Customer oriented battery design

Abstract

Battery technologies are emerging as an important candidate to balance the PV generation. Due to its fast adaption in the residential sector, new algorithms to facilitate its design are needed. In order to understand and analyze its operation, the essential components of the power balance: PV generation and consumption are modeled. The PV generation profiles are obtained from Solar Monkey B.V. monitoring data while the consumption is modeled via the Load Profile Generator (LPG) software, that simulates the behaviour of people in their house. Using a simulation approach, the battery performance can be estimated in various metrics. In this thesis, 7 different metrics were considered, such as, loss of load probability, investment revenue and resilience to power outages. Due to its estimation difficulty, the lifetime metric is further investigated by analyzing three approaches from the literature. This approaches are found to have difficulties with estimating the lifetime of large batteries. In order to fix it, a correction factor that considers the calendar lifetime of the battery is given. However, since this correction is not validated and, therefore, cannot be used for commercial applications at the moment, the warranty lifetime of the battery is used. In the quest of finding by which metrics should the battery design be optimized, the market research pointed to the existence of an entire preferences pectrum. Consequently, having a pre defined algorithm that only considers the physical aspects of the system (generation and consumption) will not give an adequate design. For this reason the battery satisfaction model (BSM) is developed. This algorithm incorporates the clients preferences into the fundamental game theory concept of payoff function, to rank the batteries according to their outcome desirability. In the results chapter, the main aspects that constraint the battery profitability are analyzed via a simplified approach. Later, the BSM is examined and is found to give an appropriate design according to client’s wishes. Moreover, when tested in extreme cases, it is found to behave as desired. In order to simplify the BSM algorithm, two approaches were taken. The first approach uses mean-shift clustering algorithm to find 5 general personas to represent the common views in the market and the second investigates the correlation between the different metrics. Finally, in the last chapter, the thesis research questions are addressed with final recommendations.