Multiple boiler system (MBS) is commonly applied in the industry to cater to the process steam demand reliably and flexibly. Optimising load allocation among the boilers is significant for the routine boiler system efficiency and carbon emission reduction. This aligns with the objective of “RENKEI control” to achieve the optimum cooperative energy efficiency between multiple elements that interact with one another. MBS can be implemented in a centralised or decentralised manner, depending on the unit allocation and process extension planning. Generally, the boiler load allocation is optimized via the complex engineering mathematic approach or modelling, which might not be user-friendly for the in-house engineer without sound mathematical knowledge and optimization background. Therefore, a user-friendly tool and method to handle problems of high combinatorial complexity with a low computational burden are needed. In this paper, the graph theoretic tool, Process Graph (P-graph), which was initially used for Process Network Synthesis (PNS), is applied to optimize the load allocation of both centralised and decentralised MBS considering the non-linear boiler part-load efficiency and existing steam piping network constraint. Besides, the fuel consumption by the existing on-site boiler operation strategy, such as parallel, tandem, and user-assigned mode, can be checked by simply modifying the P-graph structure to identify the potential improvement of MBS load optimization. The methodology proposed achieves 2.48 % and 2.37 % fuel savings in both centralised and decentralised MBS optimisation. Former case demonstrates a similar result (different by 0.02 %) compared to other authors’ work.