Due to its attractive characteristics, the TV White Space (TVWS) spectrum is considered the ideal candidate to enable the deployment of Smart Grid Networks (SGNs) via Cognitive Radio paradigm. However, the intermittent availability of the TVWS spectrum as well as its scarcity in urban scenarios could compromise the tight Smart Grid requirements in terms of reliability, latency and data rate. This degradation could be even more severe when mobile grid nodes, e.g. electric vehicles, are considered. Stemming from this, we first develop an analytical framework to account for the mobility in SG scenarios. Then, we design a switching procedure based on the use of two different bands: TVWS spectrum and ISM spectrum. The switching procedure selects, among the available spectrum bands, the one maximizing the achievable throughput at an arbitrary SGN. Such a procedure accounts for the presence of interfering SGNs on the TVWS spectrum through both their traffic and mobility patterns. By wisely using both the ISM and the TVWS spectrum, the proposed switching procedure is able to: i) increase the achievable data rate, and to ii) reduce the outage event rate, improving the reliability and the latency of the Smart Grid communications. Moreover, we show the performance of the proposed switching procedure depends largely on the time devoted to sense. Hence, the proper setting of such a parameter is critical for the performance of any SGN. For this, we derive an optimization criterion maximizing the throughput under the constraint of bounding the outage rate. The theoretical analysis is validated through extensive numerical simulations.