This document presents a study on short-term photovoltaic (PV) power forecasting using the Jordan recurrent neural network (JRNN) method. Temperature and solar radiation data from Surabaya, Indonesia were used as inputs to the JRNN model to forecast PV power. The JRNN model was trained and tested, achieving a mean square error of 0.9858 and mean absolute percentage error of 1.3311 in testing, with a processing time of 4.59 seconds. The forecasting results from JRNN were more accurate compared to an artificial neural network model, with lower error rates, though JRNN required more processing time. JRNN is an effective method for short-term PV power forecasting based on weather