CN112861242A - non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm - Google Patents

non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm Download PDF

Info

Publication number
CN112861242A
CN112861242A CN202110287600.0A CN202110287600A CN112861242A CN 112861242 A CN112861242 A CN 112861242A CN 202110287600 A CN202110287600 A CN 202110287600A CN 112861242 A CN112861242 A CN 112861242A
Authority
CN
China
Prior art keywords
wind speed
gaussian
fluctuating wind
model
kernel function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110287600.0A
Other languages
Chinese (zh)
Inventor
路明璟
孙芳锦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Liaoning Technical University
Original Assignee
Liaoning Technical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Liaoning Technical University filed Critical Liaoning Technical University
Priority to CN202110287600.0A priority Critical patent/CN112861242A/en
Publication of CN112861242A publication Critical patent/CN112861242A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/10Geometric CAD
    • G06F30/13Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • G06F30/28Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/004Artificial life, i.e. computing arrangements simulating life
    • G06N3/006Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Geometry (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Computing Systems (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Biomedical Technology (AREA)
  • Algebra (AREA)
  • Civil Engineering (AREA)
  • Fluid Mechanics (AREA)
  • Architecture (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Structural Engineering (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method for predicting non-Gaussian fluctuating wind speed based on a hybrid intelligent algorithm, which comprises the following steps: generating a non-Gaussian fluctuating wind speed sample of a space point of a large-span spherical roof structure by JT conversion and AR model simulation, dividing the non-Gaussian fluctuating wind speed of each space point into a training set and a testing set, and performing normalization processing before each simulation test; respectively deducing kernel parameter combinations consisting of kernel parameters and penalty factors searched based on cuckoo and particle swarm intelligent algorithms; transforming a non-Gaussian fluctuating wind speed training set sample into a kernel function matrix by using a CS + PSO-LSSVM learning machine, mapping the kernel function matrix to a high-latitude characteristic space, then mapping input sample data to a high-dimensional characteristic space through a nonlinear function, then testing various linear algorithms on the kernel function matrix optimized by a hybrid intelligent algorithm to obtain a nonlinear model of the non-Gaussian fluctuating wind speed training sample, and predicting the non-Gaussian fluctuating wind speed testing set sample by using the nonlinear model.

