**5. Conclusions**

In this paper, we present an overview of the current state of knowledge regarding parallel power flow computation using GPU technology. The utilization of GPUs in the PF computation contributes to computational performance improvement and confirms its suitability for parallel processing. In summary, the technical trends and applications of GPUs for the PF computation are reviewed as follows:


The parallel processing using GPUs is expected to be accelerated significantly further as their hardware performance improves more and more. We would be highly recommended to consider the utilization of GPUs in the near future for electrical power system analysis with a large amount of computations. In particular, it is expected that the GPU-based parallel processing will be very useful in the case of on-line operational planning that reflects the real time changes of the system, or performing simulations considering massive contingencies. However, the studies of performing PF analysis using the GPU techniques are still in a developing phase. Although significant speedup achievements were sufficiently convinced in the studies, but a standard technique has not been ye<sup>t</sup> established completely at this moment. Also, we can expect the employment of cloud computing in the power flow analysis since the power systems are scaling up incrementally as the use of renewable energy is getting increased [68]. By exploiting the coarse and fine-grained parallelism of the PF study with the hybrid of cloud computing and GPU technologies, we can resolve the extremely massive numerical computations to evaluate the future power system properly. We have a plan to study the hybrid further soon.

**Author Contributions:** All authors contributed to this work. Writing—original draft preparation, D.-H.Y.; writing—review and editing, D.-H.Y. and Y.H.; supervision, Y.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Korea Electric Power Corporation (Grant number: R17XA05-48) and the National Research Foundation of Korea (NRF) gran<sup>t</sup> funded by the Korea governmen<sup>t</sup> (MSIT) (Grant number: 2019R1C1C1008789).

**Conflicts of Interest:** The authors declare no conflict of interest.
