The output power of photovoltaic panels fluctuates significantly with the temperature and radiation intensity. To achieve safe and stable energy harvesters, it is critical to keep the output voltage of photovoltaic panels at the maximum power point (MPP). To eliminate the limitation of both a fixed step size and the low response speed caused by the small average step size in ordinary perturbation and observation (P&O), a new sub-area-based improved variable-step-size perturbation and observation (SVS-P&O) algorithm is proposed. The working area of a power generation system is divided into two different parts using the partition function. When the operating point of the photovoltaic system is far from the MPP, a larger and fixed step size is adopted to improve the tracking speed. However, when the operating point of the photovoltaic system is close to the MPP, a variable and smaller step size is adopted to reduce the fluctuation of steady-state power. Different perturbation step sizes are used in different areas, and the step size is adjusted in real-time to track the MPP through the combination of fixed and variable step sizes. The simulation results show that the proposed SVS-P&O algorithm for MPP tracking has less oscillation at the MPP than the conventional P&O method and double the tracking speed.