Description

non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm
Technical Field
The invention belongs to the technical field of wind speed prediction, and particularly relates to a non-Gaussian pulsation wind speed prediction method based on a hybrid intelligent algorithm.
Background
With the rapid development of science and technology, the concept transformation and technical innovation of the building industry are continuously influenced, the large-span structure is widely applied to the most of a plurality of building structures, the structure is complex, the construction environment is changeable, the characteristics of the large-span structure are always concerned by people in all circles, and overseas scholars strive to pursue the goals of light weight, high flexibility and low damping of the large-span structure. The main problems faced in structural wind engineering are wind-induced dynamic response due to large structural span and high flexibility, and improved load degree estimation, prediction and control on wind load design and wind-induced vibration response. Nowadays, with the rapid development of computer science, people also more and more frequently fuse two disciplines in a crossed manner, and the numerical simulation of a large-span structure is a product under the current era background and scientific technology, and the numerical simulation method can be used for simulating a time course curve closest to the surface wind load of an actual building structure on site under specific conditions such as different wind field environments, site conditions, building structure forms and the like.
The probability model obeying normal distribution is regarded as Gaussian distribution in an academic sense, otherwise, the probability distribution is not regarded as Gaussian distribution when the probability model does not conform to the normal distribution, and the probability distribution can also be regarded as a random distribution state. Due to the fact that positive and negative wind pressure areas of pulsating wind are unevenly distributed on the top of the large-span structure, the building structure shakes violently, surface wind pressure vortex shedding can cause deformation vibration, and serious damage such as deformation or collapse of building surface materials can be caused under extreme conditions; structural members of the building are damaged when the dynamic displacement of the structure exceeds a certain limit value; the large amplitude surface vibrations can lead to fatigue failure of external components and appendages. Therefore, the complete non-Gaussian pulsating wind speed time-course data can be mastered, the damage caused by pulsating wind can be effectively controlled and prevented, and the method has important significance for building structure design and building structure disaster prevention and reduction.
The Least Square Support Vector Machine (LSSVM) has the advantages of considering both excellent estimation capability and calculation accuracy for processing small samples, replaces a secondary nonlinear equation set of the SVM by a linear equation set, and simultaneously converts inequality constraint of the SVM into linear equation constraint of the LSSVM, so that the sensitivity to the fitting problem is improved. Meanwhile, the LSSVM learning machine hopes that a large amount of data obtained by actual tests or numerical simulation can be trained and fitted, and a group of proxy models of input and output relations can be mapped. The basic principle can be explained as follows; for the estimation and fitting of the complex function, the LSSVM uses a group of training sample sets with the total number of sample data being N, and uses a nonlinear transformation function to convert the original input data from the original space RnThe method is characterized in that the method is mapped to a high-dimensional characteristic space, a high-dimensional nonlinear target function is converted into a linear function through a nonlinear transformation function, a mixed algorithm composed of a Bucky intelligent algorithm and a particle swarm algorithm is combined to optimize parameters of a kernel function, so that the least square support vector machine has good learning capacity and strong generalization capacity, the kernel parameter and punishment parameter in the kernel function are adjusted through the mixed intelligent algorithm to adjust regression analysis accuracy of the least square support vector machine, a non-Gaussian pulse wind speed time course generated through simulation is used as a training sample of the least square support vector machine, and a regression model is established to effectively predict the single-point non-Gaussian pulse wind speed of the large-span spherical roof structure.
Disclosure of Invention
