How to tune the prameters of ddpg agent ?

Dear MathWorks Community:

Sorry to interrupt you! I am a beginner of reinforcement learning and recently studied the example case "rlwatertank". Right now, I try to replicate the procedure and use it in another control problem which is very similiar to the water level control problem in "rlwatertank". The simulation time is Tf=200, and the sampling time is Ts=1. Nevertheless, during the training, the episode steps remain 2, which means it just didn't train at all before one episode terminates, therefore, the reward keeps around negative 100 (the minimum)

I am very confused why it didn't train.