A multi layered approach is proposed to provide energy and spectrum efficient designs of Cognitive Radio Based Wireless Sensor Networks (CR-WSN) at the smart grid utility. CR-WSNs consume higher power and involve complex design procedure. Since smart grid wireless applications being used currently are meant to save power, we model the CR-WSN based home area networks by keeping minimum power consumption as the key criteria. We present an efficient design methodology spanning across multiple layers. We propose some key modifications of RPL (Routing protocol for low power and lossy networks) to suit our requirements. This provides us reliable and low-latency routing support for large-scale cognitive smart grid networks. Our RPL implementation handles finer aspects of the network such as asymmetry, device capacity handling which is performed through selective routing where battery powered devices are spared from spectrum sensing operation and minimum number of routes passes through it.