AI-Driven Smart Energy Management System for Polar Research Stations

Description:	Develop an intelligent energy-management system using AI for load forecasting, renewable energy integration and fuel optimization under extreme polar conditions.
Organization:	Ministry of Earth Sciences (MoES)
Department:	National Centre for Polar andOcean Research (NCPOR)
Category:	Software
Theme:	Clean & Green Technologytd>
Youtube Link:	Not provided
Dataset Link:	Not provided
Contact info:	Not provided

1. Executive Snapshot

The AI-Driven Smart Energy Management System for Polar Research Stations (Project POLAR-EMS) is a hybrid, edge-native microgrid intelligence platform designed for the National Centre for Polar and Ocean Research (NCPOR) [PS]. Operating in hyper-extreme polar environments like Antarctica (e.g., Maitri and Bharati stations), where fossil fuel logistics cost up to $15 per liter delivered and environmental regulations under the Antarctic Treaty System mandate strict carbon minimization [RESEARCHED], this system solves the challenge of intermittent renewable integration (Solar PV and Wind) alongside diesel Combined Heat and Power (CHP) generators [KNOWN]. By combining lightweight deep learning models (Temporal Fusion Transformer / LightGBM) for multi-horizon load and renewable forecasting with Mixed-Integer Linear Programming (MILP) and Model Predictive Control (MPC) for real-time dispatch, the platform balances electrical and thermal energy demands, prevents generator thermal shock, prolongs battery storage lifespan, and slashes diesel consumption by 22–35% [INFERENCE]. Operating primarily on localized edge hardware (NVIDIA Jetson / Industrial IPC) with intermittent C-band satellite synchronization [RESEARCHED], the system ensures 100% operational autonomy even during multi-day satellite blackout periods [INFERENCE].

System Metadata

Attribute	Detail	Evidence Tag
SNO	61	[PS]
PS Number	SIH26061	[PS]
Title	AI-Driven Smart Energy Management System for Polar Research Stations	[PS]
Organization	Ministry of Earth Sciences (MoES)	[PS]
Department	National Centre for Polar and Ocean Research (NCPOR)	[PS]
Category	Software	[PS]
Theme	Clean & Green Technology	[PS]
Deadline	30 September 2026	[PS]
Ideas Counter	0/500	[PS]
Dataset Link	None provided (Public Antarctic microgrid datasets to be used)	[PS] / [RESEARCHED]
Contact	Not provided	[PS]
YouTube Link	Not provided	[PS]
Scraped At	2026-09-05	[PS]

Fundamental System Objective: To maximize renewable energy utilization and minimize diesel fuel consumption in off-grid polar research stations through edge-native AI forecasting, joint electrical-thermal load balancing, and constraint-aware microgrid dispatch without compromising station life-support systems [INFERENCE].

2. Problem Deconstruction

The explicit problem description states: "Develop an intelligent energy-management system using AI for load forecasting, renewable energy integration and fuel optimization under extreme polar conditions." [PS]

Below is the atomic breakdown of explicit and derived requirements:

Req-1: AI-Driven Load Forecasting

• Source Text: "using AI for load forecasting" [PS]
• Interpretation: Predict both electrical load (scientific equipment, lighting, IT, life support) and thermal load (space heating, snow-melting for drinking water) over short-term (15-min to 1-hour) and medium-term (24-hour to 72-hour) horizons [INFERENCE].
• Implementation Implication: Implement a multi-variate time-series forecasting model (e.g., LightGBM / Temporal Fusion Transformer) taking historical load, indoor/outdoor ambient temperature, wind chill, station occupancy schedules, and active scientific experiment runs as input features [INFERENCE].
• Confidence Tag: [PS] / [INFERENCE]

Req-2: Renewable Energy Integration

• Source Text: "renewable energy integration" [PS]
• Interpretation: Dynamically balance variable, non-dispatchable power from Antarctic Solar PV arrays (operating during Austral Summer 24-hour sunlight) and Wind Turbines (operating year-round but subject to severe katabatic wind cutoffs) with Battery Energy Storage Systems (BESS) [RESEARCHED].
• Implementation Implication: Build solar irradiance and wind power generation forecasters linked to weather radar/NOAA GFS/ECMWF data [RESEARCHED]. Implement active load-shedding and demand-response prioritization (tier 1: life support/heating, tier 2: communications/habitat, tier 3: scientific experiments, tier 4: non-critical appliances) [INFERENCE].
• Confidence Tag: [PS] / [INFERENCE]

Req-3: Fuel Optimization

• Source Text: "fuel optimization" [PS]
• Interpretation: Minimize diesel consumption across the station's Combined Heat and Power (CHP) diesel generators, accounting for generator efficiency curves (BSFC - Brake Specific Fuel Consumption), minimum run times, startup/shutdown costs, thermal energy recovery, and fuel tank degradation [RESEARCHED].
• Implementation Implication: Formulate a Mixed-Integer Linear Programming (MILP) rolling-horizon optimization engine that determines unit commitment and generation setpoints every 15 minutes to meet net load at lowest fuel mass [INFERENCE].
• Confidence Tag: [PS] / [INFERENCE]

Req-4: Operation Under Extreme Polar Conditions

• Source Text: "under extreme polar conditions" [PS]
• Interpretation: System must operate under sub-zero temperatures (down to -50°C), severe katabatic winds (>100 knots), 6-month polar nights (zero solar PV), high satellite latency/intermittent connectivity, and harsh hardware limitations [RESEARCHED].
• Implementation Implication: Edge-first architecture deployed on ruggedized industrial computers with local SQLite/Time-series database; zero cloud-dependency for real-time control; automatic offline caching and lightweight MQTT-over-Satellite sync protocol [INFERENCE].
• Confidence Tag: [PS] / [INFERENCE]

3. Stakeholders & User Needs

Stakeholder	Goal	Current Pain	System Need
Station Station Commander / Expedition Leader [KNOWN]	Ensure 100% station safety, mission continuity, and resource security during long polar winters [INFERENCE].	Constant anxiety over diesel reserves, fuel degradation, and unexpected power outages that threaten life support [RESEARCHED].	Unified executive dashboard displaying real-time fuel runway (days of fuel remaining), power reliability index, and critical alert feeds [INFERENCE].
Electrical & Mechanical Chief Engineer [KNOWN]	Keep CHP diesel gensets, wind turbines, solar panels, and battery banks operating at peak efficiency with minimal maintenance [RESEARCHED].	Manual generator synchronization, sub-optimal generator loading (wet stacking due to low loads), and sudden tripping during wind gust cutoffs [RESEARCHED].	Automated dispatch recommendations, health monitoring of gensets, thermal-electrical balance status, and maintenance alerts [INFERENCE].
Research Scientists / Glaciologists [KNOWN]	Uninterrupted power for high-precision analytical equipment, satellite earth stations, and long-term climate sensors [RESEARCHED].	Power load shedding during polar storms or low battery levels interrupting multi-day scientific runs without warning [INFERENCE].	Advance power availability notifications, scheduled experiment power booking, and non-intrusive automated load priority management [INFERENCE].
NCPOR Headquarters / MoES Logistics Directorate (Goa, India) [KNOWN]	Optimize annual Antarctic expedition fuel procurement budget, monitor station carbon footprint, and plan seasonal resupply vessels (e.g., MV Vasiliy Golovnin) [RESEARCHED].	Lack of granular, real-time data on daily fuel burn rates and renewable penetration percentage; dependency on manual monthly report emails [RESEARCHED].	Low-bandwidth telemetry synchronization giving HQ visibility into historical energy trends, fuel savings, and carbon reduction metrics [INFERENCE].

