Dr Xiaonan Wang presents the How to build resilience.io for sustainable urban energy and water systems, Energy seminar for The Energy Futures Lab at Imperial College, London on 2nd December 2016

1. How to build resilience.io for
sustainable urban energy and
water systems
Agent-Based Modelling and Resource Technology Optimization
Xiaonan Wang, Koen H. van Dam, Charalampos Triantafyllidis,
Rembrandt Koppelaar and Nilay Shah
Department of Chemical Engineering, Imperial College London, UK
resilience.io platform
Energy Futures Lab Seminar
December 2nd, 2016
1

2. 2
Motivation: Sustainable Development Goals
http://www.un.org/sustainabledevelopment/sustainable-development-goals/By 2030

3. 3
Motivation: Sustainable Development Goals
Universal and equitable access to safe and affordable drinking water and adequate sanitation and hygiene
• 1.8 billion people globally use a source of drinking water that is fecally contaminated
• 2.4 billion people lack access to basic sanitation services, such as toilets or latrines

4. 4
Motivation: Sustainable Development Goals
Universal access to modern energy services, improve efficiency and increase use of renewable sources
• One in five people still lacks access to modern electricity
• 3 billion people rely on wood, coal, charcoal or animal waste for cooking and heating

5. 5
Motivation: Sustainable Development Goals
Cities of opportunities for all, with access to basic services, energy, housing, transportation and more
• More than half of the population (~3.5 billion) live in cities today, and reaching 70 percent by 2050
• 828 million people live in slums today and the number keeps rising

6. 6
Motivation: Smart Cities
*Source: UN Habitat
(United Nations Human
Settlements Program)

7. 7
Motivation: resilience.io

8. 8
resilience.io

9. 9
Motivation: resilience.io
Open access free at source
for decision making

10. 10
Motivation: Energy-Water-Food Nexus
8%
70%30%
1%
15%

11. 11
Motivation: water and energy systems in
sustainable city development
Hydro Plant
Labour input
Hours/day
No. people
Job-types
Waste heat
Material inputs
Quantity/hour
’’ /day
’’ /week
’’ /month
’’ /year
Goods outputs
Quantity/hour
’’ /day
’’ /week
’’ /month
’’ /year
Energy input
Electricity
Heat
Fuels
Liquids Wastes
Volume/time
Solid Wastes
Mass/time
GHG Emissions
Operational cost
Currency/time for labour, energy, materials
Investment cost
Currency/facility

12. 12

13. 13
Methodology of modeling and optimization
Local data collection
(Satellite, sensor,
survey, census etc.)
Data interpretation
Model and Data
Initialization
Population and resource
parameters adjustment
Option Selection
Technology
options
Policy
options
Planning
options
Scenario construction
Benchmark scenario
Macro Socio-
Economic Scenarios
Blueprints Scenario A
Blueprints Scenario Z
Model output
Scenario trajectory
Key performance
indicators
Spatial temporal
visualization
Individual option
assessments
Validation run User defined test Scenario runs and decision making
Model operation
A data-driven open source platform

14. 14
Workflow and data structure
Socio-economic Data Demand Simulation Supply Optimization

15. 15
Workflow and data structure
Spatial-temporal socio-economic dynamics
Java implementation

16. 16
Demand-side: Agent-Based Modeling (ABM)
Example of Agent Characteristics
Agent

17. 17
Demand-side: Agent-Based Modeling (ABM)
1. Synthetic population generated from a pre-processed master table
that represents the actual population with socio-economic variants.
2. Demand for water, electricity and other resources estimated based on
agents activities.
3. Output data visualised and connected to optimisation model.
APi= {(ACTj, MDTj, SDj, PDj)}
ACTj : Activity j
MDTj : Mean departure time
SDj : Standard deviation
PDj : Probability of departure
Master Table (part)
Agent Activities Time dependent regressive
functions adopted to estimate
each agent’s water, electricity
and facilities use based on
their characteristics and
activities throughout the day
Time-variant simulation:
electricity/ water demand
every 5 minutes
Aggregated per region

18. 18
Supply-side: Resource Technology Network (RTN)
Data-driven optimization model using
mixed-integer linear programming
(MILP)
Objective function: min L (Demand, Supply,
Scenarios)
= Σα1(Capital Expenditure)
+ Σα2(Operating and Maintenance Cost)
+ Σβ (Environmental Cost)
− Σλ (Economic Benefits)
Summation over minor time periods t (e.g.,
peak, normal, off-peak hours) to guarantee
supply-demand matching over all major
time periods tm (e.g., whole year, multi-
year period)

