Research and Analysis of ARIMA Model Based on Time Series Analysis in Smart Water

In
$a$
smart water system, detecting water leaks is crucial, especially detecting outliers in the difference between inlet and outlet water flow in 8 virtual zones. This study chooses the ARIMA model in time series analysis to address this problem. The ARIMA model processes the data though autoregressive (AR), moving average (MA), and differential to complete model identification and order determination and perform stationarity and white noise tests. This model effectively identifies and handles outliers, typically detected through autocorrelation function (ACF) and partial autocorrelation function (PAC F) analysis. Outliers may appear as significant peaks on the ACF and P ACF plots or points beyond the confidence interval. Explaining the type of outliers is helpful for subsequent analysis. We also analyze each outlier in detail to examine the optimization and model tuning, R-squared, and the significance of data stability in each region. Finally, model visualization is performed to evaluate the model fit of time series analysis. These steps will help to solve the problems that occur in water leakage detection efficiently.