of iterations. In contrast, the MFAC data-driven signal processing algorithm controller output overshoot based on the random selection parameter design is very obvious, the jitter is severe, and the tracking performance is poor. Its performance is very different from the iterative control output obtained by the MFAC data-driven signal processing algorithm based on MFPA optimization; that is, the MFPA proposed in the present invention can optimize the parameters of the MFAC-based data-driven signal processing system, and then improve the tracking performance of the control system. Therefore, the MFAC data-driven signal processing system with parameter optimization based on the MFPA algorithm has better system stability and robustness than the MFAC data-driven signal processing system with randomly selected parameters.

[0174] In order to show the gap between the output of the improved flower pollination algorithm and the expected output, the present invention sets the optimization goal to minimize the error between the iterative control output and the expected output, and the designed function expression is as follows:

[0175]

[0176] Among them, y(k) is the iterative control output, K is the maximum iteration time, and the optimization goal is to expect the value of f to be as small as possible. This application takes the inverse of the minimum objective function value f as the reference benchmark, and then normalizes and forwards the iterative objective function f value. Correspondingly, the parameter optimization performance experimental results of the MFAC data-driven signal processing SISO system based on the proposed MFPA and PSO, FPA, and GSA optimization are obtained as follows Fig. 9 shown.

[0177] from Fig. 9 It can be seen that in the early stage of iteration, the GSA algorithm did not perform very well in the parameter optimization of the MFAC data-driven signal processing SISO system. When the number of iterative optimizations approached 8, it fell into a local optimal solution and stagnated. Its parameter optimization performance was the worst among the four. The convergence degree of the MFPA algorithm proposed in the present invention for the parameter optimization of the MFAC data-driven signal processing SISO system has been above 0.8. With the increase of iterations, the optimization performance of the proposed MFPA algorithm has been gradually optimized. At the same time, the proposed MFPA algorithm has greatly improved in performance compared with the FPA and PSO algorithms. Figure 8It is proved that the optimization method proposed in the present invention can help the classic FPA algorithm to achieve better performance than the algorithm can achieve, thereby demonstrating better global optimization capabilities. Compared with the classic PSO, FPA and GSA, the MFPA proposed in the present invention has better overall performance in optimizing the parameters of the MFAC data-driven signal processing SISO system, and can better guarantee the tracking performance of the system.

[0178] Table 3. Simulation results of objective function of SISO system optimized by MFPA algorithm

[0179]

[0180] Table 3 shows the maximum (worst), average, and minimum (best) values ​​of the objective function values ​​f of different algorithms obtained in 30 independent experiments when the proposed algorithm and three traditional nature-inspired algorithms are used to optimize the parameters of the MFAC-based data-driven signal processing SISO system. It can be seen from Table 3 that the MFPA algorithm proposed in the present invention has obvious advantages in the performance of optimizing the parameters of the MFAC-based data-driven signal processing SISO system, the global optimization performance of the parameters obtained by the PSO algorithm is second, and the global optimization performance of the parameters based on the FPA algorithm and the GSA algorithm is relatively poor.

[0181] The present invention constructs an input-output model according to a target single-input single-output discrete-time nonlinear system, generates an iterative input model according to the input-output model and a projection algorithm, obtains a target parameter of the iterative input model, iterates the target parameter based on an improved flower pollination algorithm, wherein a plurality of update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm, and when the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model. The present invention improves the flower pollination algorithm on the basis of the flower pollination algorithm, and then proposes an improved flower pollination algorithm with global optimization capability, applies the improved flower pollination algorithm to the parameter optimization of the data-driven signal processing model based on model-free adaptive control, avoids the problem that the system performance is affected by the random parameters when the system is applied to the wrong parameters, and further enables the performance of the data-driven signal processing model based on model-free adaptive control to be maintained in a good state.

[0182] Furthermore, if Fig.10 As shown, based on the above-mentioned data-driven signal processing method based on improved flower pollination, the present invention also provides a data-driven signal processing system based on improved flower pollination, wherein the data-driven signal processing system based on improved flower pollination includes:

[0183] A model generation module 91 is used to construct an input-output model according to a target single-input single-output discrete-time nonlinear system, and to generate an iterative input model according to the input-output model and a projection algorithm;

[0184] An iteration module 92, used for obtaining target parameters of the iterative input model, and iterating the target parameters based on the improved flower pollination algorithm, wherein a plurality of update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm;

[0185] The result generation module 93 is used to obtain the updated target parameters according to the last iteration when the number of iterations reaches the maximum number of iterations, and to generate a data-driven signal processing model based on model-free adaptive control according to the updated target parameters and the iterative input model.

[0186] Furthermore, if Fig.11 As shown, based on the above-mentioned data-driven signal processing method and system based on improved flower pollination, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Fig.11 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0187] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed in the terminal, such as the program code of the installation terminal, etc. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a data-driven signal processing program 40 based on improved flower pollination is stored on the memory 20, and the data-driven signal processing program 40 based on improved flower pollination can be executed by the processor 10, thereby realizing the data-driven signal processing method based on improved flower pollination in the present application.

[0188] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the data-driven signal processing method based on improved flower pollination.

[0189] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0190] In one embodiment, when the processor 10 executes the data-driven signal processing program 40 based on improving flower pollination in the memory 20, the steps of the above data-driven signal processing method based on improving flower pollination are implemented.

[0191] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a data-driven signal processing program based on improved flower pollination, and when the data-driven signal processing program based on improved flower pollination is executed by a processor, the steps of the data-driven signal processing method based on improved flower pollination as described above are implemented.

[0192] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or terminal including the element.

