Artificial intelligent power prediction for efficient resource management of WCDMA mobile network

This paper presents a method of predicting changes of power consumption at Node B of a wideband code division multiple access (WCDMA) mobile network due to dynamic resource allocation such as movement of unit equipment (UE), handover call from adjacent cell and accommodation of new service request. The method learns the mapping of power consumption at Node B by monitoring power changes that response to previous performed resource allocation. Estimation of the unknown function is implemented with support vector regression (SVR). The output of SVR will be used by WCDMA mobile network to decide on new service admission. Genetic algorithm (GA) is then applied to form optimal beams to cover all UEs in a cell with minimum power. This artificial intelligent call admission control (CAC) was validated using a dynamic WCDMA mobile network simulator. A few comparative results in downlink have shown that our integrated support vector regression assists genetic algorithm (SVRaGA) is capable of predicting next interval power consumption at Node B with low prediction error and improving the quality of service (QoS) by reducing dropped calls.