sdpvar index multiplication problem

[Qianyuan Zheng]

Hi everyone,

I met a problem when I want sdpvar(sdpvar) * sdpvar(sdpvar).

Part of my code is as follow:

price_f=rand(48,1)

t_var = intvar(1,1);

p_ev = sdpvar(48,1);

price_el=sdpvar(48,1);

% objective

price_ev = 0;

power = 0;

for i =0:4

    

    power = power + p_ev(t_var+i);

    price_ev = price_ev + p_ev(t_var+i) * price_el(t_var+i);

    

end

% constr

price_constr = [price_el == price_f];

where p_ev(t_var+i) and price_el(t_var+i) are both "Nonlinear scalar (real, models 'milpsubsref', 1 variable)", 
but p_ev(t_var+i) * price_el(t_var+i) has no output.

[Johan Löfberg]

Why are you adding price_el == price_f afterwards?

The model you create now is extremely hard and bad (decision variables in indexing and bilinear products)

If you really have price_el = price_f, a much better model (linear) is to simply enumerate the cases using a binary representation

which = binvar(1,48);

t_var = which*(1:48)';

powers = sdpvar(5,1);

price_evs = sdpvar(5,1);

Model = [sum(which)==1];

for i = 0:4

    for j = 1:48

        if j + i <= 48

            Model = [Model, implies(which(j),powers(i+1) == p_ev(j + i))];

            Model = [Model, implies(which(j),price_evs(i+1) == p_ev(j + i)*price_f(j+i))];

        end

    end

end

power = sum(powers);

price_ev = sum(price_evs);