Solving lengthy nonlinear complex equations in complex variables

Hello Everyone!

I am trying to solve a set of 4 complex nonlinear equations in 4 complex variables, containing lengthy algebraic expressions with large complex numbers. I have tried vpasolve, solve, fsolve but all get stuck due to voluminous computations involved & no result is appeared. The expressions/equations cannot be shortened or moulded. Please provide suggestions to solve this problem. The code is as under:

_______________________________________________________________________________________________________________________________________

>> syms Y1s Y3s Y4s Y7s

>> e1=(220000/(sqrt(3)))/(((1.8e17*(Y3s*(4.4e63 - 2.0e63i) + Y4s*(6.9e63 - 3.1e63i) + Y7s*(9.4e63 - 4.2e63i) + Y3s*Y4s*(7.8e61 + 1.4e62i) + Y3s*Y7s*(1.1e62 + 1.9e62i) + Y4s*Y7s*(1.6e62 + 2.9e62i) - Y3s*Y4s*Y7s*(5.5e60 - 3.9e60i) - 7.0e64 - 2.1e65i))/(- Y1s*(1.2e82 + 3.7e82i) - Y3s*(1.4e82 + 4.1e82i) - Y4s*(1.5e82 + 4.3e82i) - Y7s*(2.0e82 + 5.5e82i) + Y1s*Y3s*(7.7e80 - 3.4e80i) + Y1s*Y4s*(1.2e81 - 5.4e80i) + Y3s*Y4s*(1.4e81 - 6.2e80i) + Y1s*Y7s*(1.7e81 - 7.4e80i) + Y3s*Y7s*(1.9e81 - 8.6e80i) + Y4s*Y7s*(2.0e81 - 9.2e80i) + Y1s*Y3s*Y4s*(1.4e79 + 2.4e79i) + Y1s*Y3s*Y7s*(1.9e79 + 3.3e79i) + Y1s*Y4s*Y7s*(2.9e79 + 5.1e79i) + Y3s*Y4s*Y7s*(3.4e79 + 5.9e79i) - Y1s*Y3s*Y4s*Y7s*(9.7e77 - 6.9e77i) - 9.7e83 + 2.7e83i))*484)-((3.9046-44.6295*i)*1000);

>> e3=(220000/(sqrt(3)))/(((4.4e16*(Y1s*(1.7e64 - 7.8e63i) + Y4s*(3.2e64 - 1.4e64i) + Y7s*(4.3e64 - 2.0e64i) + Y1s*Y4s*(3.1e62 + 5.4e62i) + Y1s*Y7s*(4.3e62 + 7.5e62i) + Y4s*Y7s*(7.7e62 + 1.3e63i) - Y1s*Y4s*Y7s*(2.2e61 - 1.6e61i) - 3.2e65 - 9.3e65i))/(- Y1s*(1.2e82 + 3.7e82i) - Y3s*(1.4e82 + 4.1e82i) - Y4s*(1.5e82 + 4.3e82i) - Y7s*(2.0e82 + 5.5e82i) + Y1s*Y3s*(7.7e80 - 3.4e80i) + Y1s*Y4s*(1.2e81 - 5.4e80i) + Y3s*Y4s*(1.4e81 - 6.2e80i) + Y1s*Y7s*(1.7e81 - 7.4e80i) + Y3s*Y7s*(1.9e81 - 8.6e80i) + Y4s*Y7s*(2.0e81 - 9.2e80i) + Y1s*Y3s*Y4s*(1.4e79 + 2.4e79i) + Y1s*Y3s*Y7s*(1.9e79 + 3.3e79i) + Y1s*Y4s*Y7s*(2.9e79 + 5.1e79i) + Y3s*Y4s*Y7s*(3.4e79 + 5.9e79i) - Y1s*Y3s*Y4s*Y7s*(9.7e77 - 6.9e77i) - 9.7e83 + 2.7e83i))*484)-((3.4339-39.25*i)*1000);

