The paper proposes bi-level optimisation framework for electric vehicle (EV) fleet charging based on realistic EV fleet and transport demand model. The EV fleet is modelled as a single so-called aggregate battery and parameterised by using recorded data of a particular delivery vehicle fleet. This EV fleet model is used within the inner level of bi-level optimisation framework, where the aggregate charging power variable is optimised by using the dynamic programming (DP) algorithm. In the superimposed level of optimisation framework, the final state-of-charge (SoC) values of EVs being disconnected from the grid are optimised by using a multi-objective genetic algorithm-based optimisation. In each iteration of bi-level optimisation, it is needed to recalculate transport demand-related input time distributions of the aggregate battery model. To simplify this process, transport demand is modelled by using a computationally efficient response surface method, which is based on naturalistic synthetic driving cycles and agent-based simulations of EV model. The bi-level optimisation framework represents the extension of the single-level optimisation thus enabling the multi-parameter optimisation of the considered transport-energy system as well as optimisation of different economic-related aspects, e.g. investment vs. operational costs. The bi-level optimisation approach is validated by comparing its optimisation results with the previously obtained results based on a single-level optimisation approach where the final SoC values were fixed to 100%.