Continuous control problem with reinforcement learning

Hi everyone, I'm trying to control my microgrid model by using RL. I know there's a water tank example in Matlab/Simulink, what it does is replacing the PID controller with an RL agent, similarly, I also created my Simulink model and implement the RL agent with the environment and observation blocks.

My questions are:

1) In a similar water tank problem (it has a scaler desired value as 10), I am giving a load as a scalar value, but I actually want to give a vector in the same length with simulation time and use the corresponding load at each second during the simulation.

2) My observation and action have upper and lower bounds and I defined them by following lines:

obsInfo = rlNumericSpec([3 1],...

'LowerLimit',[0 0 0]',...

'UpperLimit',[1000 1000 250]');

However, my simulation does not follow the upper bound for the third bound.

Do you have any idea about these things?

Thanks for any suggestions in advance.