>> e4=(220000/(sqrt(3)))/(((5.0e3*(Y1s*(1.2e77 + 5.5e75i) + Y3s*(1.4e77 + 5.7e75i) + Y7s*(2.0e77 + 7.0e75i) + Y1s*Y3s*(1.4e74 + 2.5e75i) + Y1s*Y7s*(2.9e74 + 5.2e75i) + Y3s*Y7s*(3.7e74 + 6.1e75i) - Y1s*Y3s*Y7s*(1.1e74 - 1.7e73i) + 5.3e77 - 4.0e78i))/(Y1s*(2.4e81 - 1.7e82i) + Y3s*(2.6e81 - 1.9e82i) + Y4s*(2.7e81 - 2.0e82i) + Y7s*(3.1e81 - 2.6e82i) + Y1s*Y3s*(3.8e80 + 1.6e79i) + Y1s*Y4s*(5.9e80 + 2.8e79i) + Y3s*Y4s*(6.8e80 + 2.8e79i) + Y1s*Y7s*(8.1e80 + 3.4e79i) + Y3s*Y7s*(9.3e80 + 3.4e79i) + Y4s*Y7s*(9.9e80 + 3.5e79i) + Y1s*Y3s*Y4s*(6.9e77 + 1.2e79i) + Y1s*Y3s*Y7s*(1.0e78 + 1.7e79i) + Y1s*Y4s*Y7s*(1.5e78 + 2.6e79i) + Y3s*Y4s*Y7s*(1.8e78 + 3.0e79i) - Y1s*Y3s*Y4s*Y7s*(5.3e77 - 8.5e76i) - 4.4e83 - 8.5e82i))*484)-((1.5165-17.3338*i)*1000);

>> e7=(220000/(sqrt(3)))/(((1.2e4*(Y1s*(1.9e77 + 7.3e76i) + Y3s*(2.2e77 + 8.2e76i) + Y4s*(2.3e77 + 8.7e76i) - Y1s*Y3s*(1.1e75 - 4.2e75i) - Y1s*Y4s*(1.7e75 - 6.4e75i) - Y3s*Y4s*(2.0e75 - 7.4e75i) - Y1s*Y3s*Y4s*(1.3e74 + 2.2e73i) + 2.8e78 - 6.0e78i))/(Y1s*(2.5e82 - 4.9e82i) + Y3s*(2.7e82 - 5.5e82i) + Y4s*(2.8e82 - 5.8e82i) + Y7s*(3.5e82 - 7.5e82i) + Y1s*Y3s*(1.1e81 + 4.2e80i) + Y1s*Y4s*(1.7e81 + 6.7e80i) + Y3s*Y4s*(2.0e81 + 7.7e80i) + Y1s*Y7s*(2.4e81 + 9.2e80i) + Y3s*Y7s*(2.7e81 + 1.0e81i) + Y4s*Y7s*(2.9e81 + 1.1e81i) - Y1s*Y3s*Y4s*(1.0e79 - 3.8e79i) - Y1s*Y3s*Y7s*(1.4e79 - 5.2e79i) - Y1s*Y4s*Y7s*(2.2e79 - 8.0e79i) - Y3s*Y4s*Y7s*(2.5e79 - 9.3e79i) - Y1s*Y3s*Y4s*Y7s*(1.7e78 + 2.7e77i) - 1.2e84 - 7.0e83i))*484)-((4.8371-55.2888*i)*1000);

>> result = solve(e1,e3,e4,e7,Y1s,Y3s,Y4s,Y7s)

_______________________________________________________________________________________________________________________________________

[Star Strider]

You need to do this:

result = solve(e1,e3,e4,e7,[Y1s,Y3s,Y4s,Y7s])
result.Y1s
result.Y3s
result.Y4s
result.Y7s

I ran this for several hours whil I ran some errands, and it is still not doing anything except keeping MATLAB in the ‘Busy’ state for all that time. I will let you continue to experiment with it.

[ubtc]

Thanks for your effort. I have actually six equations in this format, which I need to solve simultaneously; But I am unable to solve only 4 right now. Up to 3, I have done. Is there any other function in matlab to do so, or would I need to swiitch math platform?

[Star Strider]

My pleasure!

I would definitely recommend re-coding these in a way that fsolve can use them. (The fzero function is not an option here.) If you give it complex initial estimates, it will use those and return complex results. If you give it real initial estimates, it will return real results (unless the results can only be complex). It may be necesary to iterate several times with different random initial estimates over a wide range of values, positive and negative, to estimate all of them, assuming they can be uniquely estimated. I would normally suggest plotting it, however with so many parameters, that is not an option. It appears to be first-order (linear) with respect to each parameter (I did not examine the equations closely), so it is likely not necessary to be concerned about complex-conjugate roots.

[ubtc]

Thanks for your suggestions! "fsolve" happens to solve the equations, whereas "vpasolve" remains stuck.