4. Requirements Traceability

Feature / Subsystem	Direct / Derived Requirement	Source Justification
Multi-Horizon AI Load & Renewable Forecaster	Req-1 (Load Forecasting) & Req-2 (Renewable Integration)	[PS] Explicitly requires AI load forecasting and renewable energy integration.
Thermal-Electrical Joint Unit Commitment (MILP Engine)	Req-3 (Fuel Optimization)	[PS] Optimizing fuel requires accounting for CHP thermal output, preventing auxiliary electric heater usage when waste heat can be captured [INFERENCE].
Dynamic Priority Load-Shedding & Demand Response	Req-2 (Renewable Integration) & Req-4 (Polar Conditions)	[INFERENCE] When wind drops suddenly during polar night, automated load control prevents total microgrid blackouts.
Edge-First Offline Architecture with Local TSDB	Req-4 (Extreme Polar Conditions)	[RESEARCHED] Polar stations rely on satellite links (C-band/Iridium) with latency >600ms and frequent outage periods. Control loops must run locally.
Battery Health & Cold-Weather Thermal Management Engine	Req-2 (Renewable Integration) & Req-4 (Polar Conditions)	[RESEARCHED] Lithium / Lead-acid battery charge acceptance drops drastically at temperatures below 0°C; BESS must be thermally managed using CHP waste heat [RESEARCHED].
Fuel Tank Inventory & Degradation Tracker	Req-3 (Fuel Optimization)	[RESEARCHED] Fuel (Jet A-1 / Polar Diesel) stored in 300,000L tanks at Bharati/Maitri undergoes temperature fluctuations and requires precise inventory tracking [RESEARCHED].
Satellite-Optimized Telemetry Sync Engine	Req-4 (Extreme Polar Conditions)	[RESEARCHED] Bandwidth between Antarctica and mainland India (NCPOR Goa) is constrained (4–10 Mbps shared across voice/data) [RESEARCHED].

5. Constraints & Environment Realities

1. Geographical Isolation & Logistics: Stations like Bharati (69°24'S, 76°11'E) and Maitri (70°45'S, 11°44'E) are isolated for 8–9 months during Antarctic winter [RESEARCHED]. All fuel is shipped once per year via specialized icebreakers [KNOWN]. Fuel cost includes freight and transport risk, exceeding $10–$15 per liter [RESEARCHED].
2. Extreme Meteorological Conditions: Temperatures drop to -50°C in winter [KNOWN]. Katabatic winds regularly exceed 100 km/h (reaching 200 km/h during blizzards) [RESEARCHED]. Wind turbines automatically brake at high cutoff speeds (~25 m/s) to prevent mechanical destruction, creating instantaneous power loss spikes [INFERENCE].
3. Polar Sunlight Cycle: 24 hours of solar radiation during Austral Summer (November–February), dropping to 0 hours during Austral Winter (May–August) [KNOWN]. Solar PV is completely unusable for ~4 months per year [INFERENCE].
4. Thermal-Electrical Coupling: Heating in polar stations is life-critical [KNOWN]. Waste heat from diesel generator cooling systems and exhaust (CHP) is piped to heat living modules and melt snow for drinking water [RESEARCHED]. If generators are turned off entirely due to high wind/battery power, auxiliary diesel boilers or electric heaters must turn on, altering the net fuel calculation [INFERENCE].
5. Connectivity Limits: Dedicated C-band satellite link provides 4–10 Mbps shared bandwidth [RESEARCHED]. Solar flares, geomagnetic storms, and heavy snow build-up on radomes cause multi-hour to multi-day blackout periods [INFERENCE]. Cloud-only systems are completely infeasible [INFERENCE].
6. Strict Environmental Regulations: The Madrid Protocol (Protocol on Environmental Protection to the Antarctic Treaty) strictly regulates carbon emissions, waste disposal, and fuel spill prevention [RESEARCHED].
7. Hardware & Power Infrastructure: Existing systems utilize Caterpillar / Cummins diesel generators (e.g., 3 x 100 kVA CHP units at Bharati) controlled via MODBUS RTU/TCP or CANbus protocols [RESEARCHED]. Compute hardware on site consists of ruggedized server racks with limited GPU acceleration capacity [INFERENCE].

6. Research Questions

1. RQ-1 (Thermal Coupling): What is the exact thermal-to-electrical energy ratio required by Bharati/Maitri stations across summer vs. winter months, and how does stopping a diesel generator impact heating fuel consumption? [NEEDS VALIDATION]
2. RQ-2 (Genset Efficiency): What are the Brake Specific Fuel Consumption (BSFC) curves and minimum loading limits (typically 30–40% to prevent wet stacking/carbon build-up) for NCPOR's specific 100 kVA CHP units? [RESEARCHED]
3. RQ-3 (BESS Behavior at Low Temps): What battery chemistry (LiFePO4 with internal heaters vs. Lead-Acid/Gel) is deployed at Indian stations, and what are their charge/discharge efficiency profiles under ambient winter temperatures? [NEEDS VALIDATION]
4. RQ-4 (Telemetry Protocols): Which exact industrial communication protocols (Modbus TCP, OPC UA, BacNet, CANbus) are exposed by NCPOR's power distribution control panels? [NEEDS VALIDATION]
5. RQ-5 (Weather Data Granularity): Are real-time weather observations and short-term numerical weather prediction (NWP) model outputs available on-station locally from IMD (India Meteorological Department) Antarctic weather sensors? [RESEARCHED]

7. Data Requirements & Sources

Data Type	Purpose	Ideal Real-World Source	Hackathon Substitute	Access Method	Availability	Privacy Sensitivity	Confidence Tag
Station Electrical Load	Train load forecasting model & test dynamic dispatch	NCPOR SCADA / Power House Historian Logs (Bharati/Maitri)	NREL Microgrid Load Data / Public Antarctic Base Load Profiles	CSV / Parquet File Upload	Limited Publicly [NEEDS VALIDATION]	Internal Operations (Low Sensitivity)	[RESEARCHED]
Station Thermal Load	Model space heating & snow-melting demand	NCPOR Heating Circuit Sensors	Synthetic thermal models based on outdoor ambient temperature delta	Math Script Simulation	Derived [INFERENCE]	Non-sensitive	[INFERENCE]
Solar PV Generation Data	Train solar irradiance & power output model	On-site Pyranometer & Inverter Telemetry	NREL SAM (System Advisor Model) Antarctic Solar Synthetic Data	API / CSV Export	Publicly Available	Non-sensitive	[KNOWN]
Wind Turbine Power Data	Model wind power output and high-speed cutoff events	Nacelle Anemometer & Turbine Controller	OpenEI / NREL Wind Prospector polar wind datasets	CSV Download	Publicly Available	Non-sensitive	[KNOWN]
Meteorological Observations	Predict renewable generation & heating demand	IMD Antarctic Weather Station Feed / NOAA GFS / ECMWF	Open-Meteo Historical Weather API (Svalbard/McMurdo/Larsemann Hills coordinates)	REST API / JSON	Publicly Available	Public Data	[KNOWN]
Genset Operational Data	Calculate fuel burn, thermal recovery & efficiency	SCADA PLC Modbus Registers (Voltage, Current, kW, Fuel Rate)	Generator BSFC Curve Mathematical Simulator	Python Modbus Simulator Script	Generated	Non-sensitive	[ASSUMPTION]
Satellite Bandwidth Constraints	Test low-network telemetry sync	NCPOR SATCOM Router SNMP Metrics	NetEm / Traffic Control (tc) Bandwidth & Delay Emulator Script	Network Emulation	Simulated	Internal	[ASSUMPTION]