Based on the defects of the prior art, the technical problem to be solved by the invention is to provide a non-Gaussian fluctuating wind speed prediction method based on a hybrid intelligent algorithm, which is characterized in that JT transformation and AR model numerical simulation is used for non-Gaussian fluctuating wind speed sample data, a cuckoo search algorithm (CS) is used for optimizing nuclear parameters, a particle swarm algorithm (PSO) is used for optimizing penalty factors to obtain a complete nuclear function parameter combination, the CS + PSO-LSSVM of the hybrid intelligent algorithm is used for simulating single-point wind speed of a large-span spherical roof structure and predicting, an actual wind speed and predicted wind speed adaptability curve, an Average Error (AE) and a Root Mean Square Error (RMSE) evaluation method are calculated.
In order to solve the technical problems, the invention is realized by the following technical scheme:
the invention provides a method for predicting non-Gaussian fluctuating wind speed based on a hybrid intelligent algorithm, which comprises the following steps:
the first step is as follows: generating a non-Gaussian fluctuating wind speed sample of a space point of a large-span spherical roof structure by JT conversion and AR model simulation, dividing the non-Gaussian fluctuating wind speed of each space point into a training set and a testing set, and performing normalization processing before each simulation test;
the second step is that: respectively deducing a kernel parameter combination consisting of kernel parameters and penalty factors searched based on the intelligent algorithm of cuckoo and particle swarm, forming an RBF kernel function, and establishing a learning machine model based on a hybrid intelligent algorithm (CS + PSO-LSSVM) according to Mercer theorem;
the third step: transforming a non-Gaussian fluctuating wind speed training set sample into a kernel function matrix by using a CS + PSO-LSSVM learning machine, mapping the kernel function matrix to a high-latitude characteristic space, mapping input data to a high-dimensional characteristic space through a nonlinear function, testing various linear algorithms on the kernel function matrix optimized by a hybrid intelligent algorithm to obtain a nonlinear model of the non-Gaussian fluctuating wind speed training sample, and predicting the non-Gaussian fluctuating wind speed testing set sample by using the nonlinear model;
and fourthly, comparing a large-span spherical roof structure space point test sample obtained by JT transformation and an AR model with a non-Gaussian fluctuating wind speed result predicted by CS + PSO-LSSVM based on a hybrid intelligent algorithm, and for the AR model, obtaining an approximate solution by preferentially interfering an input process through probability and then interfering a potential non-Gaussian white noise through a frequency interference algorithm. The Average Error (AE) of the predicted wind speed and the actual wind speed, the root Mean Square (MSE) and the adaptability curve are calculated to evaluate the effectiveness of the method.
Preferably, the JT transform + AR model in the first step simulates non-gaussian fluctuating wind speed as follows:
lognormal system SL:y(t)=τ+δln(u(t)-ξ)
Bounded system SB:
Figure BDA0002981137320000041
Unbounded system SU:
Figure BDA0002981137320000042
Figure BDA0002981137320000043
Figure BDA0002981137320000044
Figure BDA0002981137320000045
Figure BDA0002981137320000046
In the formula (1), u is a standard gaussian process with a mean value of 0 and a variance of 1, y is a non-gaussian process, and δ, λ, ξ and τ are conversion coefficients, and by solving a negative function, an inverse JT conversion function can be obtained.
In the formula (2), h (i) is the unit impulse response of the system, and when i is more than or equal to 0 and less than or equal to 9999, the value is equal to the output value, and when i is more than or equal to 10000, h (i) is equal to 0. The third and fourth central moments (i.e., skewness and peak) are calculated from the above formula to obtain the edge distortion moment relation MMDR as formula (3), and since the input is a Gaussian process, the MMDR is a formula
Figure BDA0002981137320000047
Preferably, in the second step, n training samples (x) are given1,x2,…xn) The RBF kernel function is expressed by the following formula:
K(xi,xj)=exp[-(xi-xj)2/2σ2] (4)
yi(xi)=ωTφ(xi)+b+ei,i=1,2,3,…,n (5)
in the formula (4), i, j is 1,2, …, N, γ is confidence space, i.e. penalty factor, σ is kernel parameter, b is bias value, eiIs the error variable of the sample point i, also called the relaxation variable, now the Lagrangian alphaiForming a support vector and introducing the support vector into a Lagrange function;
Figure BDA0002981137320000051
and finally, introducing an LSSVM model constructed in an even space, wherein the LSSVM model is obtained by KKT optimization conditions:
Figure BDA0002981137320000052
clearing the variable weight vector omega and the relaxation variable e in the formulaiThen, the following matrix equation is obtained;
Figure BDA0002981137320000053
in the formula (8), omegaij=φ(xi)Tφ(xj),
i,j=1,2,…,N;Y=(y1,y2,…,yN)T,P=(1,1,…,1)T;α=(α12,…,αN)TAnd I is an identity matrix, and according to the Mercer criterion, a minimum support vector machine model estimated by a characterization function after a kernel function replaces a conversion function is expressed as follows:
Figure BDA0002981137320000054
preferably, in the third step, a widely used Logistic mapping method is selected as a chaos theory model for the nuclear parameters, the process of searching for the initial nest position of the cuckoo population is improved and optimized, the iteration is set for 100 times, and the model is as follows:
