Error Not enough input arguments using Fmincon with nlcon and linear constraints

I am trying to minimize a function say fun using fmincon, it has 18 elements, having linear and non-linear constraints. linear constraints are applied on solution s, s(1) to s(7). and 2 nonlcon constraints are applied on solution s(8) to s(18).optimization is good for linear constraints but shows error using nonlcon function(@test3), not enough input arguments.

C_test_43.csv, file is required to run this code, it is attached with.

% code
clc
clear
M = csvread('C_test_43.csv');
time=50;
C=M(1:time);
Cmax=max(C);
T=10; 
n=time/T;
X=zeros(1,n);
X(1,1)=8;
x=zeros(1,time);  
x(1,1)=8;
E=1+(25-1).*rand(1,time);
batcharg=1;
Ve=(25-(2*batcharg))/Cmax; 
Vt=(25-(2*batcharg))/Cmax;
V=(25-(2*batcharg))/Cmax; 
m=11; 
fr = 0 + (16.8*10^9-0).*rand(m,time);  
fmax=max(fr);
fmax=max(fmax);  
p = 0 + (2*10^-3-0).*rand(m,time);    
pmax=max(p);
pmax=max(pmax);
Q=rand(1,time); 
QD=rand(1,time);
a=0.1;
bb=0.025;
k=7.8*10^-21; 
L=6400; 
Hm=-50 + (-75-50).*rand(m,time);  
phi=0.1;
w=1*10^6; 
N=-174; 
tau=1;
DL=tau.*fr./L;                  
D=w.*tau.*log2(1+(Hm.*p)/N.*w);  
Td=0+(20-0).*rand(1,time);
vlb=zeros(1,m+7);
vub=zeros(1,m+7);
num=m+7;
energy_ub=1500;
f=zeros(1,num); 

% code
fun=@(s)f; 
for t=1:time
  n=ceil(t/T); 
  freq=fr(1:m,t); 
f(1,1)=X(n);    
f(1,2)=x(t);     
f(1,3)=V.*C(n); 
f(1,4)=Q(t);    
f(1,5)=QD(t);   
f(1,6)=Ve;      
f(1,7)=Vt;         
  for u=8:m+7
      f(1,u)=min(DL(u-7,t)+D(u-7,t),D(u-7,t)+D(u-7,t)); 
  end
  i=1; 
  j=m+7; 
  Aeq=ones(i,j);
  for j=8:m+7
     Aeq(i,j)=-1;
  end
  beq=1;

    for j=8:m+7
       vlb(j)=0;
       vub(j)=fmax; 
    end

   vlb(1)=-batcharg; 
   vub(1)=batcharg;  
   vlb(2)=-1; 
   vub(2)=1;  
   vub(3)=energy_ub;  
   vlb(4)=-1;        
   vub(4)=1;
   vlb(5)=-1;        
   vub(5)=1;
   vlb(6)=0;         
   vub(6)=1000;
   vlb(7)=0;         
   vub(7)=20;

    if X(n)>-V*C(n)     
        vub(1)=1; 
    else
        vlb(1)=-1;
    end
    if x(t)>-V*C(t)
    vub(2)=1; 
    else
        vlb(2)=-1; 
    end 

    if Q(t)>-Ve*E(t) 
    vub(4)=1; 
    else
        vlb(4)=0; 
    end

    if QD(t)>-Vt*Td(t) 
    vub(5)=1; 
    else
        vlb(5)=0; 
    end

    nonlcon = @test3; 
    [s,fval]=fmincon(fun,freq,[],[],Aeq,beq,vlb,vub,nonlcon(tau,L,m));

% function for 2 non linear constraints

function [c,ceq] = test3(s,tau,L,m)
 for n=8:m+7
 ceq(1) =tau.*(s(n)/L);    
 ceq(2)=w(n).*tau.*log2(1+(Hm.*p)/N.*w(n));  
 end
 c = []; 
end

[Walter Roberson]

Search for Paramaterizing Functions

Answer 1

[Nicolas Schmit]

There are several problems in your code

• You must pass the optimization variable as a argument to the nonlinear constraints.
• The size of Aeq, Beq, vlb and vub do not match the size of the freq vector.
• The cost function must return a scalar.

