Load Disaggregation of Green Aviation Based on Interpretability Decision Tree and Hidden Markov Model

The development of green aviation relies on the effective monitoring and understanding of equipment energy consumption data. However, it is often challenging to balance the accuracy and interpretability of device energy consumption disaggregation. To address this issue, this paper proposes a bi-level interpretable load disaggregation model based on decision tree (DT) and hidden Markov model (HMM). The model first constructs a device feature library, including the power characteristics and state transition probability matrix of various devices under different operating states, providing a data foundation for load disaggregation. On the one hand, a pre-pruned DT is employed to establish an explicit mapping relationship between total power features and device operating states. On the other hand, HMM is used to model the temporal dependencies of device state sequences and quantify the state transition patterns, thereby constructing a feasible state space with state transition probability constraints. Based on the bi-level constraints, the model aims to minimize the power matching error and maximize the rationality of state transitions to obtain the optimal state combination of the devices. Preliminary experiments show that our model can efficiently search for the optimal operating state combinations, reaching an average relative error of 3.5%. The proposed model provides reliable and temporally interpretable technical support for energy consumption monitoring and management of electrical devices.