Factories have always been leveraging state-of-the-art technologies since its inception. At present, Artificial Intelligence(AI) technology is being developed rapidly and also being used in various domains. Factories have also started to adapt this trend of using AI. AI basically uses historical data to train amodel, so that the trained model can be used to define the relationship among different vectors. Thus,this relationship can be used to predict future states of physical system in factories. However, in case offactories, just using data with AI might not be the optimal way to predict the future states because wedo have a good domain knowledge on how these machines work. As we use machines which are built ontop of solid mathematical backgrounds, there is a good opportunity to use it along with AI for predictingfuture states of physical system. Using only mathematical models for prediction will be not be accurateenough because real machines do not work exactly as defined in the theories. On the other hand, usingAI with historical data alone, we can train a model to mimic those factories but it will require a lot ofsensor data of various parameters, which might not be always feasible in real world industries. Instead ofbuilding a model on the basis of historical data alone, we propose an alternative way to use both domainknowledge and AI in conjunction. In this work, we studied a power plant and implemented the idea ofusing both domain knowledge and AI to predict future states of physical systems used in it.