8. Existing Solutions & Gap Analysis

Existing System	Description	Strengths	Weaknesses	Relevance	Remaining Gap	Source / Confidence
Princess Elisabeth Antarctica EMS (Belgian Station)	World's first zero-emission polar station; uses Schneider Electric PLC microgrid with 9 wind turbines, solar PV, and lead-acid BESS [RESEARCHED].	Highly reliable, custom PLC-based priority load control with manual socket switches [RESEARCHED].	Rigid rule-based priorities; lacks AI-driven predictive load/renewable forecasting; does not optimize thermal-electrical CHP trade-offs dynamically [INFERENCE].	High (benchmark Antarctic zero-emission base)	Lacks predictive AI optimization and automated multi-period MILP dispatch [INFERENCE].	Princess Elisabeth Microgrid [RESEARCHED]
McMurdo / Ross Island Energy Grid (US/NZ Partnership)	Joint microgrid connecting McMurdo Station and Scott Base using Crater Hill wind turbines and Caterpillar diesel generators [RESEARCHED].	Proven high-wind penetration, automated diesel generator synchronization across two stations [RESEARCHED].	High reliance on continuous diesel backup; complex multi-station control requiring manual operator intervention during extreme storms [INFERENCE].	Moderate (large-scale Antarctic microgrid)	High operational overhead; lacks lightweight edge-AI optimization suited for smaller 30–50 person bases like Bharati [INFERENCE].	USAP / Antarctica NZ [RESEARCHED]
Qinling Station Hybrid Power System (China, 2025)	Newly commissioned hybrid system incorporating wind, solar, hydrogen, and diesel energy targeting 60% green power [RESEARCHED].	Integrates hydrogen storage for long-duration polar night backup power [RESEARCHED].	High capital expenditure; complex hydrogen handling at sub-zero temperatures; control algorithms remain proprietary [INFERENCE].	High (state-of-the-art Antarctic green tech)	Not open-source; heavily hardware-centric without a modular software EMS for existing bases [INFERENCE].	SCIO China News 2025 [RESEARCHED]
Commercial Microgrid EMS (e.g., ETAP, HOMER Pro)	Industry-standard microgrid design and control software suites.	Comprehensive power flow analysis and static optimization capabilities.	Expensive commercial licensing; cloud-heavy architectures; lacks polar-specific thermal recovery and edge AI forecasting models [INFERENCE].	Moderate (standard tooling)	Not tailored for polar satellite blackouts, extreme thermal coupling, or SIH budget constraints [INFERENCE].	Standard Industry Knowledge [KNOWN]

9. Candidate Solution Approaches

Comparison Table of Solution Approaches

Parameter	Approach 1: Rule-Based & Threshold Controller	Approach 2: Pure Reinforcement Learning (RL) Autonomous Agent	Approach 3: Hybrid AI-Predictive MPC + MILP Edge Optimization (Proposed)
Mechanism	Static IF-THEN rules based on battery State-of-Charge (SoC) and net load thresholds.	Deep Q-Network (DQN) / PPO RL agent controlling gensets and battery in real time.	AI models forecast load & renewables; MILP solves optimal unit commitment; MPC executes real-time smoothing.
Problem Fit	Basic, handles safe operation.	Poor fit for mission-critical life support due to unpredictable edge cases.	Optimal, balances predictive intelligence with deterministic mathematical constraints.
Accuracy Potential	Low (cannot adapt to changing weather fronts or experiment schedules).	High in simulated environments; unpredictable in real-world anomalies.	High (forecasting MAPE < 5-8%, MILP yields mathematically optimal cost).
Data Requirements	Very Low (only current sensor values).	Extremely High (requires millions of simulated polar environment iterations).	Moderate (6–12 months of historical weather and load logs for forecaster training).
Complexity	Very Low.	High (difficult to train and debug).	Moderate-High (well-structured separation of concerns).
Explainability	100% Deterministic.	Black-box neural network (unacceptable for station engineers).	High (MILP output provides transparent mathematical reason for every dispatch decision).
Cost	Negligible.	High training compute cost.	Low (runs on standard edge industrial PC).
Scalability	Poor (requires manual re-scripting for new station equipment).	Poor (requires retrain for physical system changes).	High (configurable parameters for any microgrid configuration).
Hackathon Feasibility	High, but won't win an AI challenge.	Medium, high risk of unstable demo during live judging.	High (clear scope: pre-trained forecaster + PuLP/SciPy MILP optimizer + Streamlit UI).
Production Feasibility	High safety, poor fuel savings.	Unsafe for Antarctic life-support deployment without heavy guardrails.	Very High (industry standard approach for mission-critical microgrids).

10. Recommended Approach & Why

The recommended approach is Approach 3: Hybrid AI-Predictive MPC + MILP Edge Optimization Framework.

Why Alternatives Were Rejected

• Approach 1 (Static Rule-Based Controller) was rejected because static threshold rules cannot anticipate upcoming weather transitions (e.g., a major katabatic storm approaching in 4 hours). A rule-based controller keeps diesel generators running inefficiently or drains batteries prematurely right before a solar/wind drought, resulting in excessive fuel burn [INFERENCE].
• Approach 2 (Pure Reinforcement Learning Agent) was rejected due to the "black-box" problem and safety risks in extreme polar environments [INFERENCE]. An RL agent attempting to learn optimal dispatch through trial-and-error can make uninterpretable decisions, such as shedding habitat space-heating during a blizzard to maximize battery life, which poses a severe risk to station personnel [INFERENCE].

Why Approach 3 Wins

1. Safety & Explainability: The optimization layer uses Mixed-Integer Linear Programming (MILP), which operates under strict physical hard constraints (e.g., minimum generator loading, mandatory life-support heating power, battery SoC bounds) [INFERENCE]. The system can never output an unsafe dispatch state [INFERENCE].
2. Predictive Efficiency: By leveraging lightweight AI forecasting models (LightGBM / Temporal Fusion Transformer), the optimizer looks ahead 24 to 72 hours [INFERENCE]. If the AI predicts 100% wind power drop in 3 hours due to severe storm cutoff, the system pre-warms the thermal storage circuit and brings a diesel generator to optimal efficiency smoothly, avoiding emergency thermal shock or blackouts [INFERENCE].
3. Edge Native & Satellite Resilient: All forecasters and solvers run locally on an edge device (e.g., NVIDIA Jetson / Rugged Industrial PC) in Python/C++ [INFERENCE]. Satellite communication is used purely for telemetry reporting and remote model re-training, ensuring 100% continuous station autonomy [INFERENCE].

