In order to estimate the remaining driving range (RDR) in electric vehicles, the remaining discharge energy (E-RDE) of the applied battery system needs to be precisely predicted. Strongly affected by the load profiles, the available E-RDE varies largely in real-world applications and requires specific determination. However, the commonly-used direct calculation (DC) method might result in certain energy prediction errors by relating the E-RDE directly to the current state of charge (SOC). To enhance the E-RDE accuracy, this paper presents a battery energy prediction (EP) method based on the predictive control theory, in which a coupled prediction of future battery state variation, battery model parameter change, and voltage response, is implemented on the E-RDE prediction horizon, and the E-RDE is subsequently accumulated and real-timely optimized. Three EP approaches with different model parameter updating routes are introduced, and the predictive-adaptive energy prediction (PAEP) method combining the real-time parameter identification and the future parameter prediction offers the best potential. Based on a large-format lithium-ion battery, the performance of different E-RDE calculation methods is compared under various dynamic profiles. Results imply that the EP methods provide much better accuracy than the traditional DC method, and the PAEP could reduce the E-RDE error by more than 90% and guarantee the relative energy prediction error under 2%, proving as a proper choice in online E-RDE prediction. The correlation of SOC estimation and E-RDE calculation is then discussed to illustrate the importance of an accurate E-RDE method in real-world applications. (C) 2015 Elsevier Ltd. All rights reserved.