1: \begin{abstract}

2: In this paper, we present a probabilistic numerical algorithm combining

3: dynamic programming, Monte Carlo simulations and local basis regressions

4: to solve non-stationary optimal multiple switching problems in infinite

5: horizon. We provide the rate of convergence of the method in terms

6: of the time step used to discretize the problem, of the size of the

7: local hypercubes involved in the regressions, and of the truncating

8: time horizon. To make the method viable for problems in high dimension

9: and long time horizon, we extend a memory reduction method to the

10: general Euler scheme, so that, when performing the numerical resolution,

11: the storage of the Monte Carlo simulation paths is not needed. Then,

12: we apply this algorithm to a model of optimal investment in power

13: plants. This model takes into account electricity demand, cointegrated

14: fuel prices, carbon price and random outages of power plants. It computes

15: the optimal level of investment in each generation technology, considered

16: as a whole, w.r.t. the electricity spot price. This electricity price

17: is itself built according to a new extended structural model. In particular,

18: it is a function of several factors, among which the installed capacities.

19: The evolution of the optimal generation mix is illustrated on a realistic

20: numerical problem in dimension eight, i.e. with two different technologies

21: and six random factors. 

22: \end{abstract}

23: