We introduce a new hybrid analysis technique for large scale PEPA models that combines stochastic simulation with ODEs, providing a finer trade-off between accuracy of the solutions and the associated computational costs. We describe a version that employs ODEs for the mean component counts and also a version using ODEs for second order moments. This enables the calculation of covariances of component counts and also improves the accuracy of the mean approximations . We look at various examples demonstrating the advantages and limitations of this approach.
Information from pubs.doc.ic.ac.uk/hybrid-anl-pepa.