Presentation and evaluation of early model outputs of use cases for iterative improvement - GTG Webinar - 21 January 2016
1 like500 views
The document presents preliminary outcomes of an agent-based modeling and resource network optimization program focused on the WASH sector in Gama, Ghana. It discusses the development of a simulation model to assess water resource demands and optimize infrastructure, incorporating various socio-economic scenarios and agent characteristics. Key insights and next steps involve refining use cases, enhancing technology selection, and addressing demographic changes impacting water and sanitation needs.
Presentation and evaluation of early model outputs of use cases for iterative improvement - GTG Webinar - 21 January 2016
- 6. Preliminary outcomes of the agent-based modelling and resource network optimisation for the WASH sector in GAMA, Ghana Harry Triantafyllidis, Xiaonan Wang, Rembrandt Koppelaar and Koen H. van Dam Department of Chemical Engineering, Imperial College London, UK IIER – Institute for Integrated Economic Research 21 January 2016 Resilience.IO platform
- 7. Outline • Introduction and context (Xiaonan Wang) • WASH use case 1 • Role of modelling and simulation • Agent-based modelling (Xiaonan Wang, Koen H. van Dam) • What and why • Applications in GAMA WASH simulation • Optimisation (Harry Triantafyllidis) • What and why • Preliminary/illustrative outcomes Use Case 1 • Towards the final model for the WASH sector in GAMA (Rembrandt Koppelaar) • Next steps • Interviews and review of use cases • Visual presentation • Discussion 2
- 8. Introduction and context Speaker: Dr Xiaonan Wang 3
- 9. Scenarios Output - 5 to 20 years of population, economic status, and migration - Socio-economic scenarios - Costs/benefits from successful completion - Additional effort and possibilities to meet targets Three use cases were selected previously First focus on Use Case 1: Assess outcomes of ongoing WASH projects and gaps towards meeting macro-level targets for planning WASH Challenges – Use case 1 4 Input Model use- Capacities and time frame of all ongoing projects - Targets and goals from local, national policies and international agreements - Calculate combined effect of on-going projects when fully completed to targets - Estimate gaps remaining Projects & Targets
- 10. Idea: use a simulation model of a synthetic population to experiment with different scenarios to generate demand profiles, which can then be used to optimise the technologies and networks with key performance metrics Approach: linking ABM and RTN 5
- 11. Agent-based model of WASH sector in GAMA Speaker: Dr Xiaonan Wang, Dr Koen H. van Dam 6
- 12. environment Agent-based modelling is a computational method that enables a researcher to create, analyze, and experiment with models composed of agents that interact within an environment. (Nigel Gilbert, 2007) Definition of agent-based modelling agent state behaviour 7
- 13. Purpose of the model Stage 1 Stage 2 Stage 3 Simulate population of GAMA Estimate demand for water resources based on activities of the population Provide long-term socio- economic scenarios 8
- 14. (Activities depend on agent characteristics) Agent activities (example) APi= {(ACTj, MDTj, SDj, PDj)} ACTj : Activity j MDTj : Mean departure time SDj : Standard deviation PDj : Probability of departure 9
- 16. Synthetic population (1/2) Agents are generated in a stochastic process based on real data collected for GAMA, leading to a representative synthetic population (~0.1% of real population) Socio-economic Gender (male or female) Age (0-14 years or 15+) Work force status (Employed / Not active or unemployed) Income status (Low income / Medium income / High income) Spatial Home location (point in district/MMDA) Work location, for those economically active, based on distance from home 11
- 17. Synthetic population (2/2) Access to infrastructure Drinking water (private pipe access / public tap or stand pipe / protected decentralised source / unprotected decentralised source / tanker and vendor provided / sachet water / bottled water) simplified to private/non- private access Non-drinking water (private pipe access / public tap or stand pipe / protected decentralised source / unprotected decentralised source / tanker and vendor provided) simplified to private/non-private access Toilet (Water closet / Kumasi VIP / Pit Latrine / Public Toilet / Bucket or Pan Latrine / No facilities) Household demands are dependent on: Access to infrastructure Income level Time of day (i.e. activities) 12
- 18. First model (shown in August 2015 webinar) 13
- 19. Model development – Progress update Expansions implemented: Detailed water use and flow from human activities (modelled) Waste water flow Commuting (basic) Data collection per MMDA Work in progress: Non-residential demands Pipelines and sewage flows Toilet use Commuting (detailed) Incorporate long-term population and economic scenarios 13
- 21. Screenshot agent-based simulation time tick 60 time tick 120 time tick 240 dense agent population 16
- 22. Example outcomes Drinking water demand profile per MMDA over 24 hour period Total potable water demands (annual): 31.9 million m3 these demands feed into the RTN for optimisation 17
- 23. Example outcomes Waste water profile per MMDA over 24 hour period Total waste water generation (annual): 814.1 million m3 these demands feed into the RTN for optimisation 18
- 24. Resource-Technology Network optimisation Speaker: Dr Harry Triantafyllidis 19
