The worldwide share of renewables in energy generation is over 28%. With the growing energy consumption patterns and urge to transition from fossil fuel-based energy to cleaner sources i.e., renewable energy generation is continuously rising. Due to the intermittent nature of the renewables, continuous load monitoring is required. Various techniques are available in literature to perform Non-Intrusive Load Monitoring (NILM) to segregate load consumption of a building to individual appliances. In this paper, the effect of various sampling frequencies is investigated. Low sampling rates from Is to 30s are used. The data sampled at different resolutions are used to train convolution neural networks to draw performance comparisons of event detection and appliance recognition at various sampling rates. And finally, the overall accuracy of the model is calculated.