<![CDATA[This research evaluates three Intermittent Demand Forecasting (IDF) methods Croston, TSB, and ADIDA for non-routine investment disbursement at PLN UIP Sulawesi, characterized by intermittent and lumpy patterns. Using a three-stage analytical approach (in-sample validation on 29 PRKs, rolling origin out-of-sample on 180 PRKs, and production forecast on 959 PRKs) with MASE, MAE, and ME metrics, ADIDA emerged as the best method with consistent win-rate dominance (58.6% in Phase 1; 64.2% in Phase 2). Hierarchical reconciliation using Weighted Least Squares (WLS) produced a coherent 2026 forecast of IDR 2,781.89 billion, outperforming Bottom-Up and Top-Down approaches in hold-out validation. These findings provide an empirical foundation for data-driven cash planning at PLN UIP Sulawesi, with the ADIDA–WLS pipeline offering a validated, auditable, and computationally efficient framework for Monthly Cash Budget preparation. The methodological framework's auditability particularly the traceability of WLS weight matrices to empirical error sources supports readiness for SAP S/4HANA Single Source of Truth (SSoT) Wave 3 implementation scheduled for October 2027, while the PRK-level forecast matrix enables hybrid statistical-expert judgment review to address administrative zero constraints. This research contributes theoretically by extending IDF literature beyond spare parts inventory into infrastructure investment cash management, and practically by providing operational tools for working capital optimization, KPI cascading, and systematic investment portfolio phasing within PLN's Transformation 2.0 agenda.]]>
