The vibration of the internal combustion balance forklift and other types of construction machinery, especially the vibration at the steering wheel, has a significant impact on drivers’ work efficiency and comfort. Since most of the vibration sources of internal combustion forklifts during the working condition are nonstationary, how to accurately identify the nonstationary vibration sources that cause steering wheel vibration is of importance and challenging. This paper proposed an identification method for steering wheel vibration sources based on wavelet coherence analysis. Wavelet packet energy decomposition was used to determine the distribution of vibration energy at the steering wheel of an inter-combustion engine forklift truck for both idling and wide open throttle (WOT) conditions. Coherence analysis and source contribution analysis for the steering wheel vibration were carried out using FFT (coherence) and CWT (wavelet coherence) methods. The two methods and the results from the two conditions were compared and discussed. Results showed that wavelet coherence analysis can better identify the signal feature at low frequencies and provide global contribution analysis for the whole nonstationary conditions. In addition, it was found that the main vibration sources of the steering wheel vibration under the idling and WOT condition were different. A simple confirmation test was further conducted which proved the effectiveness of the proposed method.