Answer 1

[Alex Sha]

Try to find numerical solutions as below;

y1s.realpart: 11.0169967890388
y1s.imagpart: -125.141575853451
y3s.realpart: 7.81970377305325
y3s.imagpart: -106.356946529629
y4s.realpart: 3.07796056296378
y4s.imagpart: -36.8035908733627
y7s.realpart: 15.4211136064375
y7s.imagpart: -178.419654796614

[ubtc]

Thanks!

Will you please elaborate that whether you solved independently for real & imaginary parts or you split up solution in real & imaginary parts? And what method did you use to get more precision?

[Alex Sha]

Hi, ubtc, the results are obtained by 1stOpt, a math software other than Matlab, much easy for using, and without requiring the guess of initial-start valus for each parameter, the total code like below:

ComplexPar Y1s,Y3s,Y4s,Y7s;
ComplexStr = i;
Function
(220000/(sqrt(3)))/(((1.8e17*(Y3s*(4.4e63 - 2.0e63*i) + Y4s*(6.9e63 - 3.1e63*i) + Y7s*(9.4e63 - 4.2e63*i) + Y3s*Y4s*(7.8e61 + 1.4e62*i) + Y3s*Y7s*(1.1e62 + 1.9e62*i) + Y4s*Y7s*(1.6e62 + 2.9e62*i) - Y3s*Y4s*Y7s*(5.5e60 - 3.9e60*i) - 7.0e64 - 2.1e65*i))/(- Y1s*(1.2e82 + 3.7e82*i) - Y3s*(1.4e82 + 4.1e82*i) - Y4s*(1.5e82 + 4.3e82*i) - Y7s*(2.0e82 + 5.5e82*i) + Y1s*Y3s*(7.7e80 - 3.4e80*i) + Y1s*Y4s*(1.2e81 - 5.4e80*i) + Y3s*Y4s*(1.4e81 - 6.2e80*i) + Y1s*Y7s*(1.7e81 - 7.4e80*i) + Y3s*Y7s*(1.9e81 - 8.6e80*i) + Y4s*Y7s*(2.0e81 - 9.2e80*i) + Y1s*Y3s*Y4s*(1.4e79 + 2.4e79*i) + Y1s*Y3s*Y7s*(1.9e79 + 3.3e79*i) + Y1s*Y4s*Y7s*(2.9e79 + 5.1e79*i) + Y3s*Y4s*Y7s*(3.4e79 + 5.9e79*i) - Y1s*Y3s*Y4s*Y7s*(9.7e77 - 6.9e77*i) - 9.7e83 + 2.7e83*i))*484)-((3.9046-44.6295*i)*1000);
(220000/(sqrt(3)))/(((4.4e16*(Y1s*(1.7e64 - 7.8e63*i) + Y4s*(3.2e64 - 1.4e64*i) + Y7s*(4.3e64 - 2.0e64*i) + Y1s*Y4s*(3.1e62 + 5.4e62*i) + Y1s*Y7s*(4.3e62 + 7.5e62*i) + Y4s*Y7s*(7.7e62 + 1.3e63*i) - Y1s*Y4s*Y7s*(2.2e61 - 1.6e61*i) - 3.2e65 - 9.3e65*i))/(- Y1s*(1.2e82 + 3.7e82*i) - Y3s*(1.4e82 + 4.1e82*i) - Y4s*(1.5e82 + 4.3e82*i) - Y7s*(2.0e82 + 5.5e82*i) + Y1s*Y3s*(7.7e80 - 3.4e80*i) + Y1s*Y4s*(1.2e81 - 5.4e80*i) + Y3s*Y4s*(1.4e81 - 6.2e80*i) + Y1s*Y7s*(1.7e81 - 7.4e80*i) + Y3s*Y7s*(1.9e81 - 8.6e80*i) + Y4s*Y7s*(2.0e81 - 9.2e80*i) + Y1s*Y3s*Y4s*(1.4e79 + 2.4e79*i) + Y1s*Y3s*Y7s*(1.9e79 + 3.3e79*i) + Y1s*Y4s*Y7s*(2.9e79 + 5.1e79*i) + Y3s*Y4s*Y7s*(3.4e79 + 5.9e79*i) - Y1s*Y3s*Y4s*Y7s*(9.7e77 - 6.9e77*i) - 9.7e83 + 2.7e83*i))*484)-((3.4339-39.25*i)*1000);
