Perancangan sistem manajemen energi pada smart building menggunakan pendekatan Multi-Agent Deep Constrained Q-Learning (MADQN). Dengan memanfaatkan perangkat cerdas seperti HVAC (Heating, Ventilation, and Air Conditioning), pemanas air, dan mesin cuci sebagai agen, sistem ini mengoptimalkan konsumsi energi dan mengurangi biaya operasional. Melalui simulasi berbasis reinforcement learning, sistem ini menga- daptasi kondisi lingkungan yang berubah dan meminimalkan pemborosan energi. Hasil simulasi menunjukkan bahwa penerapan MADQN dapat mencapai efisiensi energi yang lebih tinggi dibandingkan dengan sistem berbasis aturan konvensional.

---

This research aims to design a energy management system for smart buildings using the Multi-Agent Deep Constrained Q-Learning (MADQN) approach. By utilizing intel- ligent devices such as HVAC (Heating, Ventilation, and Air Conditioning), water heaters, and washing machines as agents, this system optimizes energy consumption and reduces operational costs. Through reinforcement learning-based simulation, the system adapts to changing environmental conditions and minimizes energy wastage. Simulation results show that the implementation of MADQN achieves higher energy efficiency compared to conventional rule-based systems.