PV arrays are usually affected by a partial shading condition, which leads to a relatively low power production. This is because the power-voltage curve of a PV system contains multiple peaks while the traditional Maximum Power Point Tracking (MPPT) algorithm is easily trapped at the Local Maximum Power Point (LMPP). Hence, a novel MPPT approach is provided, i.e., Dynamic Leader-based Collective Intelligence (DLCI). Unlike traditional meta-heuristic algorithms, this algorithm has a multiple sub-optimizer which seeks the optimum independently. Then, the current best optimum will be chosen as the dynamic leader to guide the other sub-optimizers thereafter. Three case studies are carried out, i.e., constant climate conditions, varying climate conditions, and a large-scale photovoltaic station. Simulation outcomes of Matlab/Simulink prove that DLCI outperforms the traditional Incremental Conductance (INC) and five other typical meta-heuristic algorithms. It can achieve the fastest and most stable global MPPT. Lastly, a dSpace based Hardware-In-the-Loop (HIL) test is carried out to validate the implementation feasibility of the DLCI algorithm. This work is supported by the National Natural Science Foundation of China (No. 61963020).