mi+1=μ×mi×(1-mi),i=1,2,3…,n (10)
where μ is the control variable of the Logistic mapping method, and can be generally in [0,4 ]]Value, which determines the chaos degree, i.e. diversity and regularity, of the mapping method, the larger the value of mu, the larger the chaos degree, the larger the chaos variable mi∈[0,1]。
Then according to a formula, mapping N-1 chaotic variables to a cuckoo value space [ x ] by a Logistic methodminj,xmaxj]In the above, cuckoo population x is generatedij
xij=xminj+mkj(xmaxj-xminj),i=1,2,…,n;j=1,2,…,n (11)
Finally, calculating the reverse solution of the cuckoo population by using a formula (12)
Figure BDA0002981137320000061
I.e. the optimum value of the nuclear parameter:
Figure BDA0002981137320000062
for the penalty factor, the invention utilizes the particle swarm optimization to carry out optimization, uses the individual extreme value and the global optimum value to be combined with the following formula to update the speed and the position of the particle, and sets the population number of the particle swarm to be 30.
Figure BDA0002981137320000063
Figure BDA0002981137320000064
Ebest=(xi1,xi2,xi3,…,xiN),i=1,2,3,…M (15)
Gbest=(xi1,xi2,xi3,…,xiN),i=1,2,3,…M (16)
Wherein, c1,c2Is cognitive parameter and social parameter, generally takes 2, omega is inertia constant, generally takes 0.9, r1,r2Is [0, 1 ]]The random number in the range, g, is the current iteration computation algebra. And obtaining the optimal value of the penalty factor according to the iteration termination times and the fitness condition to obtain the kernel function parameter combination. And the second step is carried out, and a complete CS + PSO-LSSVM learning machine model is established.
Therefore, the prediction model of the non-Gaussian fluctuation wind speed prediction method based on the hybrid intelligent algorithm has good learning and generalization capability under the action of the Gaussian kernel function, the errors of the training set and the errors of the testing set are within the specified range, and the accuracy is high in the whole data set. Meanwhile, according to the comparison of the simulation results of the wind tunnel test, the wind pressure coefficient distribution and the time course of the non-Gaussian fluctuating wind speed also accord with the distribution rule of the actual test result, the goodness of fit is good, and the method can be used as an effective method for predicting the non-Gaussian fluctuating wind speed. The design scheme and the theoretical basis are provided for the wind-resistant design of the large-span building construction.
The foregoing description is only an overview of the technical solutions of the present invention, and in order to make the technical means of the present invention more clearly understood, the present invention may be implemented in accordance with the content of the description, and in order to make the above and other objects, features, and advantages of the present invention more clearly understood, the following detailed description is given in conjunction with the preferred embodiments, together with the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments will be briefly described below.
FIG. 1 is a schematic diagram of a design framework of a non-Gaussian fluctuating wind speed prediction method based on a hybrid intelligent algorithm CS + PSO-LSSVM;
FIG. 2 is a schematic program flow diagram of a non-Gaussian fluctuating wind speed prediction method based on a hybrid intelligent algorithm CS + PSO-LSSVM;
FIG. 3 is a space test point layout diagram of a large-span spherical roof structure;
FIG. 4 is a time-course diagram of non-Gaussian fluctuation wind speed simulation of a large-span spherical roof structure space test point;
FIG. 5 is a comparison graph of a non-Gaussian wind speed fit curve of a predicted wind speed and an actual wind speed of a CS-LSSVM;
FIG. 6 is a comparison graph of non-Gaussian wind speed adaptive function curves of the predicted wind speed and the actual wind speed of the CS-LSSVM;
FIG. 7 is a comparison graph of the training set of the predicted wind speed and the actual wind speed of the CS-LSSVM and the prediction result;
FIG. 8 is a comparison graph of a CS-LSSVM predicted wind speed and actual wind speed test set and a prediction result;
FIG. 9 is a comparison graph of a non-Gaussian wind speed fit curve of the predicted wind speed of the PSO-LSSVM and the actual wind speed;
FIG. 10 is a comparison graph of non-Gaussian wind speed adaptive function curves of the predicted wind speed and the actual wind speed of the PSO-LSSVM;
FIG. 11 is a comparison graph of the PSO-LSSVM predicted wind speed and actual wind speed training set with the predicted results;
FIG. 12 is a comparison graph of a PSO-LSSVM predicted wind speed and actual wind speed test set with predicted results;
FIG. 13 is a comparison graph of a non-Gaussian wind speed fit curve of the predicted wind speed of CS + PSO-LSSVM and the actual wind speed;
FIG. 14 is a comparison graph of non-Gaussian wind speed adaptive function curves of the predicted wind speed and the actual wind speed of the CS + PSO-LSSVM;
