Modules
The platform has three core modules, an AI assistant across them, and a contract-review module planned for later.
Remote monitoring
Metering devices at each plant send voltages, currents, power, power factor, frequency and harmonics every few seconds. Operators see live readings and trends for each plant and are alerted when something looks wrong.
AI methods
Savitzky–Golay filtering smooths noisy signals without hiding real changes. Anomalies are detected in layers: EWMA control limits catch sudden shifts, Isolation Forest finds unusual combinations of readings, and an LSTM autoencoder learns normal behaviour over time and flags what does not fit.
Cost modeling
Generation cost per kWh is calculated from fuel, operation and maintenance, capital and other cost parameters, and can be recalculated for different scenarios.
AI methods
Levelised cost (LCOE) and marginal cost give a transparent baseline. A hybrid model combines physical plant relationships with machine-learning regression, and gradient-boosting ensembles predict costs from operating data.
Decision support and control
Forecasts and recommendations help operators plan generation, maintenance and load shedding. Control starts as advice only; assisted and automatic modes come later and only with approval.
AI methods
LSTM networks forecast daily and long-term generation and load shedding. An autoencoder combined with linear programming and XGBoost models produce recommendations. Control runs in three modes: advisory, assisted and automatic.
AI assistant
A chat assistant answers questions about the platform data and helps prepare reports.
AI methods
A large language model (LLM) that works over the platform data and generates report drafts.
Power purchase agreement reviewPlanned for a later phase
A tool to read power purchase agreements and point out clauses that need attention.
AI methods
Natural language processing and LLMs extract key clauses and flag risks.