The measurement of flame velocity is one of the research interests on the combustion process and fluid diagnosis, as it is of key research significance for the analysis of combustion composition and propulsion dynamics. In spite of its high accuracy, laser Doppler velocity measurement has limited application due to the complexity of the measurement process and increased error rate in low-speed measurement, hence giving rise to the use of photothermal deflection spectroscopy for low and medium-speed fluid measurement. Photothermal deflection spectroscopy is based on the detection of the thermal lens on the tested medium. As components in the fluid medium absorb light and form a thermal lens distribution, when a probe beam is an incident on the medium, it is deflected due to the movement of the thermal lens. Fluid velocity is obtained by measuring the height of the probe beam relative to the pump beam, as well as the flight time corresponding to the signal deflection. This paper adopted a self-built pump device to study photothermal deflection. A pump beam with a single pulse energy of 20 mJ and wavelength of 355 nm, and He-Ne laser probe with the power of 2 mW were used to measure velocity at different positions of kerosene flame. The device has a spatial resolution of 2×10^-5 cm³. Velocity was measured at planes with distances of 5, 8 and 11 mm from the kerosene wick, to obtain the distribution of horizontal velocity corresponding to the flame. The external velocity of the flame at the same horizontal plane was found to be higher than the internal velocity near the bottom of the flame; in the position near the top of the flame, the internal velocity of the flame at the same horizontal plane was higher than the external velocity; velocity distribution along the same plane was close to parabolic distribution. Velocity distribution of the three vertical planes ±2 mm away from the center of the flame was measured, to obtain the distribution of the corresponding vertical planes. Velocity on the central vertical axis near the bottom of the flame was found to be slower than that of both sides, and the velocity of the upper part of the flame was faster than that of both sides. This was consistent with the conclusions obtained from the earlier measurements of horizontal velocity distribution. The flame speed measured in the experiment ranged from 0.2 to 1.5 m·s^-1. This paper used pump beams with single-pulse energy of 20, 40 and 60 mJ, to analyze the errors introduced by dielectric breakdown during velocity measurement. Results indicate that a larger error is introduced by higher laser energy. A velocity error of 0.1 m·s^-1 was introduced by the 40 mJ beam, while a velocity error of 0.6 m·s^-1 was introduced by the 60 mJ beam. With the further optimization of signal-to-noise ratio, photothermal deflection spectroscopy shall enable the measurement of parameters, such as temperature and concentration, thus making it a powerful tool for fluid velocity measurement and combustion diagnostics.