11. Technical Architecture

Layered Architecture Diagram

+-----------------------------------------------------------------------------------+
|                            SIH DEMO & HQ DASHBOARD                                |
|  React.js / Streamlit UI: Real-time Microgrid Flow, Fuel Runway, Forecasts & Maps |
+-----------------------------------------------------------------------------------+
                                         ^
                                         | REST / WebSockets (Gzip/Protobuf)
                                         v
+-----------------------------------------------------------------------------------+
|                        EDGE EMS CORE ENGINE (On-Station IPC)                      |
|                                                                                   |
|  +-----------------------------------+     +-----------------------------------+  |
|  |       AI FORECASTING MODULE       |     |     OPTIMIZATION & DISPATCH     |  |
|  | - LightGBM / TFT Load Predictor   | --> | - MILP Unit Commitment (PuLP)     |  |
|  | - Solar PV & Wind Power Forecaster|     | - Model Predictive Control (MPC)  |  |
|  +-----------------------------------+     +-----------------------------------+  |
|                                                              ^                    |
|                                                              | Hard Constraints   |
|                                                              v                    |
|  +-----------------------------------------------------------------------------+  |
|  |                        SAFETY & INTERLOCK LAYER                             |  |
|  | - Minimum Generator Load (30%)  - Mandatory Thermal Heat Balance             |  |
|  | - Life Support Circuit Lock     - Battery SoC Upper/Lower Guards            |  |
|  +-----------------------------------------------------------------------------+  |
+-----------------------------------------------------------------------------------+
                                         ^
                                         | Industrial Protocols (Modbus / MQTT)
                                         v
+-----------------------------------------------------------------------------------+
|                       DATA AGGREGATION & STORAGE LAYER                            |
|  - SQLite / TimescaleDB (Local Time-Series Log)                                   |
|  - Edge Data Cache & Sync Queue (Low-Bandwidth Satellite Sync to NCPOR HQ)        |
+-----------------------------------------------------------------------------------+
                                         ^
                                         | Modbus TCP / RTU / CANbus
                                         v
+-----------------------------------------------------------------------------------+
|                       POLAR STATION HARDWARE SIMULATION / SCADA                   |
|  - Diesel CHP Gensets (3x100kVA)  - Solar PV Array (60kW)  - Wind Turbines (50kW) |
|  - Battery Storage (BESS 200kWh)  - Thermal Heating Loop   - Scientific Load      |
+-----------------------------------------------------------------------------------+

Detailed Data-Flow Narrative

1. Ingestion: Telemetry from station power meters, inverter controllers, diesel engine ECUs, and local IMD weather sensors is polled every 1–5 seconds via Modbus TCP / MQTT protocols into the local Edge Data Aggregator [INFERENCE].
2. Local Logging: Raw high-frequency telemetry is ingested into a local TimescaleDB/SQLite time-series database running on the station's ruggedized edge computer [INFERENCE].
3. AI Inference (Every 15 Minutes): The AI Forecasting Module pulls the last 48 hours of load and weather data alongside local weather model predictions (NOAA GFS / IMD). It executes lightweight model inference to produce 15-minute resolution forecasts for:
   • Station Electrical Load (kW)
   • Station Thermal Heating Demand (kW thermal)
   • Solar PV Generation (kW)
   • Wind Power Generation (kW, including storm cutoff probability) [INFERENCE].
4. Optimization Solving (Every 15 Minutes): The forecast profiles are fed into the MILP Optimization Engine. The engine solves a 24-hour rolling horizon problem, optimizing the binary state (ON/OFF) and dispatch setpoints (kW) for each diesel generator, BESS charge/discharge rates, and auxiliary thermal heater states to minimize total diesel fuel mass [INFERENCE].
5. Safety Interlock Verification: Before any control command is issued to the SCADA system, the output passes through a hardware-coded Safety Interlock Layer. This layer enforces non-negotiable physical safety rules (e.g., ensuring active space heating power exceeds ambient loss, preventing generator short-cycling, and verifying battery temperature limits) [INFERENCE].
6. Execution & Closed-Loop MPC: The verified control setpoints are sent to generator governors, solar/wind inverters, and BESS controllers over Modbus TCP. A fast Model Predictive Control (MPC) loop executes every 10 seconds to smooth out high-frequency renewable fluctuations using battery storage [INFERENCE].
7. Low-Bandwidth Satellite Sync: Once per hour (or when satellite link is active), the edge sync manager compresses recent operational metrics, forecast accuracy, fuel usage, and alert logs using Protocol Buffers and sends them over MQTT-over-TLS to NCPOR Headquarters in Goa [INFERENCE].

12. Technology Stack

Layer	Selected Technology	Why	Alternative	Why Rejected
Edge Compute Hardware	Ruggedized Industrial PC (e.g., Advantech / Siemens Microbox) + NVIDIA Jetson Nano/Orin	Extended operating temperature (-40°C to 60°C), fanless, low power consumption, hardware AI acceleration [INFERENCE].	Standard Cloud Server	Requires internet connectivity; fragile in polar station environments [INFERENCE].
Edge OS & Containerization	Ubuntu Linux 22.04 LTS + Docker & Docker Compose	Lightweight, reproducible deployment, isolatable microservices, robust driver support for industrial serial/Ethernet [INFERENCE].	Windows Server	High resource overhead, updates require frequent reboots, less flexible for edge container orchestration [INFERENCE].
Local Database	TimescaleDB (PostgreSQL extension) + SQLite fallback	High-performance time-series data indexing, native SQL query support, minimal RAM footprint [INFERENCE].	MongoDB / InfluxDB	InfluxDB v3 changes licensing/architecture; MongoDB lacks optimized time-series compression for constrained disks [INFERENCE].
Backend Runtime	Python 3.11 / FastAPI	Fast development, native support for ML libraries (PyTorch/LightGBM) and mathematical solvers (PuLP/SciPy) [INFERENCE].	Node.js / Express	Lacks native scientific computing, mathematical optimization, and deep learning ecosystem [INFERENCE].
AI / ML Framework	LightGBM + PyTorch (Temporal Fusion Transformer)	Extremely fast CPU/GPU inference, low memory usage, state-of-the-art accuracy on multi-variate time-series forecasting [INFERENCE].	TensorFlow / Keras	Heavier footprint, slower inference setup on constrained edge hardware [INFERENCE].
Optimization Solver	COIN-OR CBC / HiGHS (via Python PuLP / Pyomo)	High-performance open-source MILP solvers, zero licensing cost, deterministic results [INFERENCE].	Gurobi / CPLEX	Commercial license required, expensive, restrictive for open hackathon/government deployment [INFERENCE].
Industrial Comm / Protocols	PyModbus + Eclipse Mosquitto (MQTT)	Universal standard for power equipment (inverters, gensets, meters); lightweight pub/sub for edge services [INFERENCE].	OPC UA	More complex setup; Modbus/MQTT is simpler and universally supported by low-level hardware simulators [INFERENCE].
Frontend UI Dashboard	React.js + Tailwind CSS + Recharts + Leaflet.js	High responsiveness, interactive real-time microgrid power-flow animation, offline caching via Progressive Web App (PWA) [INFERENCE].	Streamlit	Streamlit is great for rapid prototyping but lacks custom UI controls for interactive microgrid power flows required in final hackathon presentations [INFERENCE].
Satellite Sync Protocol	MQTT over TLS with Gzip / Protocol Buffers	Ultra-compact binary serialization, minimizing satellite data usage to < 5 MB/day [INFERENCE].	REST JSON APIs	JSON payload overhead is 5–10x larger; wasteful over metered Antarctic satellite links [INFERENCE].