FIG. 15 is a comparison graph of the training set of the predicted wind speed and the actual wind speed of the CS + PSO-LSSVM and the prediction result;
FIG. 16 is a comparison graph of the test set of the predicted wind speed and the actual wind speed of the CS + PSO-LSSVM and the prediction result.
Detailed Description
Other aspects, features and advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, which form a part of this specification, and which illustrate, by way of example, the principles of the invention. In the referenced drawings, the same or similar components in different drawings are denoted by the same reference numerals.
As shown in fig. 1-16, the standard cuckoo algorithm of the present invention is randomly assigned to the initial random nest positions, and the calculation result depends on the initial random nest positions, which is not favorable for the convergence of the algorithm and affects the final result. The chaos theory mapping reverse learning method is used for carrying out initialization generation distribution on cuckoo nest acupuncture points, random generation distribution of a standard cuckoo algorithm is replaced, the cuckoo population search regularity is enhanced, and the improved cuckoo algorithm is more universal and diversified. Particle Swarm Optimization (PSO) is a meta-heuristic algorithm because it makes little or no assumptions about the problem being optimized and is able to search for a very large candidate solution space. The method has the advantages of simple calculation, easy realization of the purpose and less calculation parameters, but is easy to fall into the local optimum, thereby causing the situation of low convergence precision and difficult convergence. The RBF kernel function parameter combination optimization is carried out by coupling the cuckoo search algorithm and the particle swarm algorithm, a least square support vector machine is constructed, and the method has good learning capability and generalization capability.
The invention discloses a method for predicting non-Gaussian fluctuating wind speed based on a hybrid intelligent algorithm, which comprises the following steps of:
the method comprises the following steps: generating a non-Gaussian fluctuating wind speed sample of a space point of a large-span spherical roof structure by JT conversion and AR model simulation, dividing the non-Gaussian fluctuating wind speed of each space point into a training set and a testing set, and performing normalization processing before each simulation test;
step two: respectively deducing a kernel parameter combination consisting of kernel parameters and penalty factors searched based on the intelligent algorithm of cuckoo and particle swarm, forming an RBF kernel function, and establishing a learning machine model based on a hybrid intelligent algorithm according to Mercer theorem;
step three: introducing a cuckoo search algorithm and a particle swarm algorithm, optimizing RBF kernel function parameters gamma, a penalty function c and a weight function omega respectively, wherein CS optimizes gamma and PSO optimizes c, and determines optimal model parameters; transforming the non-Gaussian fluctuating wind speed training sample into a kernel function matrix by using a kernel function parameter combination optimized by CS + POS, mapping the kernel function matrix to a high-latitude characteristic space from a low-latitude space, then performing any linear algorithm on the kernel function matrix to obtain a non-linear model of the non-Gaussian fluctuating wind speed training sample, and predicting the non-Gaussian fluctuating wind speed by using the model;
step four: and comparing the test sample with the non-Gaussian fluctuating wind speed result predicted by the CS + PSO-LSSVM of the hybrid intelligent algorithm, and calculating the average error, the root mean square error and the correlation coefficient of the predicted wind speed and the actual wind speed so as to evaluate the accuracy of the method.
In step one, the JT + AR model simulates non-Gaussian fluctuating wind speed is described by the following equation:
the output of the JT model can be expressed as:
lognormal system SL:u=τ+δln(y-ξ)
Bounded system SB:
Figure BDA0002981137320000101
Unbounded system SU:
Figure BDA0002981137320000102
Wherein u is a standard gaussian process with a mean of 0 and a variance of 1, and y is a non-gaussian process, and δ, λ, ξ, and τ are conversion coefficients;
Figure BDA0002981137320000103
the output of the AR model can be expressed as:
Figure BDA0002981137320000104
wherein h (i) is the unit impact response of the system, when i is more than or equal to 0 and less than or equal to 9999, the value is equal to the output value, and when i is more than or equal to 10000, h (i) is equal to 0; the third and fourth central moments are solved from the above formula to obtain the edge distortion moment relation MMDR:
Figure BDA0002981137320000105
Figure BDA0002981137320000106
Figure BDA0002981137320000107
in step two, the RBF kernel function is expressed as follows:
K(xi,xj)=exp[-(xi-xj)2/2γ2]
wherein gamma is an RBF kernel function parameter.
