To improve the accuracy and robustness of day-ahead bus load forecasting

this paper proposes a hybrid method integrating optimal similar-day sets and deep learning.First

using bus load and meteorological data

it scores similar day via gray relational analysis and entropy weight method to form the optimal similar-day set.Next

multiple sets of extreme learning machine(ELM)model parameters are trained

and the weights and thresholds of ELM are optimized using a mutation fruit fly optimization algorithm(MFOA)to enhance the robustness of the model..Finally

it introduces a deviation degree of prediction error(DPE)objective function for the DPE-MFOA-ELM model to strengthen the bus load forecasting applicability.Simulation tests on 10 diverse bus load groups show that

compared to conventional deep machine models

the proposed method improves the prediction accuracy and robustness of bus load.