Perhaps the most concrete topic related to the results is the cycle efficiency, which according to simulations was determined to be 58.7% under nominal operating conditions. In particular, the question is whether the cycle efficiencies shown in literature are realistic. Hartmann et al. (2012) consider that the literature values of around 70% can only be reached through ideal, isentropic configurations. According to the authors, more realistic approximation based on polytropic efficiencies results as cycle efficiency of about 60%. As shown in Figure 57, the cycle efficiency is indeed largely dictated by turbomachinery efficiency. This confirms that the assumption of nominal polytropic efficiency of 85% is amongst the most influential design choices. Considering the future studies, one of the interesting possibilities is to evaluate the cycle efficiency under grid conditions. Although with proper dimensioning the system largely operates under nominal conditions, the start-ups in particular have a decreasing influence on the efficiency.
Figure 57. System efficiency as a function of number of compression and expansion stages. Curves indicate three different values for turbomachinery polytrophic efficiency. (Adapted from Grazzini &
Milazzo 2012, 470)
Another critical design selection was made in the heat exchanger effectiveness, which was assumed to be 75%–80% for charging and 80% for discharging
magnitude of losses due to dissipated heat. In some of the models in literature, such as those of Zhao et al. (2014) and Liu & Wang (2016), ideal heat exchange and constant pressure losses are assumed based on the shown energy balance equations.
Not only the former assumption leads to exaggerated cycle efficiency, with effectiveness nearing its ideal value the pressure losses would in reality decrease the cycle efficiency drastically, as shown in the figure below. As the performance of the model is clearly sensitive to this parameter, its influence could be studied separately.
Figure 58. Cycle efficiency as a function of heat exchanger effectiveness. Pressure losses neglected in ideal cases. (Adapted from Yang et al. 2014, 1040)
Thirdly, a critical design selection was made with the expander operation mode, for which the constant inlet pressure operation mode was selected. As explained, the induced throttling loss can be justified by the high isentropic efficiency. The simulations were conducted with a constant isentropic efficiency, as the expander component in Apros only applies the efficiency loss when operated below 50% of the nominal mass flow rate. Therefore, the logical improvement would be the implementation of efficiency curves, for which a possibility readily exists in the software.
Although the model can be considered reasonably accurate, the challenge in the development is the availability of reference data. As highlighted by Figure 59, the absolute error in the mass flow rate correlation of VGV2 is to large extent below 5%,
50%
defined here as the difference between the values read from the compressor chart and the values calculated from the correlation. Although the error itself is within acceptable range, the shape of the imported correlation fits was found to influence the system performance. As questions were even posed on the validity of the VGV component, additional compressor charts should be sought and implemented in Apros in order to confirm the validity.
Figure 59. Accuracy of the VGV2 in part-load operation, the mass flow decreases as the guide vane angle approaches 0º. The points indicate the absolute error between the values read from the chart and the values calculated from the correlation.
Furthermore, alternative ways to influence the part-load performance should be considered in future studies. For example, Grazzini & Milazzo (2008, 2005) present a control scheme, in which the turbomachinery configuration is switched between series and parallel connections according to the storage pressure through the use of valves. This scheme has been revised by De Bosio & Verda (2015, 175), who simply selected to connect an additional compressor in series once a certain storage pressure is reached. As the sequence components of Apros were successfully included to the model, framework for implementing such valve configuration readily exists.
Hybridisation in the form of electric resistance heating was not included in the model despite the initial plans due to limitations in Apros, which are however to be solved
in near future. The options for hybridisation are not only limited to the introduced concepts – several relatively unconventional systems were found to have been proposed during the recent years. This development trend should be encouraged, as no heat should be purposely wasted, nor should be the available electricity.
Furthermore, if the cycle efficiency cannot be improved by conventional means, the system design could be extended to combined generation. Realistic and easily implementable short-term option under the relatively low natural gas prices would be boosting the output of the system with additional firing, requiring only small quantities of the gas (NASDAQ 2016; Freund et al. 2012, 8). In certain situations the increment in efficiency could be reached through increasing the pressure ratio of the final expansion stage low enough, allowing the system to be extended to combined generation of electricity, heat and cold.
In the end, feasibility is the most important factor when considering the success of a technology. As the low-temperature approach fundamentally suffers from lower cycle efficiency compared to systems operating with higher TES temperatures, the design selection has to be justified somehow. The use of pressurised water or organic fluids would be more reasonable selection in an economic sense for a smaller system, as the synthetic thermal oils are amongst the most expensive storage materials (Freund et al. 2012, 9; Kearney et al. 2002, 5). Particularly the pressurised water would allow the challenges in the dynamic operation to be studied accurately due to the available two-phase six-equation flow model in Apros, but would likely require further downscaling as suggested by the reviewed concepts.