Smart community clustering for sharing local green energy. Yoshiki Yamagata, Hajime Seya and Sho Kuroda

Transcription

1 2014 International Conference and Utility Exhibition on "Green Energy for Sustainable Development" Smart community clustering for sharing local green energy Yoshiki Yamagata, Hajime Seya and Sho Kuroda Contact: Center for Global Environmental Research, National Institute for Environmental Studies, Japan

2 Research objective Community-based Disaster Resilient Electricity Sharing System(DRESS) In Tokyo, cars are used only for weekends, half of all cars are parking at the house during the weekdays. Vehicle to Community (V2C) technology is available now. Electricity generated from PVs can be stored into the parking cars in communities. Based on the daily geographical PV supply demand estimates, we analyzed the community electricity sharing. We have developed a district clustering algorithm to delineate optimal energy self-sufficient communities. DRESS can establish CO2 neutral and energy resilient communities against black out risks in the future. 2

3 Study area Yokohama city #of micro town level zones:1661 (cho-cho-moku. In Japanese) # of total population 3.7 million (2008) 3

4 Research questions 1. Electricity surplus from roof top PVs of all detached houses in YOKOHAMA can cover the electricity demand? PVs 0.1MJ/m August (average year) Total solar irradiance January (average year) Black solid line: max; Black dashed line: average; The others: for each day Time of day 4

5 CO2 Free Electricity Surplus We considered the scenario that PVs are installed on the roofs of all detached houses and estimated the electricity Supply. Household electricity Demand is estimated by multiplying the unit electricity demand (kwh/m 2 /h, Japan Institute of Energy, 2008) by the floor area. Calculated the Surplus as (PV supply) (household demand). Building floor map (City Planning Basic Survey, 2008, provided by Yokohama city) 5

6 PV-electricity generation E( kwh / h) H A( h) S pc Kpt K H A (h) S pc Kpt K : total (solar) irradiance (kwh/m 2 /h) : array conversion efficiency (=10%) : installation area (m 2 ) (possible area of installation on the roof (=0.3) 1/cos30 o ) : running efficiency of power conditioner (=.945) : temperature correction coefficient (=.9221~1) : other performance ratio(=.89) 6

7 Household Electricity Demand v.s. PV Supply GW Household Demand Jan Feb Mar Apr 1.2 May Jun Jul Aug Sep Oct Nov Dec PV-supply (average) PV-supply (maximum) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 7

8 Seasonal Electricity Demand v.s. PV Supply GW Demand Supply_max. Supply_ave Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total Demand is lower than the PV supply (maximum) during April September. We can achieve energy self-sufficient communities. However it is larger than the PV supply (average) in most months. 8

9 Research questions 2. All EVs trips in the city can be charged by the electricity only from the PVs? 3. How many EVs would be needed to successfully store the CO2 neutral electricity surplus?. 9

10 EV mobility simulation with MATSim We used an agent-based transport simulator called MATSim to estimate trip length and arrival time information in test site. W have prepared daily plan file of 505,335 agents in Yokohama based on The Fourth Person Trip Survey (daily movements of 722,000 people). We prepared the plan file for. 8:39 NISSAN LEAF with 24kWh battery capacity and 8km/kWh electric mileage 9:05 18:53 18:22 xml format plan file of each agent

11 Number of cars moving/parking in Yokohama # of cars (million) total # of non-used cars # of used cars Time of day The total number of registered cars in the city was 989,125, and the number of cars making trips (cars in use) was about 51%. Hence 49% of the cars was not used on the day of the survey (cars not in use).

12 1.8 GW Demand Demand for charging EVs PV supply PV supply - demand January, max Demand Demand for charging EVs PV supply PV supply - demand January, average Demand Demand for charging EVs PV supply PV supply - demand August, max Demand Demand for charging EVs PV supply PV supply - demand August, average Carbon neutral surplus electricity stored in the parking EVs is sufficient to charge the used EVs, but not enough for the whole household demand. 12

13 Cumulative distribution of electricity surplus charged to or discharged for electricity demand GW Both for EVs and usual demand (May, max) Only for EVs (May, max) Only for EVs (May, average) Both for EVs and usual demand (August, max) Only for EVs (August, max) Only for EVs (August, average)

14 How many parking EVs are required to store electricity surplus? 60% 50% 40% 30% 20% 10% 0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec MAX. AVG. We need to consider August from the black out resilience perspective. (Max. demand is 7~9 pm in August. Hot sunny day also means Max supply) 14

15 Storage capacity of the electricity surplus GW Storage capacity of the electricity surplus (12%) Storage capacity of the electricity surplus (30%) August, max Time of day 12% of parking EVs is enough to store the electricity surplus in August (max case) at the whole city aggregated level.

16 Geographical Discrepancy of Demand and Supply Electricity surplus Storage affordability = Storage capacity Electricity surplus 18 wards 1661 zones storage capacity (12%) Spatial clustering is needed! Surplus and storage capacity is in balance at the aggregate level. However, there are significant regional differences. 16

17 CO2 Neutral Energy Self-sufficient Community Restructuring by a Spatial Clustering Algorithm Algorithm 1. Define contiguity 2. Set initial status (= initial C spatial cluster) 3. Local search with Simulated annealing (SA) Obj. FCN C cluster 1 abs i cluster S D 1 circularity minimize i i cluster λ: the penalty parameter ( ) Area circularity 真円度 4 2 Perimeter Penalty for complexity of cluster s shape 0,1 17

18 Initial setup Initial status: the wards (C = 18) SA Computation 10 6 Cooling schedule T k T T 0 k 1 10 k T 5, Acceptance probability 0, Pr 1 if improvement exp else T k 18

19 Ignoring complexity of shape (λ=0) Obj.FCN objfcn 0e+00 2e+05 4e+05 6e step 19

20 Complexity of shape is moderately considered (λ=10 5 ) Obj.FCN Capacity Surplus gap Penalty objfcn, abs(), circulality objfcn: > abs(): > circ: 8.8 -> step 20

21 Complexity of shape is too much considered (λ=10 7 ) Obj.FCN Capacity Surplus gap Penalty objfcn, abs(), circulality 9.0e e e e+08 objfcn: > abs(): > circ: 8.7 -> step 21

22 Storage affordability in each time of day 7 o clock 10 o clock 13 o clock 16 o clock 22

23 JAN, MAY, AUG (Max) Clusters Some common clusters exist among the seasons Resilient hierarhical clustering exist?

24 JAN, MAY, AUG (Mean) Clusters Simularity Resilience for normal situation or extreme situation?

25 Results & Discussions In order to achieve a carbon neutral and energy resilient city with DRESS, we have conducted agent-based simulations to test the possibilities at community level. We have shown that it is possible to store enough surplus PV electricity into parking EVs and share it at community level to satisfy EVs charging demand, and even household demand in some seasons. 25

26 Results and Discussions Newly developed spatial clustering method allow us to find locally carbon neutral and energy self-sufficient community structures. Community level DRESS is superior to whole city level one, because we need less difficult and cost efficient grid network infrastructure management. This method can consider at the same time both electricity surplus-demand balances and shapes of communities to maximize the benefits. 26