Abstract: The study of ecological driving strategy in the merging zone of urban roads needs to consider the characteristics of vehicle speed in real scenarios. By analyzing the cumulative frequency, distribution trend, characteristic percentile value and other statistical characteristics of vehicle speed, the differences of vehicle speed in the merging zone during flat/peak hours and the operational characteristics of vehicle speed and acceleration in the merging zone during peak hours were derived. Finally, a multiple linear regression speed estimation model based on the relationship between longitudinal acceleration (deceleration) speed and vehicle speed was constructed. Results show that the distribution of vehicle speed in the intertwined zone during peak hours is a relatively concentrated left-skewed distribution, while the speed in the intertwined zone during the flat hours basically conforms to the Gaussian distribution. The speed disorder in the merging area during peak hours is distributed at 0, 20 km/h, and the orderly distribution of vehicle speed in the merging area during the peak hours is distributed at 7.5, 45.0 km/h. The scatter distribution of longitudinal acceleration versus speed in the peak hour merging area is unequal-waisted triangle, while the scatter distribution of longitudinal deceleration versus speed in the peak hour merging area is right-angled triangle. The longitudinal acceleration quantile shows an M-shaped and the longitudinal deceleration quantile shows an obvious rising trend. The degree of fit index reaches more than 79%. The accuracy of the speed state estimation model constructed in this study meets the requirements, and can provide some theoretical reference for the research of eco-driving speed estimation.