Chapter 2: Research Background
2.11 Juxtaposition with Other Relevant Methods
Before the completion of the literature review chapter, it is important to make comparisons with other relevant methods from various related sources. This is because such comparisons will be able to highlight the context of the work with other methods available. It will also show the contributions of the work to knowledge more clearly. Here, two broad comparisons are made and they are:
Comparisons made based on the solution techniques used. Comparisons of results obtained with those from previous work.
2.11.1
Comparison of Work with Other Solution Techniques
Several other optimisation techniques can be used to solve the problem stated in this work depending on the key factors being sought. In time-critical systems, optimisation time would be important but in terms of accuracy, there might be some impact which might be of some reasonable significance. In this application, accuracy is not a critical issue since user-behaviour changes will have greater impact due to the override capabilities available to the user. Speed
of convergence is also not a critical issue since all data including the day ahead price is available several hours before the time required for load scheduling. However, convergence can be an issue but this is addressed here by the GA using a large number of 1000 sample population which also reduces the chances of the search mechanism getting stuck in a local minima, but enhances global optimal attainment. Effectively, this implies that various optimisation techniques will generally lead to similar answers unless the problem itself cannot converge. In order to solve the possibility of non-convergence of the results, metaheuristic approach was chosen which also involves stochastic optimization methods whereby the solution found is dependent on some set of random variables generated. This was where the choice of GA was useful.
Other evolutionary-based optimization problems such as ant colony or particle swarm optimization (PSO) are expected to provide similar results expect for the speed of convergence which is expected to be higher for PSO. The key reason for higher convergence speed for PSO is due to fewer variables used which includes velocity and position of the variables in its algorithm. But as already stated, convergence speed is not a critical factor here unless an application that requires a real-time load scheduling is desired over very short time intervals [102]. However just as in GA and several other metaheuristics, there is no guarantee of an optimal solution although the chances of obtaining this is increased by increasing the number of the original random samples, while mutation as applied in GA enhances that the search does not get stuck in a local optimal solution. So in this regard, it can be stated that GA is more suited in this application that PSO.
2.11.2
Comparison of Results Obtained with Previous Work
Finally, Table 2.1 shows a summary of the comparison of results from proposed method with results from previous work. It is observable that the measured impact on consumers who participate in demand response is not available in other related work which is the key outcome of this research.
Table 2.1: Comparison of proposed work with related work
Ref. Feedback Considered
Response Mechanism Noted Impact (on User)
Scheduling Type Measured Impact on Users Manual Automated Single Group
2 29,31 40-47 51-52 66-71 88-91 Prop.
2.12 Chapter Summary
This chapter presented a review of various contributions from different researchers’, as well as several projects and activities engaged by various governments across the world towards improving the activities of the power grid system. The areas that are considered of greater interest include reviews which aimed at improving domestic EMS whereby householders are able to engage more effectively in DR programs with retailers. While identifying with most of the techniques proposed, there remains a number of issues that still needs to be
investigated in this area. These issues are primarily observed from the literature review which confirms the research gap and can be classified into five different categories which are:
1. There is an impact on consumers who participate in DR programs with respect to the difficulties in having to check varying RTP changes because they are unable to constantly monitor and react to the price changes given other activities that they might be engaged in.
2. There is also an impact on consumers whereby appliance scheduling might deny users some preferred appliance time-of-use which might cause them some amount of discomfort. This discomfort is based on undesired schedules which can render demand scheduling programs inconvenient.
3. These difficulties has discouraged customers from investing strategically in such tariff systems, thereby making investments in smart appliances or other smart grid- related accessories such as smart plugs, unattractive.
4. As a result of the above, the number of customers who originally signed up to real-time-price tariff in several cities in the USA, have been known to decline over time. This is because those customers who fail to modify their consumption behaviour may end up paying more than they would have paid in standard tariff system [3]. So there is a strong likelihood of increased cost rather than cost reduction which can eventually drive customers back to fixed tariff pattern.
5. Similarly, customers in the UK who participate in DR programs such as Economy-7 have the challenge about not being able to discern which pricing model that is best for them. That was why a 2011 OFGEM report suggested that the addition of price comparison guide will help such customers to compare tariffs and make better decisions [103].
6. In the event of wide acceptability of the technique, the issue of security will most likely surface. Therefore, appropriate security design is required in order to ensure secure data transfer within a particular load area in order to ensure user confidence while participating.
Finally, the literature review has highlighted different optimisation techniques that are used in demand response programmes. There is room to investigate further the best method by comparing different techniques given a particular case. However in this research, the optimisation method is not a key objective nor a contribution and so Genetic Algorithm was chosen. The reasons for using GA were based more on the interest of the author and its prior use as shown in the literature [104] [105] . Moreover, load scheduling is done well in advance so there is no requirement for high speed in solving the optimisation problem. The next chapter will therefore present the architecture for a proposed model built which aims to co-ordinate the events taking place at the domestic areas. This design responds to event variables from the utility as well as customer behaviour attributes to control different household appliances by promoting behavioural modification for optimal results.