Model Buildings.Electrical.DC.Loads.Examples.LinearizedLoad fails in OM due to a wrong initialization

[AndreaBartolini]

The simulation of the model Buildings.Electrical.DC.Loads.Examples.LinearizedLoad fails in OM by throwing the following error:

 division by zero at time 0, (a=-4999.489847974696) / (b=-0), where divisor b expression is: sen_nlin.P

more precisely, the error is thrown during the lambda-0 step of the homotopy process:

homotopy process ---------------------------
homotopy parameter lambda = 0
division by zero at time 0, (a=-4999.489847974696) / (b=-0), where divisor b expression is: sen_nlin.P

The equation that generates the issue is the following, placed at the top-level in the model:

Real error = (sen_nlin.P - sen_lin.P)*100/sen_nlin.P "Percentage of error between the linearized and actual power consumption";

where sen_nlin is the sensor connected in series to the NonlinearLoad, as shown in the following figure:

The problem arises because the NonlinearLoad is initialized by using the default zero_current initMode, that prescribes the current to be zero in the entire branch during the lambda-0 step, as shown in the initial equations of the NonlinearLoad model (i.e., of the Conductor model):

if initMode == Buildings.Electrical.Types.InitMode.zero_current then
  i[1] = - homotopy(actual= P/(v[1] - v[2]),  simplified= 0);
else
  i[1] = - homotopy(actual= P/(v[1] - v[2]),  simplified= P*(2/V_nominal - (v[1]-v[2])/V_nominal^2));
end if;
...
sum(i) = 0;

thus leading the sen_nlin.P=0 from the sensor sen_nlin during the lambda_0 step.

One solution can be to set the initMode to linearized in the NonlinearLoad, so that P is not equal to zero during the lambda-0 step.

BTW Dymola seems to be able to manage this singularity, but the right way is to remove them.

[mwetter]

That for reporting. I will correct it and also in IBPSA (ibpsa/modelica-ibpsa#1942)