This numerical study deals with the estimation of both unknown position and intensity of a heat source in diffusive problems, from the knowledge of local temperature data. The source is assumed to be fixed but its intensity varies with time. The originality of this paper lies in the use of reduced models to solve the inverse problem. The source position being unknown, a specific approach is proposed, involving the Modal Identification Method (MIM) allowing us to obtain a RM relative to a set of output temperatures. The source is modelled as f(r,rs)u(t), where f covers the whole domain and mimics a source centred in rs. Starting with initial guess for rs, RMs relative to outputs and their first derivatives with respect to rs, are identified. A Quasi-Newton algorithm is used for searching rs, and according to a Taylor expansion, new RMs are built for current rs to estimate u(f) and compute sensitivities. When rs cannot be modified anymore by the iterative algorithm, the Detailed Model is called to update the RMs series. The approach is first described in detail for a ID case, then expressions for 2D and 3D cases are given. An academic 3D heat diffusion problem illustrates the method.