The status is Inaccurate/Solved

1

My program:

cvx_solver sedumi
cvx_begin
variable PEGU(length(y1));%length(y1)=40513
variable PB(length(y1));
variable E(length(y1)+1);
variable nb;
expression Vqihao(length(y1));
expression Jsunhao(length(y1));
expression dE(length(y1));
expression Tm1(length(y1));
expression Pr(length(y1));
Tm1=(delta11*(m+4.5nb).a+(m+4.5nb)gdelta+CDA/21.15y1.^2)r/2/i0;
Pr=max(Tm1.yata1,-30);%Pr=max(Tm1.yata1,-30/0.94);
dE=E(2:length(y1)+1)-E(1:length(y1));
Vqihao111=8.6510^(-5)PEGU;
Vqihao222=0.0001685PEGU+8.6510^(-5)83.4220-(0.000168583.4220-0.007009)-0.007009;
Vqihao=max(Vqihao111,Vqihao222);
Jsunhao=p00+p01nb+p20dE.^2+p02nb.^2+pp00+pp10abs(dE)+pp01nb;
minimize(sum(cfVqihao)+sum(Jsunhao)+ce*(SOC0-SOCf)nbQb3.2/1000);
subject to
E(1)==C/2nb*(pp(1)SOC0+pp(2))^2;
E(length(y1)+1)==C/2nb*(pp(1)SOCf+pp(2))^2;
Pr<=PB+PEGU;
0<=PEGU<=109.4;
140<=nb<=260;
(Imingeo_mean([2/CE(1:length(y1)),nbones(length(y1),1)],2)-RImin^2nb)/1000<=PB<=E(1:length(y1))/2/R/C/1000;
1000CPB-Imax^2RCnb-Imaxgeo_mean([2C(nbones(length(y1),1)),E(1:length(y1))-2000RCPB],2)<=0;
C/2nb(pp(1)SOCmin+pp(2))^2<=E<=C/2nb*(pp(1)SOCmax+pp(2))^2;
geo_mean([E(1:length(y1)),E(1:length(y1))-2000RCPB],2)>=3600RQb/pp(1)*dE+E(1:length(y1));
cvx_end

The result:
iter seconds digits cx by
89 463.1 4.2 -5.5899036462e+01 -5.5902928336e+01
|Ax-b| = 7.3e-07, [Ay-c]_+ = 5.0E-11, |x|= 2.3e+03, |y|= 1.6e+07

Detailed timing (sec)
Pre IPM Post
2.669E+02 3.291E+02 4.400E-01
Max-norms: ||b||=3.670000e+00, ||c|| = 260,
Cholesky |add|=2, |skip| = 1, ||L.L|| = 13397.2.

Status: Inaccurate/Solved
Optimal value (cvx_optval): +61.3584

I want to know why the result is Inaccurate/Solved and how I can do for this question.
Thank you.

2

The size of dual solution is quite large (|y|=1.0e7) that often cause numerical issues.

You could try Mosek and/or improve the numerics of your model.

3

Hi,Erling.
Is the meaning of ‘improve the numerics of your model’ trying to change the range of ’ 0<=PEGU<=109.4’ or ’ 140<=nb<=260’?

4

I do not know if that will make change.

You should ask yourself why is the dual solution is large.

Improving the numerics requires careful analysis of your model. It is kind of an art and a time consuming thing that you are unlikely to get help with here. Hard work is the way forward.