19. 19
Supply-side: Resource Technology Network (RTN)
Data-driven optimization model using mixed-integer linear programming (MILP)
Constraints:
Technology and Investment balance/ limits, N(j, i, tm) = N(j, i, tm-1) + INV(j, i, tm).
Resource balance and capacity limits- mass and energy balance,
D (r, i, t, tm) = MU * P (j, i, t, tm) + Q (r, i’, i, t, tm) – Q (r, i, i’, t, tm) + IM (r, i, t, tm)
Production limits based on capacities, P(j, i, t, tm) ≤ N(j, i, tm) CF(j) CAP(j).
Flow limits based on pipe and grid connections and capacities- geometric distance related.
Other realistic factors, e.g.,
- pipe leakage/transmission loss;
- existing infrastructure: set as pre-allocation matrix.
j – Technologies: electricity generation plants, water treatment facilities …
i – Districts: spatial zones/ cells
r – Resource: raw water, wastewater, process chemicals, solid waste, electricity, labor…

20. 20
Supply-side: Resource Technology Network (RTN)
Data-driven optimization model using mixed-integer linear programming (MILP)
Objective function: Z = ΣWT(m, tm) VM(m, tm), where
WT(m, tm): weighting factors for metrics including CAPEX, OPEX, GHG
VM(m, tm) = ∑ VIJ (j, i, m) INV (j, i, tm) + ∑ VPJ (j, i, t, m) P(j, i, t, tm)
+ ∑ VQ (r, t, m) dist (i, i′) Q(r, i, i′, t, tm)
+ ∑ VI (r, t, m) IM(r, i, t, tm) + ∑ VY (r, m) dist (i, i′) Y(r, i, i′, tm)
Decision variables:
INVj, i, tm (investment of technology j in district i during time period tm)
Pj, i, t, tm (production rate of technology j in district i during time period t, tm)
Qr, i, i’, t, tm (flow of resource r from district i to i’ during time period t, tm)
IMr, i, t, tm (import of resource r to district i during time period t, tm)
Yr, i, i’, tm (distance based connection expansion [binary],
e.g., water pipeline, electrical grid, for resource r in district i during time period t, tm)

21. 21
Case study (WASH and power sectors in GAMA)
Example optimization model output:
§ Resource supply & demand matching
§ Investment suggestions in energy and water sectors
(i.e., infrastructure expansion, operational strategies).
Central source watertreatment plant
Borehole with pipes
Protected wells and spring
Sachet drinking water plant
Bottled water plant
Central waste water treatment plant
Waste stabilization pond
Aerated lagoon
Decentralized activated sludge system
Decentralized anaerobic bio-gas treatment plant
Decentralized aerobic treatment plant
Desalination plant
Chlor-alkali plant (example industrial plant)
Fossil fuel electricity station (e.g. coal, nuclear)
Hydro electricity station
PV solar electricity station
Bioethanol from sugar cane, cassava

22. 22
Case study (demand simulation)

23. 23
Case study (demand simulation)

24. 24
Case study (demand simulation)

25. 25
Case study (demand simulation)
Residential electricity demand profile
per district over 24 hour period

26. n Total residential water demand per district over 24 hour period in year 2030
n Projection of demands (m3) for 2010-2030 socio-demographic scenario
Sub-results for residential water use
26

27. Sub-results for sanitation
n Total toilet use profile per MMDA per MMDA over 24 hour period
27
Use Times

28. 28
Optimal clean water flows (thousands of m3)
among districts
Optimal un-treated waste water flows (m3)
among districts
Water sector (investment strategies)
GAMA WASH SDG - 100% water supply and wastewater treatment by 2030

29. 29
Additional treatment capacity
needs by 2025: 200,000 m3/day
Water sector (What technologies and capacity can meet
future needs?)

30. 30
Water sector (How will the proposed system(s) affect other
sectors?)
New desalination plant substantially
increases electricity needs:
350,000 kWh /day

31. 31
Additional treatment needs capacity
needs by 2025 : 200,000 m3/day
Water sector (What will be the cost and is it affordable?)
Population and Demands 2015 2025
Population 4.39 million 5.68 million
Faecal Sludge Generation 6,651 m3/day 8,708 m3/day
Waste-Water Treatment
Needs
243 thousand m3/day 423 thousand m3/day

32. 32
Additional treatment needs capacity
needs by 2025 : 200,000 m3/day
Water sector (What will be the cost and is it affordable?)
Population and Demands 2015 2025
Population 4.39 million 5.68 million
Faecal Sludge Generation 6,651 m3/day 8,708 m3/day
Waste-Water Treatment
Needs
243 thousand m3/day 423 thousand m3/day

33. 33
Additional treatment needs capacity
needs by 2025 : 200,000 m3/day
Water sector (What will be the total cost?)
Population and Demands 2015 2025
Population 4.39 million 5.68 million
Faecal Sludge Generation 6,651 m3/day 8,708 m3/day
Waste-Water Treatment
Needs
243 thousand m3/day 423 thousand m3/day
Public Decentralised (million USD) 2010-2015 2015-2025
Expenditure for treatment capacity 90 260
Expenditure for public toilets 42 192
Total Capital Costs 132 352

