ARIMA Time Series Modelling for Energy Forecasting in Wireless Sensor Networks

Energy conservation is critical in wireless sensor networks since it affects the sensor's lifespan. Reducing the frequency of transmission is one way to reduce expenses, but it must not compromise the accuracy of the data that is being received. Hence, this paper has developed an autoregressive integrated moving average (ARIMA) time-series-based model to improve the prediction approach. The proposed ARIMA-based model assures the characteristics of the nodes that remain idle for extended times to conserve energy during inactive periods. It forecasts a building's energy usage based on the data gathered (e.g., day of the week, light energy, temperature, humidity, etc.). The proposed methodology efficiently preserves the constrained battery power of wireless sensor nodes while maintaining the predicted data values within the application-defined error bounds. Through experiments, it has been shown that these predicted data values closely match the actual observed data values and requires less communication between sensor nodes and aggregators than the actual data aggregation method. The proposed methodology enhanced the prediction accuracy as compared with existing approaches. This approach produces the mean absolute error (MAE) as 45.06.