[0193] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0194] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A data-driven signal processing method based on improved flower pollination, characterized in that: The data-driven signal processing method based on improved flower pollination includes: According to the target single-input single-output discrete-time nonlinear system, an input-output model is constructed, and according to the input-output model and the projection algorithm, an iterative input model is generated; Obtaining target parameters of the iterative input model, and iterating the target parameters based on an improved flower pollination algorithm, wherein a plurality of updating schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm; When the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model.

2. The data-driven signal processing method based on improved flower pollination according to claim 1, characterized in that: The step of constructing an input-output model according to the target single-input single-output discrete-time nonlinear system and generating an iterative input model according to the input-output model and the projection algorithm specifically includes: Obtain a target single-input single-output discrete-time nonlinear system, and construct an input-output model according to the target single-input single-output discrete-time nonlinear system; According to the theorem satisfied by the target single-input single-output discrete-time nonlinear system, the input-output model is converted into an equivalent dynamic linearized data model; Generate a criterion function of a pseudo partial derivative in the equivalent dynamic linearization data model according to a projection algorithm, solve an optimal condition according to the criterion function, and generate a pseudo partial derivative formula according to the optimal condition and the criterion function; Obtain an expected system output, and generate the iterative input model according to the expected system output and the pseudo partial derivative formula.

3. The data-driven signal processing method based on improved flower pollination according to claim 2, characterized in that: The obtaining of the expected system output and generating the iterative input model according to the expected system output and the pseudo partial derivative formula specifically includes: Obtaining an expected system output, and generating a criterion function according to the expected system output; The equivalent dynamic linearization data model is updated according to the pseudo partial derivative formula to obtain an updated equivalent dynamic linearization data model, the updated equivalent dynamic linearization data model is substituted into the criterion function, and the partial derivative of the input is calculated to obtain the current optimal condition; The iterative input model is generated according to the current optimal condition and the pseudo partial derivative formula.

4. The data-driven signal processing method based on improved flower pollination according to claim 1, characterized in that: The step of obtaining the target parameter of the iterative input model and iterating the target parameter based on the improved flower pollination algorithm specifically includes: Obtaining target parameters of the iterative input model, initializing the number of individuals in the population according to the target parameters, and initializing the switching probability, the maximum number of iterations, and the dimensions of the individuals in the population; In each iteration, determine whether the current number of iterations is within the maximum number of iterations. When the current number of iterations is within the maximum number of iterations, generate a random number and determine whether the random number is greater than the switching probability. If it is greater, the global pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. When it is not greater, the local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual. The objective function value of the next-moment iterative solution of the current solution of each individual is calculated, and the global optimal solution is updated according to all objective function values.

5. The data-driven signal processing method based on improved flower pollination according to claim 4, characterized in that: The global pollination process is used to calculate the next-moment iterative solution of the current solution of each individual, specifically including: The next-moment iterative solution of the current solution of each individual is calculated using the global pollination process, wherein the global pollination process is expressed as the following formula: Among them, x i,g+1 represents the next iteration solution of the ith solution, x i,g represents the current position of the i-th solution, ξ is the scaling factor of the guide step, i is a positive integer in the set {1,2,...,N}, r is a random number generated in [0,1], t is the current iteration number, t max is the maximum number of iterations, L represents the pollination intensity, gbest g is the current optimal solution, x r1,g and x r2,g are two solutions randomly selected from the population when the number of iterations is t, v is a random number generated by a uniform distribution in the interval [0,1], and r1 and r2 are randomly generated numbers between 0 and 1.

6. The data-driven signal processing method based on improved flower pollination according to claim 5, characterized in that: The method of using the local pollination process to calculate the next-moment iterative solution of the current solution of each individual specifically includes: The local pollination process is used to calculate the next-moment iterative solution of the current solution of each individual, where the local pollination process is expressed as the following formula: Among them, x m1,g x m2,g x m3,g x m4,g are different solutions randomly selected from the population when the number of iterations is t, and ζ, ζ1, and ζ2 are step size proportional factors.

7. The data-driven signal processing method based on improved flower pollination according to claim 4, characterized in that: When the number of iterations reaches the maximum number of iterations, an updated target parameter is obtained according to the last iteration, and a data-driven signal processing model based on model-free adaptive control is generated according to the updated target parameter and the iterative input model, specifically including: When the number of iterations reaches the maximum number of iterations, the updated target parameters are obtained based on the global optimal solution obtained in the last iteration; The updated target parameters are used to replace corresponding parameters in the iterative input model to generate a data-driven signal processing model based on model-free adaptive control.

8. A data-driven signal processing system based on improved flower pollination, characterized in that: The data-driven signal processing system based on improved flower pollination includes: A model generation module, used to construct an input-output model according to a target single-input single-output discrete-time nonlinear system, and to generate an iterative input model according to the input-output model and a projection algorithm; An iteration module, used for obtaining target parameters of the iterative input model, and iterating the target parameters based on an improved flower pollination algorithm, wherein a plurality of update schemes are used to optimize the local pollination process and the global pollination process in the flower pollination algorithm to obtain the improved flower pollination algorithm; The result generation module is used to obtain the updated target parameters according to the last iteration when the number of iterations reaches the maximum number of iterations, and to generate a data-driven signal processing model based on model-free adaptive control according to the updated target parameters and the iterative input model.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a data-driven signal processing program based on improved flower pollination stored in the memory and executable on the processor, wherein the data-driven signal processing program based on improved flower pollination, when executed by the processor, implements the steps of the data-driven signal processing method based on improved flower pollination as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a data-driven signal processing program based on improved flower pollination, and when the data-driven signal processing program based on improved flower pollination is executed by a processor, the steps of the data-driven signal processing method based on improved flower pollination as described in any one of claims 1-7 are implemented.

Images