I am doing an experiment on power system state estimation using neural network  for IEEE 14 bus system with fully observed. Fully observe means all sensors are working and all measured data  from sensors are transmitted to the control center. So I set the input as measured data (available measurements) and target as the true state variables (load angle and magnitude voltage), however  the measured data (input) has different samples from the target (state variables), so can we set every type of input as vector. For example, active power injections as a first vector (first row) in the input matrix, reactive power injections as a second vector (second row) and the same for active power flow and reactive power flow. Or it is wrong and we have to set all measurements as one vector as we do when we use weighted least square, kalman filter, linear programing and so on.