The existing electric grid system is facing many challenges: aging infrastructure, growing energy demand, lack of fault-tolerance, lack of diversification of energy supply, emission of carbon dioxide, etc. This encourages the roll out of smart grid system to reinforce the existing grid system. One of the key technologies of smart grid is two-way communication between smart grid components. As smart grid covers a wide range of geographical area, wireless communication outperforms wired communication. However, it is economically not viable to assign a particular band of spectrum for smart grid communication. Besides, the problems encountered in today's spectrum management include scarcity of spectrum, inflexibility of static spectrum assignment, and underutilization of the spectrum. The concept of cognitive radio is an attractive solution for smart grid. By deploying cognitive radio based smart grid, the smart grid component will perform spectrum sensing to look for idle spectrum and opportunistically utilize the idle spectrum. In view of that, the spectrum usage pattern has to be investigated before any justification of a suitable framework for cognitive radio based smart grid is made. In this project, we conducted several spectrum measurements in Malaysia. Antenna and other measurement equipment were set up at Universiti Tenaga Nasional, MIMOS, and Universiti of Malaya to capture 24-hour measurement data. We focused on TV band, GSM band, 3G band, LTE band, ISM band, and WiMAX band. To verify the spectrum usage, the duty cycle (DC) method was used as an indicator of the activeness of the spectrum band. We found that the bands are moderately to heavily underutilize. Based on these spectrum occupancy measurement results, we suggest a feasible cognitive radio based smart grid framework.