One of the basic elements that any plant process facility consists of is control tank system. Custom tank is one type of these control tank systems. Proper controlling of this element will help the facility to work smoothly and it will increase the reliability of the whole system. This study is looking into predicting the state of variables that completely represent the dynamics the custom tank (height of fluid or output flow rate). This prediction can be used in controlling the custom tank (predictive control). The study involve MATLAB SIMULINK simulation program for the custom tank along with different prediction models. The obtained results showed that introducing Multilayer Perceptron (MLP) Neural Network architecture improve the prediction significantly where different algorithms, Recursive Kalman Filter (RKF) and Extended Kalman Filter (EKF) have been used simultaneously to estimate fluid height and output flow. It further shows that introducing centered finite difference or derivative free with EKF improve the performance of the network.