- 25. Idea: use a simulation model of a synthetic population to experiment with different scenarios to generate demand profiles, which can then be used to optimise the technologies and networks with key performance metrics Approach: linking ABM and RTN 20
- 26. Preparing the data 21 Current technology in place Demands from ABM Technical specifications Flow costs, capital expenditure, operational costs, import values, emissions Pipe network (?) across districts MMDAs coordinates
- 27. How the optimization works Production rates – satisfy demands Technology balance – investments Import resources? Resource surplus – better to flow it around or build infrastructure? Minimize cost/CO2 emissions while treating the network as a whole 22
- 28. Technological infrastructure - potable water 23
- 29. Technological infrastructure - wastewater 24
- 30. Resources in network 25 1. raw source water 2. electricity 3. labour hours 4. potable water 5. sludge 6. pot distributed water 7. carbon dioxide 8. influent wastewater 9. raw waste water 10. drink water sachet 11. treated effluent 12. influent faecal sludge electricity Labour hours Raw source water Potable water
- 31. Characteristics of test model 26 15 MMDAs 2-year simulation (not fixed) with ability to split a single day for 5 periods, each reflecting different demands 17 tech types, 12 resources Pipe connections Pre-allocated tech units on base year (2010) Optimize capex/opex/CO2 Satisfy % of demands in potable water and treated waste water
- 32. Preliminary outcomes of use case 1 27
- 33. Example – USE CASE 1 - work in progress Detailed implementation: The formulated model has 5,191 constraints and 36,455 variables 20% of demands in potable water and treated waste water with existing infrastructure + investments: KPI('capex'.2010) = ~16,8 billion USD KPI('opex'.2010) = ~ 36 million USD KPI('CO2'.2010) = ~8k tons / m3 70% : KPI('capex'.2011) = 0 USD KPI('opex'.2011) = ~ 127 million USD KPI('CO2'.2011) = ~ 29k tons / m3 28
- 34. Automated Visualized output 29 MMDA’s on map • Associated optimal flows • Geo-localized • Connectivity • Bandwidth of flow visualized with the width of each edge • MMDA’s size scaling up with demands
- 35. Provide insights under various input scenarios and for specific targets Implement current status of infrastructure and experiment with costs / alternative tech types Insight where to build new infrastructure for cost reduction Let the model figure out from scratch the optimal allocation of the network and associated resource flows Employ long-term provision depending on different future policies or needs (ABM-RTN re-feed) Project technical importance of any component (predict impact of flaws / breakdowns in network) Calculate environmental impact of decisions! 30
- 36. Next steps use cases and stakeholder engagement Speaker: Rembrandt Koppelaar 31
- 37. Next Steps • Refine and improve current Use Case 1, build functionality for Use Case 2 (improved water) and 3 (toilet use) • Refine technology selection for calculation CAPEX and OPEX • Add functionality for water and sewage pipe system expansion • Simulation of toilet use during the day • Integrate financial charges for pipe connections, water use and toilet use by simulated population based on PURC tariffs, public, private fees • How much does the infrastructure use cost the population, and what revenues are generated per MMDA? 32
- 38. Next Steps • Build capability to include scenarios for population and household demographics change to 2030 • What infrastructure is needed and what is the cost to respond to demographic changes and associated water demands, toilet usage, and wastewater treatment needs? • Include rainfall patterns and agricultural water demands • How do changes in rainfall and affect water needs, and what would be the implication of increased irrigation on (waste)water treatment? 33
- 39. Interviews and Review of Use Cases • MLGRD, MWRWH, AMA, TEMA, Zoomlion, CWSA Key Points: • Affordability In neighbourhoods income of population is mixed, it is challenging to know if new infrastructure is affordable for residents • Toilet categories Include toilets at schools, marketplaces, bus stops • Floating population Many people moving in and out of districts in daytime posing large challenges on infrastructure (450,000 estimates for TEMA) 34
- 40. Decide on visual presentation • Presentation of data to make interpretation easier and faster • Examples from reports by MLGRD, MWRWH, CWSA: Bar charts with table:Pie charts:Line charts: 35
- 41. Decide on visual presentation Sankey Diagram of Flows: Example estimates for Ga South 2010 in m3/day (2012 district boundaries) 36
- 42. Decide on visual presentation Horizontal stacked bar charts for categories per district • Use of different water sources (improved / unimproved) • % discharge of liquid wastes and human excreta • % of waste-water treated per district 37
- 43. Decide on visual presentation Map with Flows between districts: Example flows of potable water from treatment based on main water trunk lines 38
- 44. Preliminary outcomes of the agent-based modelling and resource network optimisation for the WASH sector in GAMA, Ghana Harry Triantafyllidis, Xiaonan Wang, Rembrandt Koppelaar and Koen H. van Dam c.triantafyllidis@imperial.ac.uk xiaonan.wang@imperial.ac.uk rembrandt@iier.us k.van-dam@imperial.ac.uk Department of Chemical Engineering, Imperial College London, UK IIER – Institute for Integrated Economic Research 21 January 2016 Resilience.IO platform