Accurately counting the number of pests killed by solar insecticidal lamp (SIL) is crucial as it provides valuable insights into local pest density. However, a challenge arises when pests adhere to the high voltage metal mesh, causing continuous discharging and resulting in inaccurate counting using existing voltage counting methods. To address this issue, this research proposes a novel method that combines the counting of discharge sound and discharge voltage. The proposed method analyzes the discharge sound signal in the time domain, utilizing the time difference between pulses to filter out redundant data generated by discharges. The processed discharge sound data and discharge voltage data are jointly counted by machine learning, which significantly improves the accuracy and feasibility of counting. Comparative analysis shows that the accuracy of this method is 90.02%, which is 11.24% higher than that of voltage counting method (78.78%) and 34.63% higher than that of comparison method (55.39%). Finally, the accuracy of verification in real farmland environment is 93.28%. These results demonstrate the effectiveness of this paper method in achieving accurate pest counting. Furthermore, this paper method boasts lightweight implementation and easy deployment in consumer electronic devices such as SIL used for pest early warning.