Conventional optimization approaches to generator reactive power reserve are mostly based on the deterministic load increase pattern. However

this pattern is unable to effectively evaluate the reactive power reserve level under the most critical load increase pattern when the forecast errors of load and renewable energy output are taken into account. To address this problem

a nonlinear bilevel programming model of optimal generator reactive power reserve dispatch method based on the closest critical point of voltage stability limit surface is proposed. The upper model transforms the solution of critical point and its load increase pattern into the iteration process of second-order approximation of limit surface and continuation power flow under the given nodal power injection. The lower model solves the reactive power reserve under the critical load increase pattern

while the modified power injections are returned to the upper level. To demonstrate the validity of the proposed method

first the existence of optimal solution and the effectiveness of convergence criteria are verified. Further

it is shown by an analysis of the numerical simulation that the proposed method is able to improve reactive power level and enhance the voltage stability margin of power system under the critical load increase pattern considering the uncertainties of load increase pattern. This work is supported by the Key Program of National Natural Science Foundation of China(No. 51337005)and National Basic Research Program of China(973 Program)(No. 2014CB23903).