(220000/(sqrt(3)))/(((5.0e3*(Y1s*(1.2e77 + 5.5e75*i) + Y3s*(1.4e77 + 5.7e75*i) + Y7s*(2.0e77 + 7.0e75*i) + Y1s*Y3s*(1.4e74 + 2.5e75*i) + Y1s*Y7s*(2.9e74 + 5.2e75*i) + Y3s*Y7s*(3.7e74 + 6.1e75*i) - Y1s*Y3s*Y7s*(1.1e74 - 1.7e73*i) + 5.3e77 - 4.0e78*i))/(Y1s*(2.4e81 - 1.7e82*i) + Y3s*(2.6e81 - 1.9e82*i) + Y4s*(2.7e81 - 2.0e82*i) + Y7s*(3.1e81 - 2.6e82*i) + Y1s*Y3s*(3.8e80 + 1.6e79*i) + Y1s*Y4s*(5.9e80 + 2.8e79*i) + Y3s*Y4s*(6.8e80 + 2.8e79*i) + Y1s*Y7s*(8.1e80 + 3.4e79*i) + Y3s*Y7s*(9.3e80 + 3.4e79*i) + Y4s*Y7s*(9.9e80 + 3.5e79*i) + Y1s*Y3s*Y4s*(6.9e77 + 1.2e79*i) + Y1s*Y3s*Y7s*(1.0e78 + 1.7e79*i) + Y1s*Y4s*Y7s*(1.5e78 + 2.6e79*i) + Y3s*Y4s*Y7s*(1.8e78 + 3.0e79*i) - Y1s*Y3s*Y4s*Y7s*(5.3e77 - 8.5e76*i) - 4.4e83 - 8.5e82*i))*484)-((1.5165-17.3338*i)*1000);
(220000/(sqrt(3)))/(((1.2e4*(Y1s*(1.9e77 + 7.3e76*i) + Y3s*(2.2e77 + 8.2e76*i) + Y4s*(2.3e77 + 8.7e76*i) - Y1s*Y3s*(1.1e75 - 4.2e75*i) - Y1s*Y4s*(1.7e75 - 6.4e75*i) - Y3s*Y4s*(2.0e75 - 7.4e75*i) - Y1s*Y3s*Y4s*(1.3e74 + 2.2e73*i) + 2.8e78 - 6.0e78*i))/(Y1s*(2.5e82 - 4.9e82*i) + Y3s*(2.7e82 - 5.5e82*i) + Y4s*(2.8e82 - 5.8e82*i) + Y7s*(3.5e82 - 7.5e82*i) + Y1s*Y3s*(1.1e81 + 4.2e80*i) + Y1s*Y4s*(1.7e81 + 6.7e80*i) + Y3s*Y4s*(2.0e81 + 7.7e80*i) + Y1s*Y7s*(2.4e81 + 9.2e80*i) + Y3s*Y7s*(2.7e81 + 1.0e81*i) + Y4s*Y7s*(2.9e81 + 1.1e81*i) - Y1s*Y3s*Y4s*(1.0e79 - 3.8e79*i) - Y1s*Y3s*Y7s*(1.4e79 - 5.2e79*i) - Y1s*Y4s*Y7s*(2.2e79 - 8.0e79*i) - Y3s*Y4s*Y7s*(2.5e79 - 9.3e79*i) - Y1s*Y3s*Y4s*Y7s*(1.7e78 + 2.7e77*i) - 1.2e84 - 7.0e83*i))*484)-((4.8371-55.2888*i)*1000);

The outcome will be:

y1s.realpart: 11.0169967890388
y1s.imagpart: -125.141575853451
y3s.realpart: 7.81970377305327
y3s.imagpart: -106.356946529629
y4s.realpart: 3.07796056296379
y4s.imagpart: -36.8035908733627
y7s.realpart: 15.4211136064375
y7s.imagpart: -178.419654796614

Answer 2

[ubtc]

Thank you, everyone!

I have solved the problem by using fsolve. (I had not tried this properly earliear). The answer is as following:-

ans =
 
   1.0e+02 *
 
   0.1102 - 1.2514i   0.0782 - 1.0636i   0.0308 - 0.3680i   0.1542 - 1.7842i

It is same as pointed out by Mr. Alex Sha above but only limited to four decimal places. However, as you can see in original equations, I have simplified the equations using simplify & Variable-precision arithmetic (up to 2 decimal points)

eq = simplify(eq,'Steps',10)
eq = vpa(eq, 2)

I shall definitely try to solve without using "vpa" via "fsolve" method. However, Is there some more accurate way other than "fsolve", to solve such lengthy nonlinear equations in complex variables (more than 8 equations in same no. of variables) with more precision without using "vpa"?

[Star Strider]

I wish now that I had posted my Comment as an Answer!