13. AI/ML Justification

1. Solar PV & Wind Power Generation Forecaster

• Purpose: Predict solar PV and wind power output over a 15-minute to 24-hour horizon to enable proactive generator scheduling [INFERENCE].
• Model Type: LightGBM Regressor + Physics-Informed Wind Turbine Power Curve Filter [INFERENCE].
• Input: Historical solar/wind generation, ambient air temperature, wind speed, wind direction, solar zenith angle, surface albedo (snow cover reflection factor), numerical weather prediction (NWP) forecasts [INFERENCE].
• Output: Expected power output profile (kW) with 10% and 90% confidence intervals [INFERENCE].
• Data Required: 6–12 months of 15-minute weather and renewable generation logs [INFERENCE].
• Why AI is Preferable: Wind patterns in polar coastal regions (Larsemann Hills) exhibit highly non-linear katabatic gusts and sudden turbulence drop-offs that simple linear or persistence models fail to capture [RESEARCHED].
• Evaluation Metric: Mean Absolute Percentage Error (MAPE < 8%) and Continuous Ranked Probability Score (CRPS) [INFERENCE].
• Failure Mode: Unexpected sensor icing leads to false zero-power predictions.
• Human Oversight: Override switch allows station engineer to lock forecast to zero during active blizzards or physical turbine maintenance [INFERENCE].

2. Station Electric & Thermal Load Forecaster

• Purpose: Predict combined electrical and space-heating energy demand across station habitats and science labs [INFERENCE].
• Model Type: Temporal Fusion Transformer (TFT) or Multi-Layer Perceptron (MLP) with temporal embeddings [INFERENCE].
• Input: Hour of day, day of season, indoor/outdoor temperature delta, wind chill index, station headcount/occupancy, scheduled experiment runs [INFERENCE].
• Output: Predicted electrical load curve (kW) and thermal heat load curve (kW-thermal) [INFERENCE].
• Data Required: Historical station power meter logs and thermal circuit sensor logs [INFERENCE].
• Why AI is Preferable: Human behavior, scheduled scientific experiments, and sudden wind-chill temperature drops create complex multi-variable interactions that rule-based systems cannot model accurately [INFERENCE].
• Evaluation Metric: Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE < 3 kW) [INFERENCE].
• Failure Mode: Unscheduled power-intensive experiment causes unexpected load spike.
• Human Oversight: Station scientists log major power-consuming experiments 24 hours in advance via the EMS UI calendar [INFERENCE].

3. Multi-Objective Fuel & Genset Dispatch Optimizer

• Purpose: Determine optimal ON/OFF commitment and kW setpoints for diesel generators, BESS, and thermal heaters [INFERENCE].
• Model Type: Mixed-Integer Linear Programming (MILP) rolling horizon solver (Mathematical Optimization, non-AI neural network) [INFERENCE].
• Input: AI load & renewable forecasts, battery State-of-Charge (SoC), generator fuel consumption curves (BSFC), thermal balance requirements, minimum up/down time limits [INFERENCE].
• Output: 24-hour dispatch matrix (Genset 1/2/3 kW setpoints, BESS Charge/Discharge kW, Auxiliary Boiler kW) updated every 15 minutes [INFERENCE].
• Data Required: Physical hardware spec sheets and generator fuel curves [INFERENCE].
• Why AI/Neural Nets Should NOT Be Used Here: CRITICAL DESIGN DECISION: Real-time dispatch control in life-critical polar microgrids must be 100% deterministic, verifiable, and constrained [INFERENCE]. Neural networks (RL/DL) can produce uninterpretable actions or violate physical bounds [INFERENCE]. MILP guarantees a mathematically optimal solution while strictly enforcing hard safety constraints [INFERENCE].
• Evaluation Metric: Percentage reduction in daily fuel mass (liters) compared to baseline rule-based operation [INFERENCE].
• Failure Mode: Optimization solver fails to converge within time limit (timeout > 30s).
• Human Oversight: Automatic fallback to rule-based threshold controller if solver fails [INFERENCE].

14. Data Architecture & Offline/Low-Network Design

Database Schema (Local SQLite / TimescaleDB)

-- High-frequency sensor telemetry (1-5 sec)
CREATE TABLE sensor_telemetry (
    timestamp TIMESTAMPTZ NOT NULL PRIMARY KEY,
    genset_1_kw REAL,
    genset_2_kw REAL,
    genset_3_kw REAL,
    solar_pv_kw REAL,
    wind_turbine_kw REAL,
    bess_soc_pct REAL,
    bess_power_kw REAL,
    station_elec_load_kw REAL,
    station_thermal_load_kw REAL,
    ambient_temp_c REAL,
    wind_speed_mps REAL
);

-- AI Forecast Storage (Updated every 15 mins)
CREATE TABLE ai_forecasts (
    forecast_id UUID PRIMARY KEY,
    generated_at TIMESTAMPTZ NOT NULL,
    target_timestamp TIMESTAMPTZ NOT NULL,
    pred_solar_kw REAL,
    pred_wind_kw REAL,
    pred_elec_load_kw REAL,
    pred_thermal_load_kw REAL,
    confidence_lower REAL,
    confidence_upper REAL
);

-- Optimization Control Schedule (15-min setpoints)
CREATE TABLE dispatch_schedule (
    schedule_id UUID PRIMARY KEY,
    created_at TIMESTAMPTZ NOT NULL,
    execution_time TIMESTAMPTZ NOT NULL,
    genset_1_status INTEGER, -- 0=OFF, 1=ON
    genset_1_setpoint_kw REAL,
    genset_2_status INTEGER,
    genset_2_setpoint_kw REAL,
    bess_target_kw REAL, -- Positive=Discharge, Negative=Charge
    aux_boiler_kw REAL,
    fuel_burn_est_liters REAL
);

-- Satellite Sync Status Queue
CREATE TABLE sync_queue (
    record_id BIGSERIAL PRIMARY KEY,
    payload_type VARCHAR(50),
    payload_data BLOB, -- Compressed Protobuf
    created_at TIMESTAMPTZ NOT NULL,
    synced_at TIMESTAMPTZ DEFAULT NULL
);

Low-Network & Offline Synchronization Architecture

1. Autonomous Local Operation: The Edge EMS core engine operates completely disconnected from the WAN/internet [INFERENCE]. It reads local sensor data, stores it in SQLite/TimescaleDB, computes AI forecasts, and executes MILP optimization locally [INFERENCE].
2. Data Compression & Serialization: Operational summary logs (5-minute averaged metrics) are serialized using Google Protocol Buffers (Protobuf) and compressed via Gzip, reducing telemetry data size by up to 92% compared to standard JSON [INFERENCE].
3. Store-and-Forward Queue: When satellite connection is degraded or unavailable, compressed data packets accumulate in the local sync_queue table [INFERENCE].
4. Adaptive Bandwidth Sync: The Sync Manager periodically pings the NCPOR HQ broker. When an active link is detected, it transmits pending sync queue packets using low-overhead MQTT QoS-1 with exponential backoff and automatic resume capability [INFERENCE].
5. Conflict Handling: The Edge Station is the single source of truth for operational state. NCPOR HQ only sends high-level administrative configurations (e.g., updated model weights or fuel price parameters) [INFERENCE]. In case of sync conflicts, the station's edge database state strictly overrides HQ [INFERENCE].