34. 34
Water sector (Will it be affordable?)
GAMA – 15 MMDA values 2015 (million USD) 2025 (million USD)
Total operational costs per year 55.6 80.5
Revenues from public toilet use 33.0 82.0
Costs per Citizen per year
(USD)
12.7 11.6
GAMA – 15 MMDA values 2015 2025
Greenhouse emissions in tonnes per
year
2011 7516
Total jobs for sewerage system 82 625

35. 35
Case study (supply technologies)
Key Metrics Units
Nameplate capacity m3 of drinking water output per
year
Capacity factor of
operation
0-100%
Resources, especially
raw source water,
electricity, and labour
hours used
m3 of raw water, kWh of
electricity, hours of labour to
produce 1 m3 potable water
Side products or wastes e.g., tonnes of CO2 emissions
Capital investment CAPEX in USD
Operating cost OPEX in USD
Technology process block example:
Water purification by sun http://desolenator.com/
Potential to be more cost-effective and
selected as a promising alternative for drinking water
supply when its capacity can be scaled up to 10 times.
Demand of drinking water ~4.7 million m3 per year
estimated by ABM for the whole region.

36. 36
Water-Energy (supply optimization)

37. 37
Energy sector (supply technologies)
Technology Sub-Type People fed Typical capacity (MJ)
Photovoltaic P-Si, ground 4334 630,720,000
CSP Trough 9865 1,576,800,000
Hydropower Small 1 63,072,000
Hydropower large 1561 37,212,480,000
Coal IGCC 101104 15,768,000,000
Gas CCGT 4 7,884,000,000
Coal IGCC + CCS 101104 15,768,000,000
Gas CCGT + CCS 4 7,884,000,000
Nuclear PWR 62 9,460,800,000
Wind Onshore 63 94,608,000
Wind Offshore 3 113,529,600
Biofuel Cassava - 109618 2,743,632,000
Biofuel Sugar Cane - 191434 2,743,632,000
31
= ( )
2
t p wP C Avλ ρ

38. 38
Power generation
technology mix in
GAMA from
model suggested
investment
strategies
(Land use is considered)
Energy sector:
baseline scenario
Energy sector (supply optimization)

39. 39
Optimal electrical power grid extension suggested for the
electricity network (CAPEX of around 73.39 million USD).
Energy sector (investment strategies)
Optimal transmission schedule among
districts (MW of power).

40. 40
Energy sector (scenario tree analysis)

41. 41
Case study (supply technologies)
Electricity generated from renewable
technology/ source
FIT (USD/
kWh)
Wind with grid stability systems 0.139
Solar PV with grid stability/storage
systems
0.161
Hydro (≤ 10 MW) 0.134
Hydro (10 - 100 MW) 0.135
Landfill Gas, Sewage Gas and Biomass 0.140
Biomass (plantation as feed stock) 0.158
Technology process block example:
Biogas process
Energy-Water-Food Nexus
Decentralized anaerobic biogas treatment plant
Combing wastewater and food scraps treatment, heat and power generation, and fertilizer by-product
Capital investment: 3,092 USD
Operational costs: 0.07 USD per m3 of sewage waste treated
Revenues from selling co-generated electricity: 480 USD per year
Total value of 0.6 million USD in 2030, with 724 job opportunities

42. 42
Roadmap (using resilience.io)
Ø Workshop on model use and project development in Ghana

43. 43
Roadmap (using resilience.io)
Ø Workshop on model use and project development in Ghana
Slum neighbourhood in
Accra, Ghana

44. 44
Next steps (developing resilience.io)
Ø Design your city in the most efficient way using games
Data driven technical innovations

45. 45
Next steps (developing resilience.io)
Ø Game theory into agent based models
Arneth, A., Brown, C. and Rounsevell, M.D.A., 2014. Global models of human decision-making for land-based mitigation and
adaptation assessment. Nature Climate Change, 4, pp.550-557.

46. 46
Acknowledgements
Wonderful and diversified team:
Dr. Koen H. van Dam
Dr. Charalampos Triantafyllidis
Dr. Rembrandt H.E.M. Koppelaar
Dr. Kamal Kuriyan
Dr. Wentao Yang
Prof. Nilay Shah
Jen Ho Ker
Niclas Bieber
Stephen Passmore
Andre Head
Dr. Miao Guo
Prof Peter Head
Great and supportive agencies:
Department of Chemical Engineering, Imperial
College London, UK
The Ecological Sequestration Trust
DFID UK
Future Cities Africa
Ghana technical group
IIER – Institute for Integrated Economic
Research
Energy Futures Lab
The analysis was carried out as part of the DFID funded Future Proofing African Cities for Sustainable Growth
project. The authors are grateful to DFID for financial support (grant number 203830).

47. Thank you! Any questions?
Xiaonan Wang xiaonan.wang@imperial.ac.uk
tel (UK): +44 (0) 7874 349693
http://www.imperial.ac.uk/people/xiaonan.wang
47