BRICS SMART CITY 2030 — Category 3: Smart-City Technologies and Services
JalDrishti 2030
An interactive companion to the research paper “JalDrishti 2030: An IoT–AI Digital Twin for Predictive Water-Stress and Intervention Planning in Bengaluru”. Ten steps, from the problem the paper opens with to a traceability table mapping every module below back to a paper section. Each step states plainly whether what it shows is real, synthetic, proposed, or validated.
Jyatin Kumar Singh · supervised by Dr. Parveen Kumar, Mittal School of Business, Lovely Professional University
The problem
RealWater stress in Bengaluru isn't one number
A citywide supply-demand balance hides which specific wards run dry. In 2024 a large share of the city's borewells failed while a groundwater assessment flagged dozens of wards and peripheral areas as critical — even as aggregate supply capacity was being increased elsewhere. Adding water to the system doesn't route it to the neighbourhood that has none.
Paper reference: §1 Introduction
Research concept
ProposedA planning layer, not a control system
JalDrishti proposes a six-stage cycle sitting above existing utility infrastructure: Sense → Integrate → Predict → Simulate → Optimise → Decide. It recommends where to intervene and why, for human review — it does not operate the network.
Paper reference: §4.1 Analytical framework
Water-stress index
RealRanking wards from their indicators
Six indicators, normalised with direction reversal so a larger number always means more stress, weighted by entropy derived from the ward set itself rather than assigned by hand. The arithmetic is real and runs live below. Most of the indicators feeding it are synthetic — see the data-availability step.
Paper reference: §4.4 Water-Stress Index
Highest-ranked wards, entropy weighting
- 1BLR-W07Critical0.79140% conf.
- 2BLR-W03Critical0.77140% conf.
- 3BLR-W14Critical0.75740% conf.
- 4BLR-W35Critical0.75640% conf.
- 5BLR-W36Critical0.71040% conf.
Prediction
RealA model has to earn its place
Demand forecasts are validated against a simple seasonal baseline on a held-out chronological window. A more complex model is only adopted if it beats that baseline — otherwise the baseline ships. This rule is implemented and enforced in code, not just described.
Paper reference: §4.5 Predictive modelling
Demand forecast · BLR-W01
Baseline MAE 5.06 · Candidate MAE 7.07
Digital twin
SyntheticA representative network, not a calibrated one
The Level 2 hydraulic layer is a stand-in distribution topology you can stress with a leak, a pump outage or a supply cut, with a minimum-pressure feasibility gate. Its pressures come from an illustrative falloff formula today. The paper's proposal is a full EPANET/WNTR solve — that integration hasn't been built yet.
Paper reference: §4.7 Level 2 — hydraulic digital twin
Leak, pump-outage and supply-reduction perturbations are available in the interactive view. Pressures there come from an illustrative distance-and-elevation formula — the proposed EPANET/WNTR solve is not yet wired in.
2030 scenarios
ProposedFour futures, not one forecast
Business as usual, supply augmentation, predictive intervention, and integrated water resilience. Each shifts demand, loss and supply multipliers through the index and the network so a plan can be tested against more than one assumption about how the next four years go.
Paper reference: §4.13 Scenarios (Table 5)
ABusiness as usual
Existing management and verified programmes continue.
demand ×1 · loss ×1 · supply ×1
BSupply augmentation
Verified supply additions occur.
demand ×1.04 · loss ×1 · supply ×0.82
CPredictive intervention
Forecasts guide intervention priorities.
demand ×0.97 · loss ×0.86 · supply ×0.95
DIntegrated water resilience
Supply, demand, loss, recharge and equity measures are combined.
demand ×0.92 · loss ×0.74 · supply ×0.86
Intervention planning
SyntheticThere is no single best plan
A prototype multi-objective portfolio search generates candidate intervention sets and keeps only those nothing else beats on every objective at once — cost, residual stress, reach, and equity. The non-dominated sort is real. The intervention effect sizes and costs it searches over are fabricated: the paper gives relative cost bands, not benefit figures. The planned NSGA-II solver and hydraulic verification of leading solutions are not yet built.
Paper reference: §4.10–4.11 Optimisation & equity
Pareto front at a 900 L budget cap
Data availability
RealThe finding underneath everything else
A Phase 2 audit graded every variable this model needs against the resolution it needs it at. Ward consumption, non-revenue water and supply reliability all came back at the bottom of the scale — not published anywhere at ward or DMA resolution. This negative result is the reason the rest of the site is built the way it is.
Paper reference: §4.3 Phase 2 data-availability audit (Table 3)
Variables graded E — not available at ward resolution
- EWater consumptionNo sufficiently usable public ward or DMA monthly or annual consumption series identified.
- ESupply reliabilityNo systematic ward or DMA record of supply hours, interruption frequency or pressure identified.
- ENon-revenue waterOnly city-level loss evidence identified; no ward or DMA measurements published.
Roadmap
ProposedThree phases from prototype to validated twin
Phase 1, this site, demonstrates the methodology end to end on synthetic and real public data. Phase 2 adds a backend, real ingestion and a persisted analytical engine. Phase 3 adds the real hydraulic solver and optimiser and produces the first results that could earn a Validated label.
Paper reference: §7 Future work
Interactive research prototype
This site. Full methodology, synthetic + available real data.
Data & analytical engine
FastAPI, PostGIS, real ingestion, persisted WSI & forecast runs.
Digital twin & optimisation validation
EPANET/WNTR, NSGA-II, scenario evaluation, research validation.
Research paper
RealEvery feature traces back to a section
None of this website's mechanics were invented for the demo. Each module below implements a specific, citable part of the paper.
Paper reference: Full paper
Feature → paper section
- Six-stage Sense→Decide framework§4.1Proposed
- Phase 2 data-availability audit§4.3, Table 3Real
- Water-Stress Index (normalisation + entropy weighting)§4.4Real
- Forecast baseline-vs-candidate validation rule§4.5Real
- Level 2 hydraulic digital twin§4.7Proposed
- Intervention library & cost bands§4.9, Table 4Real
- Multi-objective optimisation objectives§4.10Proposed
- Equity-weighted vulnerability index§4.11Synthetic
- Sensitivity / rank-stability check§4.12Real
- Four 2030 scenarios§4.13, Table 5Proposed
- Governance — decision support, not control§6.5Real
Start with the data audit — it explains every synthetic label that follows.
Open the data-availability audit