Load control with communications for autonomous power systems

摘要

Investigations were carried out into the capability of a load control system with central control, communications and distributed intelligence to increase the amount of energy supplied to useful loads connected to an autonomous power system. The initial part of this work included computer modeling of communications systems and autonomous power systems. This work led to the construction of equipment for experiments earned out on a wind-diesel test rig at UMIST, and a diesel-photovoltaic test rig at the Centre for Renewable Energy Systems in Greece. Initial investigations produced good results with simple algorithms and slow networks. An algorithm based on the bubblesort technique learned the sizes of loads, and was able to provide very accurate control of loading, but only when used with a very fast communications network. Subsequent testing was carried out on an autonomous power system on the Hebridean Island of Rum. Load controllers with communications were installed in several locations around this island's autonomous power system to lest the reliability of Power Line Carrier communications. A central controller equipped with power measurement devices was connected to the power system. This central controller manipulated loads on the autonomous power system during a closed loop test. Using relaying techniques a network of load controllers could pass messages across the entire island. The connections between the load controllers were reconfigured in response to changing network conditions. This is known as an ad hoc network. A description of the simulation techniques used in the computer representation of the communications system is given. The experimental hardware used in the laboratory and site tests is described. The results of these tests are examined and further work for load control systems with communications is proposed. This work has contributed to the knowledge of load control by investigating the improvements in the load factor of autonomous power systems brought about by ad hoc communications networks, simple control algorithms, and machine learning algorithms.