Artificial neural network (ANN) has emerged as a promising tool for component modeling in power electronics due to its high accuracy and flexibility. However, the availability of large training data is an essential requirement for reliable predictions given by ANN, which is often hard to satisfy in power electronics applications. Therefore, this paper proposes a novel loss modeling approach based on knowledge-aware ANN for planar magnetic components, which is implemented by combining small training data with specific domain knowledge. After investigating the principles of ANN, theoretical explanations of knowledge-aware ANN are illustrated and several cases are performed, both of which validate the generalization of the proposed method. Compared with existing modeling tools, the results show that the proposed method realizes an excellent balance between accuracy and computational efficiency.