Cell-free massive multiple input multiple output (CF-mMIMO) offers users a more consistent service quality than a traditional cellular topology. Only a set of access points (APs) need specific user group to meet the performance requirements after high traffic hours. This paper addresses the optimization of load balancing and energy efficiency (EE)-based CF-mMIMO (EE-CF-mMIMO) networks to minimize total downlink power consumption (PC), with transmit powers and hardware dissipation. Despite the non-convex nature of the optimization issue, a better solution of a mixed-integer second-order cone program (MISOCP) is used for optimal solution from the transmit power and hardware dissipation. The two algorithms with low complexity are also included that take use of the issue formulation's optimal transmit powers and intrinsic group sparsity. The proposed MISOCP-based EE-CF-mMIMO method achieved better results by comparing with joint optimization for user association and content caching (JUACC) and deep Q-network-based CF-mMIMO (DQN-CF-mMIMO) with 95.63% of EE and 29.45% of PC.