To enable granular spatial planning under high distributed energy resource penetration, this paper proposes a multi-source hierarchical forecasting framework that reconciles net electricity consumption and installed solar self-consumption capacity. Utilizing real-world smart metering and production unit records from the E-REDES Open Data Portal, demand and capacity dynamics are modeled across four administrative levels in mainland Portugal: national, district, municipal, and parish tiers. Candidate baseline statistical models, including AutoARIMA, AutoETS, and MSTL, are evaluated alongside Bottom-Up, Top-Down, Middle-Out, and Minimum Trace (MinTrace) reconciliation techniques to guarantee strict spatial coherence. The empirical evaluation demonstrates that AutoARIMA paired with MinTrace reconciliation achieves optimal cross-hierarchy alignment, yielding the most robust continuous growth trajectories for solar self-consumption capacity. For net load forecasting, Top-Down reconciliation preserves macro-scale aggregate precision, whereas Bottom-Up captures localized consumption volatility driven by behind-the-meter PV adoption at the granular parish level. Furthermore, spatial disaggregation of the reconciled forecasts identifies critical high-growth expansion clusters, equipping distribution system operators with actionable insights to prioritize local grid reinforcements and effectively mitigate hosting capacity constraints. Overall, this framework optimizes model selection to support data-driven distribution network planning.