15. Security, Privacy & Governance

1. Sensitive Operational Information: Microgrid control systems for national research facilities (NCPOR / MoES) are classified as Critical Energy Infrastructure [INFERENCE]. Real-time fuel reserves, generator statuses, and satellite communication logs must be protected against unauthorized access and cyber tampering [INFERENCE].
2. Access Control & Role-Based Security (RBAC):
   • Viewer Mode: Research scientists can view current power status and schedule upcoming scientific experiment power demands.
   • Operator Mode: Station Engineers can modify dispatch parameters, set manual overrides, and initiate generator testing.
   • Administrator Mode: System Administrators & NCPOR HQ staff can update system software, AI model weights, and hardware configurations [INFERENCE].
3. Authentication & Authorization: Hardened OAuth2 with JWT tokens for local Web UI access; mutual TLS (mTLS) with pre-shared cryptographic certificates for all Edge-to-HQ satellite communications [INFERENCE].
4. Encryption Standards: AES-256 encryption for disk storage (at rest); TLS 1.3 encryption for all Modbus-over-TLS or MQTT satellite communications (in transit) [INFERENCE].
5. Data Ownership & Governance: All collected telemetry, forecast logs, and fuel performance data remain the exclusive property of NCPOR, Ministry of Earth Sciences, Government of India [PS] / [INFERENCE].
6. Regulatory Compliance & Safety Validation: Software recommendations must adhere to IEEE 1547 standards for microgrid operation and Madrid Protocol environmental standards [RESEARCHED]. The system provides a permanent immutable audit trail for all automatic generator switching actions [INFERENCE].

16. MVP Scope (Hackathon Build)

MUST BUILD (Core Hackathon MVP)

• Interactive Microgrid Simulator Core: Python-based mathematical simulation of a hybrid polar microgrid (3 x 100 kVA Diesel CHP Gensets, 60 kW Solar PV, 50 kW Wind Turbine, 200 kWh Battery Storage, Station Heat/Elec Loads) running on real Antarctic weather datasets [INFERENCE].
• AI Load & Renewable Forecasting Module: Pre-trained LightGBM / XGBoost models forecasting 24-hour ahead solar PV, wind power, and electrical load curves with live graph outputs [INFERENCE].
• MILP Fuel Optimization Dispatch Engine: Rolling-horizon PuLP optimization script solving generator unit commitment and battery setpoints every 15 minutes, demonstrating explicit fuel savings (Liters/day) against baseline rule-based dispatch [INFERENCE].
• Real-time Operator Dashboard (React.js / Streamlit):
  • Animated Power Flow Diagram (Genset -> Renewable -> Battery -> Load)
  • Real-Time Fuel Runway Counter (Days remaining based on burn rate)
  • Load & Renewable Forecast vs. Actual overlay charts
  • Manual Emergency Override Panel [INFERENCE].
• Simulated Satellite Sync & Bandwidth Monitor: Visual demonstration showing telemetry compression and store-and-forward queue behavior during simulated satellite blackouts [INFERENCE].

SHOULD BUILD (Differentiators if Time Permits)

• Thermal Balance & CHP Heat Recovery View: Visualizing waste heat flow from diesel exhausts to space heating and snow-melting circuits [INFERENCE].
• Battery Health & Cold-Weather Thermal Loss Model: Modeling battery discharge efficiency degradation as temperature drops below 0°C [INFERENCE].
• Scientific Experiment Power Booking Calendar: Interface allowing scientists to reserve high-power slots for analytical instruments [INFERENCE].

NICE TO HAVE (Post-Hackathon)

• Physical hardware Hardware-in-the-Loop (HIL) integration via Modbus RTU PLC kit.
• Automated retraining pipeline at NCPOR HQ.

DO NOT BUILD DURING HACKATHON

• Actual physical hardware deployment in Antarctica.
• Commercial satellite airtime connection (simulated locally).
• Complete replacement of existing station SCADA PLCs.

17. Live Demo Simulation Plan

To deliver a compelling live presentation without access to real-time physical sensors in Antarctica, the team will construct a High-Fidelity Polar Microgrid Hardware/Software Simulator Engine:

1. Simulated Hardware Inputs:
   • Weather Data Feed: Replay historical 1-minute interval weather logs from Antarctic research stations (Larsemann Hills / Svalbard coordinates) containing solar irradiance, wind speed, wind gusts, and ambient temperature [RESEARCHED].
   • Load Data Feed: Synthetic load profile derived from standard polar station operations (base habitat load + fluctuating lab equipment + heating thermal spikes during wind storms) [INFERENCE].
   • Sensor Fault Injector: Live demo toggle buttons on the presenter UI to simulate real-world emergency events:
     • Trigger Katabatic Storm: Wind speed instantly jumps to 35 m/s -> forces wind turbines into mechanical auto-brake cutoff.
     • Trigger Satellite Blackout: Disconnects network interface -> demonstrates edge engine continuing 100% autonomous dispatch and accumulating store-and-forward queue.
     • Trigger Solar Eclipse/Blizzard: Drops solar PV output to zero.
2. Preserving Real Data Flow: The simulator broadcasts data over standard Modbus TCP / MQTT loopback sockets locally on the demo laptop. The Edge EMS software reads these sockets exactly as if it were connected to physical station inverters and engine ECUs [INFERENCE].
3. Clarity on Simulation: The demo interface will feature a clear status banner: "Data Stream: High-Fidelity Antarctic Historical Simulation (Larsemann Hills Dataset)" ensuring total transparency with SIH judges [PS] / [INFERENCE].

18. Implementation Roadmap

Milestone Timeline (Targeting 30 September 2026 Deadline) [PS]

Phase 1: Research & Data Prep      Phase 2: Core Algorithm Build      Phase 3: Integration & UI        Phase 4: Validation & Demo
[Oct 2025 - Dec 2025]              [Jan 2026 - Mar 2026]              [Apr 2026 - Jun 2026]            [Jul 2026 - Sep 2026]
----------------------------------------------------------------------------------------------------------------------------------
* Gather Antarctic weather datasets * Build PyModbus Simulator      * Integrate AI + MILP + Backend  * Conduct stress tests
* Define CHP fuel & BSFC curves     * Train LightGBM Forecasters     * Build React.js/Tailwind UI     * Simulated satellite blackout
* Finalize PuLP MILP formulation    * Develop PuLP Optimization      * Implement Low-BW Sync Queue    * Prepare SIH Grand Finale Demo

