economic benefits. As an offshore power system is often remotely located, it often experiences more adverse environments and higher failures than its onshore counterpart. More powerful tools are therefore handling uncertainties occurring in onshore power-system maintenance.

An adaptive maintenance advisor is proposed here for handling uncertainties in condition-based maintenance of offshore power systems. Type-2 fuzzy logic is known for being potentially more versatile than type-1 fuzzy logic in handling uncertainties. Relative merits of these two types of fuzzy logic are investigated in this research.

The proposed type-2 fuzzy adaptive maintenance advisor is shown in Figure 1, in which a multi-phase reliability model of chosen connected equipment calculates the reliability indices at various load points supplied from a substation. Figure 1 also shows the other major block of the proposed overall architecture (Maintenance Activity Optimiser), which optimises the initial plan of maintenance of an offshore power system by considering only major system variables in the design stage.

The adaptive maintenance advisor implements and adjusts the day-to-day maintenance of chosen connected equipment by first adopting the current or initial maintenance schedule as generated by an overall optimisation for all connected equipments, and then adjusting the maintenance schedule for each chosen equipment during operation according to the shifts in control, set-point, operation, measurement and human-judgment, detected from all other connected equipments and grid. These shifts will certainly contain a lot of uncertainties.

An algorithm for carrying out type-2 fuzzy logic analysis and learning of the proposed adaptive maintenance advisor contains three parts: namely the input, type-2 fuzzy logic, and parameter re-evaluation modules. Type-2 fuzzy rules are converted from expert/ domain knowledge and mapped into linguistic values. Inputs and outputs of the fuzzy inference engine are linked by inference rules. The widely used Gaussian functions are adopted to describe the probability and possibility in a range rather than single crisp values. Mendel’s type reduced method is used here.

Although there are many factors affecting the reliability of chosen equipment, the present work as shown in Figure 2 focuses on the effects of changing weather, load factors, working environments and equipment age. Type-1 and type-2 fuzzy logic have been compared in case studies to demonstrate the design versatility, ability and efficiency of using type-2 over type-1 fuzzy logic in our present research problem. Additional operational uncertainties occurring during sensing and measurements are being incorporated by further changing the footprint of uncertainty of membership functions.