In the third step, setting the particle swarm size m to be 30, randomly generating an initial position of a kernel function parameter, determining the range of a parameter C to be optimized, setting the maximum iteration times, similarly setting the iteration times of cuckoos, initializing the number and the position of bird nests, and determining the range of a kernel function parameter gamma;
and terminating iteration according to a termination condition, determining an optimal parameter, and establishing a CS + PSO-LSSVM non-Gaussian fluctuating wind speed prediction model of a hybrid intelligent algorithm.
The following describes in detail the prediction of the single-point non-gaussian pulsating wind speed of the large-span spherical roof structure according to the embodiment of the present invention with reference to the accompanying drawings, wherein the steps are as follows:
the first step is as follows: the ground roughness is B level, a large-span spherical roof structure space point non-Gaussian fluctuating wind speed sample with the height of 35 meters, the roof diameter of 80 meters, the height of an eave opening of 25 meters and the average wind speed of 15m/s is generated by using a JT conversion and AR model simulation, a Davenport spectrum is adopted as a wind speed power spectrum, the time interval is 0.5s, 3 space points are averagely arranged along the height of the roof, the non-Gaussian fluctuating wind speed of each space point is divided into a training set and a testing set, and normalization processing is carried out before each simulation test.
The second step is that: and (3) establishing a JT transformation and AR model autoregressive model respectively, and generating non-Gaussian fluctuation wind speed time-course curves of 1600 sampling time points in 7 simulated space wind speed points 200S. And respectively taking the non-Gaussian fluctuating wind speeds of the first 1200 sampling time points of each space single point as a training set, and taking the wind speed time courses of the last 400 sampling points as a test set. And (3) establishing a CS + PSO-LSSVM prediction model, wherein the flow chart is shown in the figure 1 and the figure 2.
The third step: introducing a CS + PSO-LSSVM prediction model, optimizing RBF kernel function parameters and penalty factors, and determining an optimal kernel parameter combination: and (3) performing cuckoo search optimization on the kernel function parameters, performing particle swarm optimization on the penalty factors, and coupling the two algorithms to obtain a hybrid intelligent algorithm. And establishing a CS + PSO-LSSVM prediction model based on a hybrid intelligent algorithm. Determining the number of particle swarms, weight coefficients and iteration times, mapping a non-Gaussian fluctuating wind speed training set to a high-dimensional characteristic space by using a CS + PSO-LSSVM prediction model through a nonlinear function to obtain a nonlinear model of a non-Gaussian fluctuating wind speed training sample, and predicting a non-Gaussian fluctuating wind speed test set by using the linear model to obtain a training regression prediction model.
The fourth step: inputting the non-Gaussian fluctuating wind speeds of the next 400 single-point sampling time points into a learning machine as a test set, predicting the non-Gaussian fluctuating wind speeds of the prediction set by using a regression prediction model output by a training set, comparing a test set sample with a non-Gaussian fluctuating wind speed result obtained based on a hybrid intelligent algorithm CS + PSO-LSSVM prediction model, comparing the training set with the non-Gaussian fluctuating wind speed result obtained by training and learning of the hybrid intelligent algorithm, and evaluating the effectiveness of the method by calculating the error (AE) and the root Mean Square Error (MSE) of the predicted wind speed and the actual wind speed.
The comparison result of the CS + PSO-LSSVM prediction method by utilizing the CS-LSSVM and the PSO-LSSVM is as follows:
Figure BDA0002981137320000121
the implementation process of the invention is intuitively given through the steps, and the analysis result shows that the correlation coefficients of the results of the prediction method of the hybrid intelligent algorithm CS + PSO-LSSVM are all larger than 0.9 (the correlation coefficient is larger than 0.9, which shows that the correlation is strong), and the root mean square error shows that the prediction result of the hybrid intelligent algorithm better converges to the actual wind speed. The method couples the cuckoo search algorithm and the particle swarm algorithm to obtain a hybrid intelligent algorithm, performs advantage complementation by utilizing the CS algorithm and the PSO algorithm to a certain extent, integrates the training advantages of the Gaussian kernel function, optimizes the kernel function parameter combination, further improves the accuracy of a prediction result, and provides a method with higher accuracy and higher calculation efficiency for the prediction of the non-Gaussian fluctuating wind speed.
While the foregoing is directed to the preferred embodiment of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof, and the scope thereof is determined by the claims that follow.

