Case study / 08
Six agents coordinating one physical energy system.
A distributed BDI system for monitoring, forecasting, orientation optimization, and predictive maintenance.
- Role
- AI & System Architect
- Scope
- Agent design · Forecasting · Event backbone · Operations
- Status
- Research and engineering project
- Year
- 2025–2026
01 / CONTEXT
The problem
Coordinate predictions and physical actions in real time while keeping responsibilities, conflicts, and operating data understandable.
02 / RESPONSE
The response
Six BDI agents communicate through MQTT and combine shared time-series data, LSTM forecasts, anomaly detection, arbitration, and actuator commands.
03 / EXPERIENCE
Product flow
- 01
Observe
Collect solar, equipment, and weather signals.
- 02
Predict
Forecast production and conditions.
- 03
Propose
Recommend orientation and maintenance.
- 04
Arbitrate
Resolve priorities and constraints.
- 05
Act
Apply commands and measure feedback.
05 / SYSTEM
System architecture
An event-driven loop converts live context into coordinated optimization and maintenance.
06 / TRADE-OFFS
Architecture decisions
BDI responsibilities
Each agent owns a coherent goal and action boundary.
MQTT backbone
Publish/subscribe decouples live services.
Coordinator arbitration
Conflicts are resolved before physical action.
07 / QUALITY
Security & reliability
- A coordinator resolves competing recommendations.
- Redis shares operational context.
- InfluxDB preserves time-series observations.
08 / EVIDENCE
Outcomes & evidence
- A six-agent model spanning monitoring to action.
- A reusable event protocol for decisions.
- A platform for forecasting and maintenance experiments.
09 / STACK