The central question to be addressed in this thesis is whether Active Network Management (ANM) implemented through a Multi-Agent System (MAS) can offer a realistic basis for managing a distribution network of the future. These networks will contain many more Distributed Energy Resources (DER) than present networks resulting in a more regular occurrence of constraints being violated because of the network’s increased usage. A review of ANM schemes is presented in order to show their impact on the global and local operation of distribution networks, the economic impacts and benefits they have, and the limitations that arise from the centralisation of operations.

This thesis shows that a distributed MAS implementation of ANM could potentially remove limitations associated with a centralised approach to ANM. However, present MAS ANM approaches raise concerns with regards to their operation on realistic networks, the communication burden they create, the safety and security of their operation and the economic considerations taken when implementing decisions.

This thesis proposes a MAS ANM scheme with an agent allocated to every node of the network, implementing a fully decentralised decision making algorithm. These agents are all peers and use message propagation so voltage approximations can be calculated and DER corrective actions shared without the need for a complete, central network model to exist.

Agents select a beneficial control action to satisfy voltage constraints in order to minimise corrective costs associated with the DER actions. This is achieved by combining the cost of various DER corrective actions and a cost allocated to voltage excursions outside the target control range. This approach removes the need for a central moderator, avoiding potential single point failures and reducing communication overheads of present MAS ANM schemes.

Furthermore to ensure this MAS ANM does not affect the present GB energy market operation, the cost associated with DER corrective actions reflects the real revenue incurred by DER owners and compensated by the MAS ANM scheme operator.

Finally to confirm that the MAS ANM meets the main aim of the thesis, it is shown to successfully operate upon many test distribution network configurations, rapidly bring voltage excursions from load changes back within limits. The solutions found are comparable to an Optimal Power Flow (OPF) supplied with a full network model and set of DER costs, confirming its ability to provide similar results to centralised ANM schemes with the benefits over the operational limitations of proposed centrally moderated MAS ANM schemes.