Claims (4)

1. A method for predicting non-Gaussian fluctuating wind speed based on a hybrid intelligent algorithm is characterized by comprising the following steps:
the method comprises the following steps: generating a non-Gaussian fluctuating wind speed sample of a space point of a large-span spherical roof structure by JT conversion and AR model simulation, dividing the non-Gaussian fluctuating wind speed of each space point into a training set and a testing set, and performing normalization processing before each simulation test;
step two: respectively deducing a kernel parameter combination consisting of kernel parameters and penalty factors searched based on the intelligent algorithm of cuckoo and particle swarm, forming an RBF kernel function, and establishing a learning machine model based on a hybrid intelligent algorithm according to Mercer theorem;
step three: introducing a cuckoo search algorithm and a particle swarm algorithm, optimizing RBF kernel function parameters gamma, a penalty function c and a weight function omega respectively, wherein CS optimizes gamma and PSO optimizes c, and determines optimal model parameters; transforming the non-Gaussian fluctuating wind speed training sample into a kernel function matrix by using a kernel function parameter combination optimized by CS + POS, mapping the kernel function matrix to a high-latitude characteristic space from a low-latitude space, then performing any linear algorithm on the kernel function matrix to obtain a non-linear model of the non-Gaussian fluctuating wind speed training sample, and predicting the non-Gaussian fluctuating wind speed by using the model;
step four: and comparing the test sample with the non-Gaussian fluctuating wind speed result predicted by the CS + PSO-LSSVM of the hybrid intelligent algorithm, and calculating the average error, the root mean square error and the correlation coefficient of the predicted wind speed and the actual wind speed so as to evaluate the accuracy of the method.
2. The hybrid intelligent algorithm-based non-gaussian pulsating wind speed prediction method of claim 1, wherein in step one, the JT + AR model modeling non-gaussian pulsating wind speed is described by the following equation:
the output of the JT model can be expressed as:
lognormal system SL:u=τ+δln(y-ξ)
Bounded system SB:
Figure FDA0002981137310000021
Unbounded system SU:
Figure FDA0002981137310000022
Wherein u is a standard gaussian process with a mean of 0 and a variance of 1, and y is a non-gaussian process, and δ, λ, ξ, and τ are conversion coefficients;
Figure FDA0002981137310000023
the output of the AR model can be expressed as:
Figure FDA0002981137310000024
wherein h (i) is the unit impact response of the system, when i is more than or equal to 0 and less than or equal to 9999, the value is equal to the output value, and when i is more than or equal to 10000, h (i) is equal to 0; the third and fourth central moments are solved from the above formula to obtain the edge distortion moment relation MMDR:
Figure FDA0002981137310000025
Figure FDA0002981137310000026
Figure FDA0002981137310000027
3. the hybrid intelligent algorithm-based non-gaussian pulsating wind speed prediction method according to claim 1, wherein in step two, the RBF kernel function is expressed as follows:
K(xi,xj)=exp[-(xi-xj)2/2γ2]
wherein gamma is an RBF kernel function parameter.
4. The method for predicting the non-Gaussian fluctuating wind speed based on the hybrid intelligent algorithm according to claim 1, wherein in the third step, the particle swarm size m is set to be 30, the initial position of the kernel function parameter is randomly generated, the range of the parameter C to be optimized is determined, the maximum number of iterations is set, the number of cuckoo iterations is also set, the number and the position of bird nests are initialized, and the range of the kernel function parameter γ is determined;
and terminating iteration according to a termination condition, determining an optimal parameter, and establishing a CS + PSO-LSSVM non-Gaussian fluctuating wind speed prediction model of a hybrid intelligent algorithm.
CN202110287600.0A 2021-03-17 2021-03-17 non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm Pending CN112861242A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110287600.0A CN112861242A (en) 2021-03-17 2021-03-17 non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110287600.0A CN112861242A (en) 2021-03-17 2021-03-17 non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm

