The design and the optimisation on the power and the co-generation sector are usually made using thermodynamic modelling based on the first and second principles. The tools needed for the supervision of the equipment are therefore derived from the same models as a practical and efficient way to set them up. That solution reveals itself very practical and cost effective for large, highly optimised plants. There, sufficient resources can be spent for supervision and optimisation purpose as small differences in efficiency represent huge sums of money. With the constant development of renewable energies for the power generation and in the co-generation, the need to search for new, simple, easy to use supervision solutions has become of paramount importance. Moreover, in a near future, most of those plants will have to be integrated into a de-centralised combined generation and consumption system with remote optimisation and supervision functionalities. Such a system is often called virtual power plant. The pressure to find solutions offering a high versatility in use while remaining cost efficient is rising. Neural Networks are already applied with proven great benefits in various industries. The power market's traditional conservative approach has slowed down a wider use of such a technology. The time is set to move into that direction. The present paper shall explain the whereabouts of NN and the possibility of application for versatile virtual power plants. A discussion of the advantages and the disadvantages of both solutions shall emphasise the application fields of NN. A possible approach for a practical and efficient implementation of such systems shall be developed.