|
Energy-Aware Opportunistic Charging and Energy
% _; h5 a) b9 z+ G: R& a+ Y3 {Distribution for Sustainable Vehicular Edge and Fog * {2 d; f, `) e J7 D
Networks ' U$ ` }- \ K, s3 n" \ a+ t
5 ]( V* ~/ N6 S/ ^- G; `+ l
" y% O7 H; _! c* N0 A, s
The fast-growing popularity of electric vehicles
- a0 D6 g( B8 O+ O# ]* L(EVs) poses complex challenges for the existing power grid + o: e/ L" W. q0 M+ P9 C
infrastructure to meet the high demands at peak charging hours. 4 c4 i: V8 l! ^9 v! x
Discovering and transferring energy amongst EVs in mobile
( }2 ?) n7 {$ t8 _/ s' `; qvehicular edges and fogs is expected to be an effective solution for
; b# Z9 ~% V; D3 \% u9 x) Ybringing energy closer to where the demand is and improving the - ]: E3 \4 H5 \' |& W$ m
scalability and flexibility compared to traditional charging # u9 J! L4 n; b2 F4 c$ f& Q2 L5 u
solutions. In this paper, we propose a fully-distributed energy1 ]$ I4 t7 ^5 G1 r. w' _! L, y
aware opportunistic charging approach which enables distributed
$ x' ?) i+ k5 C% wmulti-layer adaptive edge cloud platform for sustainable mobile
) P1 G; ]" Y$ Fautonomous vehicular edges which host dynamic on-demand , r$ u4 V! y) s
virtual edge containers of on-demand services. We introduce a 3 f6 H6 \( |/ H, T9 @ l" q
novel Reinforcement Learning (Q-learning) based SmartCharge + E, U/ d4 ?7 m9 V9 X
algorithm formulated as a finite Markov Decision Process. We . L8 l1 t. X# D4 R
define multiple edge energy states, transitions and possible actions & M3 W/ I4 J" V2 q; E' a4 w( L
of edge nodes in dynamic complex network environments which " D, R- \2 {: J8 ]' E
are adaptively resolved by multilayer real-time multidimensional 6 }3 c- Z$ |* t( i2 G; D3 e4 Z
predictive analytics. This allows SmartCharge edge nodes to more
( q' h2 x3 L; r1 Baccurately capture, predict and adapt to dynamic spatial-temporal
# B- G, s1 g% I! Denergy supply and demand as well as mobility patterns when ( z% J3 k0 u5 @+ F: ^1 c
energy peaks are expected. More specifically, SmartCharge edge 2 _. t6 J S) G3 F" f
nodes are able to autonomously and collaboratively understand
3 h, H+ R: A- _" T0 J- [4 gwhen (how soon) and where the geo-temporal peaks are expected 8 o, U. }2 @; Z) R4 o8 z3 i* _
to happen, thus enable better local prediction and more accurate
' |; P6 U" [. _global distribution of energy resources. We provide multi-criteria b- X. F1 G( t- v5 X
evaluation of SmartCharge against competitive protocols over 0 b% a+ u# h, `( P9 s# @
real-world San Francisco Cab mobility traces and in the presence . {$ b+ ?+ S: t! c$ G& t
of real-world users’ energy interest traces driven by Foursquare : Z) o5 N; E7 w6 s m
San Francisco dataset. We show that SmartCharge successfully
8 V6 ], d1 c6 _/ ?7 J, ?predicts and mitigates congestion in peak charging hours, reduces
, T% R" {/ n( u$ }9 J3 q5 Ythe waiting time between vehicles sending energy demand requests
% S4 \' i) f7 _" `and being successfully charged as well as significantly reduces the
' p+ S7 E% R! F% ~0 e9 {; S9 z4 {total number of vehicles in need of energy. 9 l4 n0 o' t# r1 _/ e9 N! |5 U
- L5 q* R, Q& i
" {" _7 i% X W/ W: ?/ g5 y. Y
6 m* Y/ T+ q. F( v/ i/ _- d5 x& [* Y/ _7 t
|