Non-intrusive load monitoring technique could help air source heat pump (ASHP) load data collection for its benefits of lower cost and easier installation. This paper proposed a load dactylogram extracting method of ASHP based on load event monitoring. By capturing transit events, the key parameters ΔT and ΔP are obtained, according to which ASHP starting state is recognized. A practical method of automatic generation of load dactylogram is proposed through adaptive learning, the load dactylogram library is formed by clustering identified load events. Accordingly, ASHP load can be identified and its electric quantity can be calculated. Compared with the load results of coal-to-power users measured by sensor, the running result shows that the start-stop-times of ASHP are correct and the accuracy of power consumption is above 84%, and the dactylogram template library of ASHP is more unique.