In this paper, we consider the problem of sensor selection for estimating an unknown parameter in an energy constrained network. We assume that measurement noises are correlated and thus sensor observations are conditionally dependent given the underlying parameter. To determine optimal sensor activations, we propose a tractable framework in which the estimation error is minimized subject to an energy constraint. The resulting optimization problem is nonconvex in nature, and we develop both a convex relaxation approach and a greedy algorithm to find its near-optimal solutions. We point out that the prior formulations in the literature is valid only for the case of weakly correlated noise. We show that the problem of sensor selection with weak noise correlation, that maximizes the trace of Fisher information, can be transformed into the problem of maximizing a convex quadratic function over a bounded polyhedron. This problem structure simplifies the optimization approaches employed for sensor selection. Furthermore, we generalize our framework of sensor selection to solve the problem of sensor scheduling, where a greedy algorithm is proposed to determine non-myopic (multi-time ahead) sensor schedules. Lastly, numerical results are provided to illustrate the effectiveness of our approaches, and to reveal the effect of noise correlation on the estimation performance.