Integrated planning and operation of distributed power generators and energy storages for active distribution networks by new ML-based two-level optimization model

• Two-level framework integrates DG planning with real-time dispatch control. • COA optimizer used for siting and sizing of PV, CHP units, and BESS devices. • SQP applied for hourly dispatch, with full SoC, ramp, and voltage limits. • ML-based forecasts (MRMI-LSTM, WT–MIIG–MLP) handle load, solar, and price. • IEEE 33-bus study cuts losses 35.8%, costs 22%, and boosts voltage by 85%. The increasing integration of renewable energy sources (RES) and distributed generators (DGs) into modern distribution systems presents both opportunities and challenges in achieving efficient, reliable, and cost-effective operations. This paper proposes a novel two-level optimization framework based on Machine Learning (ML) algorithms for Active Distribution Networks (ADN), which simultaneously optimizes DG placement, sizing, and real-time energy management. At the outer level, a Coot Bird Search Algorithm (CBSA) is employed to determine the optimal siting and sizing of photovoltaic (PV), combined heat and power (CHP), and battery energy storage systems (BESS), while ensuring unique bus assignments and realistic capacity limits. The inner level applies Sequential Quadratic Programming (SQP) to optimize hourly dispatch over a 24-hour period, considering operational constraints such as battery state of charge (SoC), CHP ramping limits, voltage bounds, and network power losses. To account for real-world uncertainties, advanced forecasting methods are integrated into the framework, including Modified Relief-Mutual Information-Long Short Term Memory (MRMI-LSTM) for load and solar generation, and a cascaded Wavelet Transform-Mutual Information Interaction Gain-Multilayer Perceptron (WT-MIIG-MLP) model for electricity prices. Simulation results demonstrate that the proposed method reduces network power losses by 35.8%, enhances voltage stability by 85.1%, and lowers operational costs by 22% compared to the base case without Distributed Energy Resources (DERs). These findings highlight the effectiveness of the proposed hybrid framework in achieving economic operation while enhancing technical performance, offering a practical solution for the sustainable integration of distributed energy resources in modern power systems.