Swing options have emerged as efficient tools for managing power price risk. Bundled with electricity forward contracts, they address the need for some structural flexibility in both the time and volume of the contracted delivery. We present an efficient numerical method to evaluate power swing options, using a state-space forest in combination with an analytic approximation of the conditional densities of the underlying processes. We consider a multi-Ornstein-Uhlenbeck model for the real spot price extracted from Nord Pool Elspot for the period 2008-10. As a particular extension to previous studies, positive and negative spikes are modeled as separate mean reverting stochastic processes. We present some numerical results and investigate option price characteristics through the option value surface.