A new method is presented for the optimal control design of one-dimensional heat equation with the actuators being concentrated at discrete points in the spatial domain. This systematic methodology incorporates the advanced concept of proper orthogonal decomposition for the model reduction of distributed parameter systems. After deigning a set of problem oriented basis functions an analogous optimal control problem in the lumped domain is formulated. The optimal control problem is then solved in the time domain, in a state feedback sense, following the philosophy of adaptive critic neural networks. The control solution is then mapped back to the spatial domain using the same basis functions. Numerical simulation results are presented for a linear and a nonlinear one-dimensional heat equation problem.