• Phase 1: Problem Deep Dive & Data Preparation (October 2025 – December 2025)
  • Extract and clean public polar weather datasets (Larsemann Hills / McMurdo / NREL SAM) [KNOWN].
  • Formulate mathematical BSFC fuel curves for 100 kVA diesel CHP generators [RESEARCHED].
  • Establish baseline fuel consumption under traditional rule-based dispatch [INFERENCE].
• Phase 2: Core Algorithm & Engine Development (January 2026 – March 2026)
  • Train and validate LightGBM / TFT models for load and renewable forecasting [INFERENCE].
  • Implement rolling-horizon MILP unit commitment optimizer in Python (PuLP) [INFERENCE].
  • Build the Python Modbus TCP synthetic microgrid simulator [INFERENCE].
• Phase 3: Systems Integration & User Interface (April 2026 – June 2026)
  • Connect AI forecaster, optimizer, and simulator via FastAPI microservices [INFERENCE].
  • Develop frontend interactive operator dashboard (React.js / Recharts) [INFERENCE].
  • Implement low-bandwidth Protobuf serializer and SQLite offline queue [INFERENCE].
• Phase 4: Validation, Stress Testing & SIH Final Prep (July 2026 – September 2026)
  • Perform false-positive, extreme weather, and satellite disconnection stress tests [INFERENCE].
  • Measure exact fuel savings percentage across 365-day simulated run [INFERENCE].
  • Freeze code, polish demo script, and submit final dossier prior to 30 September 2026 [PS].

19. Testing & Validation Plan

1. Functional Testing: Verify that MILP dispatch output setpoints strictly match physical generator capacity bounds and battery SoC limits [INFERENCE]. Test manual override capabilities [INFERENCE].
2. Performance & Latency Testing: Measure optimization solver execution time on low-power edge hardware (NVIDIA Jetson / Raspberry Pi 4). Benchmark must achieve complete MILP solution in < 15 seconds for a 24-hour horizon [INFERENCE].
3. AI Model Accuracy Testing: Evaluate solar, wind, and load forecasting models across unseen Antarctic winter and summer test sets. Target accuracy: MAPE < 8% for 1-hour ahead, MAPE < 12% for 24-hour ahead [INFERENCE].
4. False Positive & Edge Case Testing:
   • Inject sudden 200% load spike -> verify BESS and secondary generator ramp up without microgrid tripping [INFERENCE].
   • Inject sudden katabatic wind cutoff -> verify system smoothly starts standby diesel generator before battery exhausts [INFERENCE].
5. False Negative Testing: Ensure AI model uncertainty bounds prevent under-estimation of station heating loads during sub-zero cold waves [INFERENCE].
6. Offline & Satellite Resilience Testing: Sever all external network interfaces for 72 simulated hours. Verify zero data loss in local SQLite TSDB and complete autonomous control execution [INFERENCE].
7. Fuel Savings Validation: Run a comparative 30-day simulation comparing the Proposed Hybrid AI-MPC system against a traditional static threshold controller. Objective metric: Demonstrate >= 20% total diesel mass reduction [INFERENCE].

20. Risk Register

Risk	Likelihood	Impact	Severity	Mitigation Strategy
R-1: Optimization Solver Timeout on Edge PC	Medium	High	High	Implement strict 10-second timeout on MILP solver. Fall back to a fast rule-based heuristic controller if solver times out [INFERENCE].
R-2: Severe AI Forecast Drift During Polar Storms	High	High	High	Incorporate dynamic confidence intervals. Expand safety spinning reserve capacity when forecast uncertainty exceeds 20% [INFERENCE].
R-3: Battery Thermal runaway or low-temp lockout	Low	High	High	Enforce physical safety constraints in MILP solver to prioritize CHP waste heat routing to BESS thermal jackets when ambient temp < -10°C [RESEARCHED].
R-4: Hardware Failure of Primary Edge Server	Low	Critical	High	Implement dual-redundant active-passive edge hardware nodes with automated heartbeats [INFERENCE].
R-5: Loss of Satellite Sync for > 7 Days	Medium	Low	Medium	Edge database designed with circular time-series buffer and heavy Protobuf compression, retaining up to 90 days of local metrics [INFERENCE].
R-6: Generator Wet Stacking due to low load operation	Medium	Medium	Medium	Include hard constraint requiring any active diesel generator to operate at minimum 35% nominal load [RESEARCHED].

21. Cost & Scalability

Hackathon Phase Costs

Item	Cost (INR)	Justification
Compute / Local Dev Hardware	₹0	Student laptops and free open-source software (Python, React, Docker) [INFERENCE].
Open-Source Datasets & Tools	₹0	Public NREL, NOAA, and Open-Meteo Antarctic datasets; PuLP/CBC solvers [KNOWN].
Edge Hardware Prototype	₹15,000	Single NVIDIA Jetson Nano / Raspberry Pi 4 for live physical edge demo (Optional) [INFERENCE].
Total Hackathon Cost	~₹0 - ₹15,000	Highly cost-effective software prototype [INFERENCE].

Production Deployment Costs (Per Polar Station - Estimated)

Item	Cost Estimate	Justification
Dual Industrial Edge Servers	$8,000 - $12,000	Siemens / Advantech fanless, wide-temperature (-40°C to 60°C) dual redundant IPCs [INFERENCE].
Modbus / CANbus Gateway Hubs	$3,000 - $5,000	Industrial-grade protocol converters and optoisolators for SCADA connection [INFERENCE].
On-Site Integration & Testing	$25,000 - $40,000	Specialized polar engineering personnel logistics during annual expedition [RESEARCHED].
Software Licensing & Maintenance	$0 (Open-Source Core)	Government-owned custom codebase with ongoing MoES maintenance [INFERENCE].
Total Production Capex	~$36,000 - $57,000	Payback Period < 6 Months (A 25% diesel saving on a base burning 300,000L/year saves ~75,000L fuel = >$750,000/year saved!) [RESEARCHED] / [INFERENCE].

22. Differentiation

Dimension	Generic Hackathon Submission	Our Solution (POLAR-EMS)	Why It Wins
Core Control Logic	Basic Python if/else scripts or simple rule engines.	Hybrid AI-Predictive MPC + MILP Mathematical Optimization Engine.	Solves complex multi-variable unit commitment while guaranteeing 100% hard safety constraints [INFERENCE].
Polar Specificity	Treats microgrid like a standard sunny solar farm in mainland India.	Joint Thermal-Electrical Modeling (CHP Waste Heat + Snow Melting + Katabatic Storm Cutoff).	Directly addresses the real-world life-support realities of Antarctic bases (Bharati/Maitri) [RESEARCHED].
Network Architecture	Requires continuous cloud connectivity (Firebase / AWS).	Edge-Native Local TSDB + Low-Bandwidth Protobuf Satellite Sync Queue.	Guarantees continuous 100% station autonomy during multi-day Antarctic satellite blackouts [INFERENCE].
Safety & Governance	Black-box Neural Network / Reinforcement Learning.	Deterministic MILP Solver with Hardware Safety Interlocks.	Fully explainable, safe, and deployable in mission-critical government polar infrastructure [INFERENCE].
Economic Value Realization	Vague claims of "saving energy".	Quantified Fuel Runway (Days) & Exact BSFC Mass Burn Optimization Calculation.	Proves direct ROI (>75,000 Liters fuel saved annually per station) to MoES judges [RESEARCHED] / [INFERENCE].