Publications (1)

Publication Number Publication Date
CN112861242A true CN112861242A (en) 2021-05-28

Family

ID=75995154

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110287600.0A Pending CN112861242A (en) 2021-03-17 2021-03-17 non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm

Country Status (1)

Country Link
CN (1) CN112861242A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116089844A (en) * 2023-04-11 2023-05-09 南京信息工程大学 non-Gaussian feature verification method for pose data of unmanned aerial vehicle
CN117115377A (en) * 2023-10-18 2023-11-24 云南滇能智慧能源有限公司 Wind farm energy model creation method, device, equipment and storage medium
CN117973429A (en) * 2024-04-01 2024-05-03 南京信息工程大学 Model parameter ratio estimation method applied to non-Gaussian noise filtering

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7979363B1 (en) * 2008-03-06 2011-07-12 Thomas Cecil Minter Priori probability and probability of error estimation for adaptive bayes pattern recognition
CN107153874A (en) * 2017-04-11 2017-09-12 中国农业大学 Water quality prediction method and system
CN109447178A (en) * 2018-11-13 2019-03-08 淮北师范大学 A kind of svm classifier method based on mixed kernel function
CN109871625A (en) * 2019-02-26 2019-06-11 西南交通大学 Non-gaussian wind pressure analogy method based on Johnson transformation
US20200301408A1 (en) * 2017-05-25 2020-09-24 Johnson Controls Technology Company Model predictive maintenance system with degradation impact model
AU2020104000A4 (en) * 2020-12-10 2021-02-18 Guangxi University Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7979363B1 (en) * 2008-03-06 2011-07-12 Thomas Cecil Minter Priori probability and probability of error estimation for adaptive bayes pattern recognition
CN107153874A (en) * 2017-04-11 2017-09-12 中国农业大学 Water quality prediction method and system
US20200301408A1 (en) * 2017-05-25 2020-09-24 Johnson Controls Technology Company Model predictive maintenance system with degradation impact model
CN109447178A (en) * 2018-11-13 2019-03-08 淮北师范大学 A kind of svm classifier method based on mixed kernel function
CN109871625A (en) * 2019-02-26 2019-06-11 西南交通大学 Non-gaussian wind pressure analogy method based on Johnson transformation
AU2020104000A4 (en) * 2020-12-10 2021-02-18 Guangxi University Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
孙芳锦;梁爽;张大明;: "基于回归神经网络的大跨度结构风压场预测", 地震工程与工程振动, no. 05, 15 October 2014 (2014-10-15) *
李春祥;丁晓达;郑晓芬: "基于混合智能优化LSSVM的非高斯脉动风速预测", 振动与冲击, vol. 36, no. 20, 28 October 2017 (2017-10-28) *
杜雪;李军;: "基于高斯过程的短期风电功率概率预测", 测控技术, no. 03, 18 March 2018 (2018-03-18) *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116089844A (en) * 2023-04-11 2023-05-09 南京信息工程大学 non-Gaussian feature verification method for pose data of unmanned aerial vehicle
CN116089844B (en) * 2023-04-11 2023-06-16 南京信息工程大学 non-Gaussian feature verification method for pose data of unmanned aerial vehicle
CN117115377A (en) * 2023-10-18 2023-11-24 云南滇能智慧能源有限公司 Wind farm energy model creation method, device, equipment and storage medium
CN117973429A (en) * 2024-04-01 2024-05-03 南京信息工程大学 Model parameter ratio estimation method applied to non-Gaussian noise filtering
CN117973429B (en) * 2024-04-01 2024-06-07 南京信息工程大学 Model parameter ratio estimation method applied to non-Gaussian noise filtering

Similar Documents

Publication Publication Date Title
CN112861242A (en) non-Gaussian fluctuating wind speed prediction method based on hybrid intelligent algorithm
CN111860982A (en) Wind power plant short-term wind power prediction method based on VMD-FCM-GRU
Ramezani et al. Application of artificial neural networks in optimal tuning of tuned mass dampers implemented in high-rise buildings subjected to wind load
CN110866592B (en) Model training method, device, energy efficiency prediction method, device and storage medium
CN105354363A (en) Fluctuation wind speed prediction method based on extreme learning machine
CN109885916B (en) Mixed test online model updating method based on LSSVM
CN110942205A (en) Short-term photovoltaic power generation power prediction method based on HIMVO-SVM
CN110991721A (en) Short-term wind speed prediction method based on improved empirical mode decomposition and support vector machine
CN105205495A (en) Non-stationary fluctuating wind speed forecasting method based on EMD-ELM
Jallal et al. Air temperature forecasting using artificial neural networks with delayed exogenous input
CN116894384B (en) Multi-fan wind speed space-time prediction method and system
CN114492922A (en) Medium-and-long-term power generation capacity prediction method
CN104992008A (en) Hilbert space multi-kernel function multiplication based wind speed prediction method
CN112149883A (en) Photovoltaic power prediction method based on FWA-BP neural network
CN114119273A (en) Park comprehensive energy system non-invasive load decomposition method and system
CN108256693B (en) Photovoltaic power generation power prediction method, device and system
Kaveh et al. Feasibility of PSO-ANFIS-PSO and GA-ANFIS-GA models in prediction of peak ground acceleration
CN116345555A (en) CNN-ISCA-LSTM model-based short-term photovoltaic power generation power prediction method
CN114862023A (en) Distributed photovoltaic power prediction method and system based on four-dimensional point-by-point meteorological forecast
Yaïci et al. Artificial neural network modelling for performance prediction of solar energy system
El Shahat et al. An artificial neural network model for wind energy estimation
CN113435595A (en) Two-stage optimization method for extreme learning machine network parameters based on natural evolution strategy
Marinke et al. Particle swarm optimization (PSO) applied to fuzzy modeling in a thermal-vacuum system
CN117114438A (en) Building area energy system cold and hot load data driving prediction method with flexibility and interpretability
CN116681154A (en) Photovoltaic power calculation method based on EMD-AO-DELM

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination