This paper presents a novel generative probabilistic forecasting approach derived from the Wiener-Kallianpur innovation representation of nonparametric time series. Under the paradigm of generative artificial intelligence, the proposed forecasting architecture includes an autoencoder that transforms nonparametric multivariate random processes into canonical innovation sequences, from which future time series samples are generated according to their probability distributions conditioned on past samples. A novel deep-learning algorithm is proposed that constrains the latent process to be an independent and identically distributed sequence with matching autoencoder input-output conditional probability distributions. Asymptotic optimality and structural convergence properties of the proposed generative forecasting approach are established. Three applications involving highly dynamic and volatile time series in real-time market operations are considered: (i) locational marginal price forecasting for merchant storage participants, {(ii) interregional price spread forecasting for interchange markets,} and (iii) area control error forecasting for frequency regulations. Numerical studies based on market data from multiple independent system operators demonstrate superior performance against leading traditional and machine learning-based forecasting techniques under both probabilistic and point forecast metrics.