Light-emitting diode (LED) arrays have attracted increased attention in the area of high power intelligent automotive headlamps because of their superiority in disposing of the power limit of an individual LED package and controllably luminous intensity and illumination pattern. The optical and chromatic performances of an LED array do not equal to the sum of individual LED packages’ performances, as the thermal interactions between individual LED packages can’t be ignored in the actual application. This paper presents a thermal-electrical-spectral (TES) model to dynamically predict the optical and chromatic performances of the LED array. The thermal-electrical (TE) model considering the thermal coupling effect in the LED array is firstly proposed to predict the case temperature of each individual LED package, and the Spectral power distributions (SPDs) of individual LED package is then decomposed by the extended Asym2sig model to extract the spectral characteristic parameters. Finally, the experimental measurements of the designed LED arrays operated under usage conditions are used to verify the TES model. Some validation case studies show that the prediction accuracy of the proposed TES model, which is expressed as a quadratic polynomial function of current and case temperature, can be achieved higher than 95%. Therefore, it can be concluded that this TES model offers a convenient method with high accuracy to dynamically predict the optical and chromatic performances of LED arrays at real usages.