KuryenteWatch: Hydropower Drought & Grid Stress Forecaster for Brownout Prevention

Problem Statement

El Niño creates a severe “energy-water nexus” crisis in the Philippines: high temperatures cause a surge in cooling and air conditioning power demand (pushing Luzon and Visayas peak loads past historical records), while severe drought simultaneously depletes the reservoirs of critical hydroelectric power plants (including Ambuklao, Binga, San Roque, Agus, and Pulangi). When hydropower capacity drops by 30% to 50% during dry spells, the National Grid Corporation of the Philippines (NGCP) is forced to declare multiple “Yellow” and “Red” Grid Alerts, triggering unscheduled rotational brownouts (manual load dropping) across businesses, cold storage facilities, hospitals, and residential communities. Electric cooperatives and distribution utilities (DUs) lack predictive visibility into local feeder overload before sub-stations trip.

Proposed Solution

KuryenteWatch is an AI-driven energy-water resilience and automated demand-response platform for Philippine grid operators, distribution utilities (Meralco, electric coops), and large commercial enterprises.

1. Hydropower & Thermal Load Prediction: Machine learning models track dam water elevation trends, water discharge limits mandated by the National Water Resources Board (NWRB), and ambient heat indices to forecast available hydropower generation versus expected air-conditioning load curves 24 to 48 hours in advance.
2. Predictive Grid Alert Engine: Forewarns distribution utilities and LGUs of high-probability Yellow/Red alert windows, pinpointing which transmission corridors and feeder lines risk thermal overloading.
3. Automated Demand-Response (ADR) Coordinator: When an alert is imminent, the AI coordinates voluntary load reduction among participating commercial establishments, malls, factories, and BPO offices—signaling them via API or SMS to switch to back-up generator sets or adjust thermostat setpoints by 1.5°C in exchange for peak electricity tariff credits (Interruptible Load Program - ILP).
4. Community Power Status Bot: Delivers transparent, advance rotational brownout schedules in Tagalog and English directly to citizens and MSMEs, replacing the sudden, unannounced blackouts that ruin appliances and interrupt economic activity.

Target Market

• Primary Beneficiaries: 20+ million electricity consumers, small businesses, cold storage chains, and BPO centers vulnerable to rotating power cuts.
• Primary Buyers (B2B / B2G): Department of Energy (DOE), Energy Regulatory Commission (ERC), National Electrification Administration (NEA), 121 Electric Cooperatives, private distribution utilities (Meralco, Visayan Electric, Davao Light), and large commercial property operators (malls, hotel chains, industrial parks).

How AI Is Used

• Time-Series Energy Demand Forecasting: Deep learning models (LSTM / Chronos / Prophet) model weather-correlated electricity consumption patterns downscaled to municipal feeder lines.
• Hydro-Meteorological Generation Modeling: Non-linear physics-based regression correlates river inflow, dam reservoir head levels, and evaporation rates to estimate available megawatt output from run-of-river and reservoir hydro plants.
• Smart Load-Shedding Optimization: Multi-agent reinforcement learning determines the most equitable and economically least-disruptive sequence of load shedding across industrial and commercial zones, shielding hospitals, water treatment plants, and dense residential sectors.
• Why it fails without AI: Grid balancing during compound climate events involves thousands of variables across generation, transmission, and weather that traditional rule-based SCADA software cannot forecast adaptively.

PLDT / Smart / ePLDT / Rokid Angle

• ePLDT VITRO Data Centers: As one of the largest energy consumers in the country, ePLDT’s hyperscale data centers can integrate KuryenteWatch to optimize their cooling loads and participate in the Interruptible Load Program (ILP), demonstrating national grid citizenship.
• PLDT Enterprise B2B Power IoT: Marketed as a mission-critical Smart Grid Energy Management service to large enterprise clients, manufacturing zones, and utility cooperatives.
• Smart Communications: Powers the real-time IoT gateway communications between distributed sub-station meters and the central analytics cloud.

Demo Scenario (3 minutes for judges)

1. The Grid Under Pressure (0:00 - 0:45): The presenter showcases a live Luzon grid simulation in mid-April during El Niño: 42°C heat index in Manila drives peak power demand to 14,500 MW, while San Roque and Angat hydro outputs drop by 350 MW due to water rationing. The simulated NGCP system is 15 minutes away from a catastrophic Red Alert blackout.
2. The AI Forecaster (0:45 - 1:30): The KuryenteWatch screen displays the warning detected 6 hours prior. The AI forecasts a 220 MW deficit between 2:00 PM and 4:00 PM and initiates an automated demand-response event.
3. Automated Demand Response (1:30 - 2:15): The system triggers an automated API dispatch to 45 enrolled commercial buildings (malls, IT parks, ePLDT facilities). The dashboard shows smart chillers automatically dialing back by 1.5°C and auxiliary diesel generators kicking in, immediately shedding 180 MW from the grid without interrupting mall operations.
4. Preventing the Brownout (2:15 - 3:00): The grid frequency stabilizes from 59.2 Hz back to 60.0 Hz. The Red Alert is avoided, saving residential neighborhoods and small businesses in 8 cities from sudden rotational blackouts.

30-Hour Build Plan

Suggested stack: Python (FastAPI, Pandas, Scikit-Learn, Statsmodels/Prophet), Next.js, Tailwind CSS, Recharts / Tremor, Semaphore SMS API, PostgreSQL.
Public data / APIs: Wholesale Electricity Spot Market (WESM) Market Data, DOE Philippine Energy Plan datasets, PAGASA Dam Water Bulletins, Open-Meteo Historical Weather API.

Risks & Privacy Notes

• Utility Data Sensitivity: Industrial power consumption profiles can be commercially sensitive. The prototype utilizes public aggregate grid data and synthetic enterprise load profiles; production implementations require role-based encryption and NDAs.
• Safety Safeguards: Demand-response shedding triggers must have automated overrides so critical life-support systems (hospitals, trauma centers) are permanently blacklisted from any automated load-drop commands.

AI Use Disclosure (draft)

KuryenteWatch utilizes ensemble time-series machine learning models to forecast electrical demand and regression algorithms to estimate hydroelectric output. Generative AI tools are utilized solely to draft standardized utility advisory alerts and operator summaries. Grid dispatch simulations are benchmarked against historical WESM operational thresholds.

Rubric Scoring