Setting an adequate operating power reserve (PR) to compensate unpredictable imbalances between generation and consumption is essential for power system security. Operating power reserve should be carefully sized but also ideally minimized and dispatched to reduce operation costs with a satisfying security level. Although several energy generation and load forecasting tools have been developed, decision-making methods are required to estimate the operating power reserve amount within its dispatch over generators during small time windows and with adaptive capabilities to markets, as new ancillary service markets. This paper proposes an uncertainty analysis method for power reserve quantification in an urban microgrid with a high penetration ratio of PV (photovoltaic) power. First, forecasting errors of PV production and load demand are estimated one day ahead by using artificial neural networks. Then two methods are proposed to calculate one day ahead the net demand error. The first perform a direct forecast of the error, the second one calculates it from the available PV power and load demand forecast errors. This remaining net error is analyzed with dedicated statistical and stochastic procedures. Hence, according to an accepted risk level, a method is proposed to calculate the required PR for each hour. (C) 2017 Elsevier Ltd. All rights reserved.