• No results found

Chapter 2: Research Background

2.7 Optimisation Techniques for Load Scheduling

2.7.1 Individually-Based Load Scheduling

An individually-based load scheduling is a system that possesses a localised decision-making algorithm that controls only the load of a specific domestic load and allocates time schedules for appliance use within the household. The schedules proposed are solely designed to accommodate the interest of household without any interaction or assessment of the impact on the comparative load profiles of the community. Such approach is best situated

where it is difficult to encourage a community-based load scheduling program due to privacy issues, diverging interests amongst various householders within the community, or when it is not cost effective to do so.

An example is based on the work of authors in [52, 53] who analysed the results of a pilot scheme that was launched by the end of 2010, to evaluate washing machine load potential for integration in smart grid. The aim of this individually- based scheduling was to encourage certain consumers to schedule their laundry only when their photovoltaic panels were supplying electricity for over a period of more than 15 weeks. The results showed a possible peak load reduction of 5% using a coincidence factor based on average weekly washing machine load profiles. Although it was not clear if the reduction was exclusively attributed to the washing machine scheduling, the results were encouraging. A consequence of the result was pivotal in the design of a new smart grid pilot known as ‘Your energy moment’, launched in December 2012 in Netherlands and it covered more than 250 households.

In other instances as presented in [41-48, 54-65] where individually-based scheduling was applied, the authors discussed the merits of identifying specific loads also known as schedulable loads and then engaging them specifically for load scheduling purposes. In [41] the authors presented an appliance commitment algorithm that schedules thermostatically-controlled loads based on price and consumption prediction while prioritising customer comfort. Only controllable thermostatically controlled appliances (C-TCAs) such as HVACs and water heaters were scheduled using consumer level DR programs. Controllable non-thermostatically controlled loads (non-TCAs) such as washing

machine and dryers are considered straightforward to schedule, unlike non- controllable appliances which the authors considered non-schedulable. Result shows the generation of day-ahead consumption schedule using forecasted day-ahead energy price forecast [41]. Although the algorithm used is fast and robust, it did not include analysis using probability of appliance use nor did it consider analysis based on historical load profile. A consequence is the non- availability of intelligent decision making capabilities which helps to reduce user involvement in the operations thereby improving user comfort.

The authors in [43] scheduled Electrical Water Heater (EWH) using a novel Traversal-and-Pruning (TP) algorithm. The problem was presented as a Mixed Integer Non-Linear Programming (MINLP) problem which can be solves in a variety of ways such as: PSO [44], Genetic Algorithm (GA) [66] and Simulated Annealing (SA) [45]. However, the authors resolved that these methods are for solving general MINLPs but on a special occasion that requires removal (pruning) of unlikely outcomes, the need for TP algorithm for solving specific appliance commitment problem becomes important. Results are a solution-tree analysis of varying temperature settings whereby branches that deviates from ambient temperature specifications were pruned and avoided in subsequent iterations. An optimal path was determined which coincides with the cheaper heating costs during the day.

The authors in [57] simulated an effective autonomous appliance scheduling for households who are both producers as well as consumers of electricity. The domestic smart scheduler, embedded in the smart meter, is based on each

device’s time of use (TOU) probabilities and RTP of energy. Results showed that the schedulable appliances requests were altered by systematically switching them on and off avoiding peak or high cost durations of electricity during the day. But the methodology did not include re-distributing the appliances at other convenient times in the day when such high energy prices are encountered. Rather is it only a single on/off mechanism whenever the price of energy exceeds a certain predefined threshold.

In [58] the authors demonstrated how coordinated scheduling of residential DER could be achieved using PSO [59-65]. The user would first assign values such as hourly consumption and discharging times with which the scheduler operates. Thereafter, the desired energy services such as: electric vehicle, space heater, water heater, pool pump and photo voltaic (PV) system; which are considered random particle trajectories, are optimised by the scheduler to achieve maximum benefits using PSO. The optimal benefit was obtained by increasing repulsion among the particles which added more randomness to the particle trajectory, in order to prevent premature convergence.

In summary, individually-based load scheduling pertains to a household’s requirements and can be useful when it is difficult to accommodate events on the larger community. Since not everyone will be happy to permit sharing of the details of their load profiles to a third party, this type of load scheduling will be ideal for such customers. But if such data is free from abuse and with increased security, group-based load scheduling will be ideal as there are more advantages when scheduling is performed based on the events on the larger

community than when it is individually-based because the load profile of the community will have a more reduced peak load when scheduling is considered on a micro-grid scale than when scheduling is considered on individual bases.