23. Future Scope

1. Hydrogen Electrolyser & Fuel Cell Integration: Expanding the MILP dispatch engine to model green hydrogen production during excess summer solar/wind periods, storing hydrogen for zero-emission winter heating (aligning with China's Qinling base benchmark) [RESEARCHED].
2. Multi-Station Swarm Optimization: Extending satellite coordination so Bharati, Maitri, and upcoming Maitri II stations share predictive logistics data and coordinate fuel supply vessel demands [RESEARCHED].
3. Computer Vision Radome Snow-Clearing Inspection: Integrating edge cameras on wind turbine blades and radomes using YOLOv8 to detect ice build-up and trigger automated heating elements [INFERENCE].

24. Assumptions & Open Items Log

Consolidated Assumptions Log

• [ASSUMPTION-1]: The target polar research station utilizes standard 3-phase AC microgrid distribution with industrial Modbus TCP/RTU or CANbus interfaces on diesel generators and inverters.
• [ASSUMPTION-2]: Historical weather data from coastal East Antarctica (Larsemann Hills / Schirmacher Oasis) is representative of operational conditions for Bharati and Maitri stations.
• [ASSUMPTION-3]: Diesel generators at the station are equipped with Combined Heat and Power (CHP) heat-exchangers capturing at least 40–50% of exhaust/jacket waste heat for thermal space heating.
• [ASSUMPTION-4]: Station scientists and engineers are willing to input scheduled high-power experiments into a 24-hour advance booking interface.

Consolidated Validation Log (Items Requiring Validation)

• [NEEDS VALIDATION-1]: Exact brand, model, and PLC interface protocols of existing diesel CHP generators currently running at Bharati and Maitri stations (e.g., Caterpillar vs. Cummins).
• [NEEDS VALIDATION-2]: Current battery storage (BESS) installed capacity (kWh), battery chemistry (LiFePO4 vs Lead-Acid), and operating thermal bounds at NCPOR bases.
• [NEEDS VALIDATION-3]: Exact satellite bandwidth allocation and daily data quota for non-scientific operational SCADA telemetry transmitted between Antarctica and NCPOR Goa.

Final Recommendation

Adopt Approach 3: Hybrid AI-Predictive MPC + MILP Edge Optimization Framework. Build a Python-based edge EMS core utilizing LightGBM for load/renewable forecasting, PuLP (COIN-OR CBC) for 15-minute rolling-horizon fuel minimization, an SQLite time-series local cache, and a React.js operator dashboard. Simulate the physical polar microgrid using real Antarctic meteorological data.

Recommended Stack

• Edge Compute: Ubuntu Linux / Docker / Python 3.11 / FastAPI
• AI & Optimization: LightGBM / PyTorch / PuLP (CBC Solver) / SciPy
• Database & Storage: TimescaleDB / SQLite / Protobuf
• Protocols: PyModbus / Eclipse Mosquitto (MQTT)
• Frontend: React.js / Tailwind CSS / Recharts / Leaflet.js

Biggest Dependency

Access to realistic time-series load profiles and weather observations from high-latitude polar environments (achieved via NREL SAM and Open-Meteo Antarctic historical APIs).

Biggest Risk

AI forecasting performance degradation during sudden, unpredicted polar katabatic wind storms, causing premature battery depletion or generator thermal shock (mitigated by dynamic confidence bounds and strict hardware safety interlocks).

Biggest Technical Challenge

Formulating a unified, computationally efficient Mixed-Integer Linear Program (MILP) that simultaneously balances electrical power constraints AND thermal heat recovery requirements in real time on resource-constrained edge hardware.

First 5 Actions

1. Extract and clean 1-year historical weather and solar/wind data for Larsemann Hills, Antarctica coordinates from Open-Meteo / NREL.
2. Develop a standalone Python mathematical model of a 100 kVA diesel CHP generator including its Brake Specific Fuel Consumption (BSFC) curve.
3. Formulate the core PuLP MILP unit commitment script to optimize dispatch between 2 generators, 1 solar PV array, 1 wind turbine, and 1 BESS.
4. Train a LightGBM regressor on weather features to forecast 24-hour ahead renewable power output.
5. Assemble the FastAPI backend loop and connect it to a lightweight React.js power-flow dashboard.

Validation Before Coding

Validate the mathematical formulation of the MILP objective function to ensure that generator startup costs, minimum run-times (to prevent wet stacking), and mandatory thermal space-heating demands are strictly satisfied before writing UI code.

Buildability Score /100

92 / 100 (Extremely high feasibility for a 4–6 student team: all software dependencies are open-source, public polar weather data is accessible, and the scope avoids physical hardware dependencies through high-fidelity socket simulation).

Confidence

HIGH (Supported by established microgrid control literature, Antarctic Treaty environmental frameworks, and documented operational benchmarks from Belgian, US, and Chinese polar stations).

Research Sources

1. Title: Bharati Research Station & Maitri Infrastructure Specifications [RESEARCHED]

   • URL: https://ncpor.res.in/antarcticas/display/376-maitri-
   • Source Type: Official Government Portal (NCPOR / MoES)
   • Supported Claim: Station operational capacity, 3x100 kVA CHP power units, 300,000L fuel farm, C-band satellite link, and location coordinates [PS] / [RESEARCHED].
   • Reliability: Official Government Source (Highest).

2. Title: Princess Elisabeth Antarctica Research Station: Micro Smart Grid [RESEARCHED]

   • URL: http://www.antarcticstation.org/station/smart_grid
   • Source Type: Official Polar Organization Documentation (International Polar Foundation)
   • Supported Claim: Zero-emission polar microgrid architecture, priority load-shedding mechanisms, wind/solar/battery integration in Antarctica [RESEARCHED].
   • Reliability: Official Operating Organization Documentation (High).

3. Title: China's Antarctic Station Goes Green with Scaled New Energy System (Qinling Station) [RESEARCHED]

   • URL: http://english.scio.gov.cn/chinavoices/2025-04/09/content_117814292.html
   • Source Type: Government News Portal (SCIO China, April 2025)
   • Supported Claim: Hybrid microgrid integration (wind, solar, hydrogen, diesel) achieving 60% green power in Antarctica [RESEARCHED].
   • Reliability: Official Government News Source (High).

4. Title: Mapping Renewable Energy among Antarctic Research Stations [RESEARCHED]

   • URL: https://www.mdpi.com/2071-1050/16/1/426
   • Source Type: Peer-Reviewed Academic Publication (MDPI Sustainability, 2024)
   • Supported Claim: Microgrid generation patterns, diesel backup dependency, and extreme thermal-electrical coupling challenges in Antarctica [RESEARCHED].
   • Reliability: Academic Peer-Reviewed Journal (High).

5. Title: Artificial Intelligence-Based Energy Management and Control Strategies for Renewable-Powered Smart Microgrids [RESEARCHED]

   • URL: https://www.mdpi.com/2571-8797/8/5/136
   • Source Type: Peer-Reviewed Academic Publication (MDPI Clean Technologies, August 2026)
   • Supported Claim: Comparison of Rule-Based, ANN, and MPC/RL control strategies for microgrid voltage, battery SoC, and fuel optimization [RESEARCHED].
   • Reliability: Academic Peer-Reviewed Journal (High).