This work is concerned with the integration and coordination of decentralized combined heat and power (CHP) systems in commercial buildings. Although extensive research has been performed on theoretically optimizing the design, sizing and operation of CHP systems, less effort has been devoted to an understanding of the practical challenges and the effects of uncertainty in implementing advanced algorithms to real-world applications. This paper provides details of an undergoing field trial involving the installation of a dynamic controller for the optimal operation of an existing CHP engine, which provides electricity and heat to a supermarket. The challenges in developing and applying an optimization framework and the software architecture required to implement it are discussed. Deterministic approaches that involve no measure of uncertainty provide limited useful insight to decision makers. For this reason, the methodology here develops a stochastic programming technique, which performs Monte Carlo simulations that can consider the uncertainty related to the exporting electricity price. The method involves the formation of a bi-objective function that represents a compromise between maximizing the expected savings and minimizing the associated risk. The results reveal a risk-return trade-off, demonstrating that conservative operation choices emerging from the stochastic approach can reduce risk by about 15% at the expense of a noticeably smaller reduction of about 10% in expected savings.