I suggest you have a look at the examples in the documentation of fmincon.

https://www.mathworks.com/help/optim/ug/fmincon.html

[ikram shahzad]

Thanks to Walter Roberson and Nicolas Schmit for suggestions. I got out of previous error, but now encountering Error using fmincon (line 609) Supplied objective function must return a scalar value.

[Walter Roberson]

You have

f=zeros(1,num); %for m apps,1 row and 4m coloumns, variables to be found.

fun=@(s)f; %to use non linear constraint

This says that no matter what the input is, a vector of zeros is to be returned.

You then store other things into f, but those will be ignored: when you define an anonymous function with reference to a variable, then the value of the variable is captured as of the time the anonymous function is defined.

If you were to move the

fun=@(s)f; %to use non linear constraint

to just before

    [s,fval]=fmincon(fun,xo,[],[],Aeq,beq,vlb,vub,nonlcon(xo,tau,L,m)); %s=[Rn,Rtau,Hn,DL,DO,Bn,bt,Pm]

then the changed value of f would be captured and returned when the function was invoked. However, fmincon requires that the function return a scalar value.

You also have

      nonlcon = @test3; %non linear constraints

and you have

    [s,fval]=fmincon(fun,xo,[],[],Aeq,beq,vlb,vub,nonlcon(xo,tau,L,m)); %s=[Rn,Rtau,Hn,DL,DO,Bn,bt,Pm]

so test3 will be invoked once with xo, tau, L, m and would be expected to return a function handle that could be invoked each time to test whether the current variables met the constraints. However, test3 returns a numeric value, not a function handle.

You could try

    fun = @(s) 0;
    [s,fval]=fmincon(fun, xo, [], [], Aeq, beq, vlb, vub, @(s) nonlcon(s,tau,L,m)); %s=[Rn,Rtau,Hn,DL,DO,Bn,bt,Pm]

This will at least execute, though ignoring all of the values you carefully placed into f.

Note: we also see in ctp3 that you have

sum=0;

and we see in test3 that you have

   ceq = sum(ceq.^2);

Was the intention that ceq should result in 0? Since you defined sum as 0?

It is not a syntax error to define sum as a variable, but you can be very sure that defining sum as a variable will confuse people reading your code. If you continue to use sum as the name of a variable, sooner or later you will encounter a situation where you try to use sum as a function in the context where it is a variable, and then you will get rather confused as you try to figure out what is happening.

[ikram shahzad]

Thanks to Walter Roberson and Nicolas Schmit for suggestions. I tried the examples in following link but these are very simple https://www.mathworks.com/help/optim/ug/fmincon.html#busow0u-1

I also tried

x = fgoalattain(@myfun,x0,goal,weight,A,b,Aeq,beq,...
                lb,ub,@mycon)

where mycon is a MATLAB function such as

function [c,ceq] = mycon(x)
c = ...         % compute nonlinear inequalities at x.
ceq = ...       % compute nonlinear equalities at x.

function[c ,ceq]=test3(s,tau,L) results a vector but fmincon needs a scalar. when i use fmincon it shows error :

 (line 609) Supplied objective function must return a scalar value.

when i use fgoatattain, it gives error

too many input arguments

moreover, i am confused to select values for goal in my code, what is criteria to choose goal values for fgoalattain?

Thanks for your help

[Walter Roberson]

You still have

fun=@(s)f;

which tells it to ignore the input s and to always return the content of f, a constant vector. Are you sure that is what you want to do? Is your purpose to try to find a point that violates the constraints the least -- that you do not actually care about the return value of myfun and only want to try to figure out where some points are that (nearly) satisfy the constraints?

Perhaps you want something like

fun = @(s) f*s;

??

[ikram shahzad]

Thanks Thanks to Walter Roberson. my simulation got successful by defining fun=@(s)X(n)*s(1)+x(t)*s(2)+V*C(n)*s(3)+Q(t)*s(4)+QD(t)*s(5)+Ve*s(6)+Vt*s(7)+f(1,u)*s(u); Thank you very much.