The emergence of the field of smart grid data mining in the past years has an increase of interest in load profile analysis. The load profile clustering is used to discover the customer power consumption patterns from the AMI data. This paper examines how the wavelet-based clustering algorithm improves the capability to discriminate among the load profiles clusters in manufacture industry according to their AMI time series data. We cluster the manufacture customers in our sample according to their monthly power consumption behaviour in 2012. Combining the different wavelet level and k-means algorithm, the results can find out the daily and weekly power consumption patterns. The knowledge from load profile analysis will add empirical understanding of the problems to the related research groups and contribute to the future best practice in the energy industry.