In recent years, behind-the-meter distributed photovoltaic (PV) systems have developed rapidly, and their unobservable characteristics pose huge challenges to demand response aggregated baseline load (ABL) estimation. To improve the ABL estimation accuracy under high penetration of distributed PVs, an ABL estimation method based on identification and decoupling is proposed. First, a PV user identification and decoupling method based on weather state-driven features is proposed. This separates PV users from non-PV users. Secondly, estimation models are established respectively according to the characteristics of two types of users' net load: the estimation model based on weather type classification is established for PV users, and the piecewise linear regression model of temperature and load is established for non-PV users. Finally, a final ABL estimation is obtained by adding the estimated values of the two types of users. The simulation results show that, compared with the direct estimation methods and other classification estimation methods, the proposed method has an average decrease of 42.07% on the mean absolute percentage error (MAPE) and an average decrease of 23.93% on the normalized root mean square error (NRMSE). This work is supported by the Science and Technology Project of State Grid Corporation of China (No. SGHEYX00SCJS2000037).