This study examines how climate change affects photosynthetically active radiation (PAR) and solar agrivoltaic systems (AVS) in Nigeria's six geopolitical regions and seasonal cycles using sophisticated machine learning and hybrid analytical models. December-January-February (DJF) PAR peaks in the Central and Northwest (120-126 W/m2), but decreases in the South and Southeast (109-114 W/m2), according to regional-seasonal analysis. The humid South/Southeast experiences declines in JJA (74-90 W/m2), while the Northeast maintains high values (100-130 W/m2). A hybrid of particle swarm optimization and controlled autoregressive integrated moving average (PSO-CARIMA) was evaluated in conjunction with six machine learning algorithms. PSO-CARIMA outperformed all of them, achieving a coefficient of determination (R2) up to 0.999 and a root mean square error (RMSE) as low as 0.0020. SARIMA and CARIMA performed best in the South/Southeast (R2 = 0.998, RMSE <= 0.0086) and Southwest, respectively. Shared Socioeconomic Pathways (SSP) climate projections show notable changes: SSP245 (2.5 degrees C) moderate emission scenario (-12 % Borno) with gains in Lagos/Niger; SSP585 (3.5 degrees C) high emission scenario yields extremes, from -13 % (Niger, SON) to + 28 % (Kebbi, JJA); SSP126 (1.5 degrees C) low emission scenario shows JJA PAR losses of -23 % in Borno/Yobe but March-April-May (MAM) gains of up to 18 % in the Northeast/Southwest. AVS efficiency reacts nonlinearly: polycrystalline silicon (p-Si) exhibits higher variability, while monocrystalline silicon (m-Si) shows greater resilience, with SSP585 gains of + 19.64 % (Maiduguri, JJA) in arid zones but losses in humid ones. The findings highlight the necessity of region- and climate-specific AVS deployment plans and PAR forecasting.