Public transportation has significantly reduced operational costs and carbon emissions through the adoption of electric buses. Efficient charging schedules plays a crucial role in achieving cost savings for electric bus operations. However, reliance solely on the conventional power grid as the energy source poses limitations due to its overdependence on non-renewable energy, such as fossil fuels. This study focuses on an innovative scenario where photovoltaic-storage-charging (PSC) stations supply electricity to an electric bus fleet, presenting a novel and promising charging scheduling challenge. Unlike previous studies, this research not only addresses when and where buses are charged but also optimizes electricity allocation within PSC stations to minimize the total charging costs of a bus fleet. We formulate a mixed integer programming with time discretization technique and a time-expanded network to accommodate various routes, vehicles, and charging stations. Charging costs are assessed based on photovoltaic generation expenses and time-of-use (TOU) electricity tariffs. To address this problem, a Lagrangian relaxation-based framework is devised. Additionally, a dynamic programming algorithm incorporating a bi-criteria labeling method is developed to solve the Lagrangian relaxation problem. Data from an electric bus fleet in Jiading, Shanghai, is collected as a case study to validate the model and algorithm. Optimization results indicate that electric buses primarily utilize electricity generated by photovoltaic panels during daytime operations. Upon returning to the depot at night, they prioritize the depletion of stored energy in the onsite batteries. In cases of inadequate supply, the power grid steps in to meet operational charging demands.