Transfer learning-based physics-informed convolutional neural network for simulating flow in porous media with time-varying controls