I want to automate a 2 input-1 output Hammerstein-Wiener model estimation through genetic algorithms

I tried to automate the process of estimation of the parameters for an hammerstein-wiener model using the existing matlab funtions. However I was not able to put it to work. If I run the functions in the command window without the genetic algorithms it all works properly but when I introduce the genetic algorithm estimations, specially in the cost function it starts to give back errors, maybe related with the code in the cost function I think is related with not initializing the parameters for the model but I do not know how to solve this. The code I implemented is as follows:

% Define GA options

options = optimoptions('ga','PopulationSize',5,'MaxGenerations',5);

% Define GA optimization function

fun = @(x)modelTuning(x);

% Data for training 2 inputs, 1 output

timedata = iddata (Ensaio2t2,[Ensaio2in(:,1), Ensaio2in(:,2)]);

% initmodel = idnlhw ([[2 2],[3 3],[6 4]]);

% Define bounds

lb = [1 1 1 1];

%

ub = [10 10 10 10];

% Run GA optimization

[x,fval] = ga(fun,4,[],[],[],[],lb,ub,[],options);

testmodel = idnlhw ([[x(1) x(2)],[x(3) x(4)],[6 4]]);

model20234t2 = nlhw(timedata,testmodel);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function error = modelTuning(x)

% tmodel = idnlhw ([[2 2],[3 3],[6 4]]);

tmodel = idnlhw ([[x(1) x(2)],[x(3) x(4)],[6 4]]);

timedata = iddata (Ensaio2t2,[Ensaio2in(:,1), Ensaio2in(:,2)]);

modelt2 = nlhw(timedata, tmodel);

[~,q,~]=compare (timedata, modelt2);

error = 100-q;

end

The error:

Error using idnlhw

The "idnlhw" command should be called with the syntax:

model = idnlhw(Orders, InputNonlinearity, OutputNonlinearity, PV-pairs), or

model = idnlhw(PV-pairs).

Type "help idnlhw" for more information.

Error in modelestimation>modelTuning (line 24)

tmodel = idnlhw ([[x(1) x(2)],[x(3) x(4)],[6 4]]);

Error in modelestimation>@(x)modelTuning(x) (line 6)

fun = @(x)modelTuning(x);

Error in createAnonymousFcn>@(x)fcn(x,FcnArgs{:}) (line 11)

fcn_handle = @(x) fcn(x,FcnArgs{:});

Error in makeState (line 58)

firstMemberScore = FitnessFcn(state.Population(initScoreProvided+1,:));

Error in galincon (line 24)

state = makeState(GenomeLength,FitnessFcn,Iterate,output.problemtype,options);

Error in ga (line 416)

[x,fval,exitFlag,output,population,scores] = galincon(FitnessFcn,nvars, ...

Error in modelestimation (line 17)

[x,fval] = ga(fun,4,[],[],[],[],lb,ub,[],options);

Caused by:

Failure in initial user-supplied fitness function evaluation. GA cannot continue.

[Ivo Cabral]

I am a student only starting to use MATLAB last year so I do not know if the resolution to this is simple or not but If anyone can give me any help I would appreciate it. Thank you in advance!

Answer 1

[Tianyu]

Hi Ivo,

Here are a few suggestions you can try to modify your code

I do not have the data, so allow me to use iddata1 for example

>> load iddata1.mat z1

% 1, you need to pass your data to your function, using the following syntax

>> fun = @(x,z1)modelTuning(x,z1);

...

>> [x,fval] = ga(@(x)fun(x,z1),4,[],[],[],[],lb,ub,[],options);

% In function definition, use

>> function error = modelTuning(x,z1)

% 2, From the error information, you can see that the "idnlhw" is not called with correct syntax, i.e. "SYS = IDNLHW(ORDERS,InputNL,OutputNL)". ORDERS should be integers. Something like this should work

>> idnlhw ([[1 1],[1 1],[1 1]]);

% If you want to treat x as integers, add integer constraint in GA algorithm. See https://www.mathworks.com/help/gads/ga.html?searchHighlight=ga&s_tid=srchtitle_support_results_1_ga#d124e52414