Welcome to NOAH
Wastewater Volume Generation Forecasting
Atlantic City, New Jersey

Problem Definition

Atlantic County Utilities Authority (ACUA) treats wastewater generated within its service area, which includes the casino and shore tourist community for Atlantic City, New Jersey. Wastewater volume is a surrogate for energy consumption for ACUA, as larger volumes require higher energy expenditures, and vice versa. In this preliminary feasibility study conducted with WKG Consultants, several artificial neural network (ANN) models were developed to forecast wastewater volumes for the Atlantic City for three different forecasting periods; one-day ahead, seven-days ahead, and thirty-days ahead. The types ofinput variables consisted of recent historical weather conditions preceding the prediction period, “future” weather conditions, which in this case, were know a-priori, and recent historical wastewater volume data. Recent historical wastewater volume data was used to account for the fact that there typically is a “memory” or correlation between recent wastewater volumes and future volumes. The weather variables, consisting of precipitation and temperature, are used to account for the fact that wastewater volumes are generally influenced by weather patterns. For example, during a hot dry spell, residents may water lawns more frequently, and a larger number of vacationers and day trippers may go to the shore, increasing water consumption. Because water consumption determines wastewater volume, this type of consumer demand relationship between wastewater volume and weather was deemed appropriate.

 

Results

Tables 1 presents the statistical results for the daily ANN forecasting model, which achieved the high correlation coefficient of 0.86 during validation.

Figure 1 below compares measured versus forecasted wastewater volume for the validation data for this model.

The figure clearly demonstrates that the ANN model accurately tracks the higher and lower wastewater volumes. It should be noted that the mean absolute change between consecutive days in wastewater volume over the study period is approximately 0.62 million gallons, while the mean absolute error for the Complex ANN model was 0.46 million gallons during validation. Consequently, the ANN model is not merely “keying” off the previous day’s wastewater volume, but is extracting some relationships between the input predictor variables and the future wastewater volume.

Tables 2 and 3 summarize the statistical results for the 7-day and 30-day ahead forecasting periods, and Figures 2 and 3 depict the model validation results for these two periods, respectively.

The high correlation coefficients and low absolute mean errors, particularly with respect to the total wastewater volumes over each of the prediction periods, demonstrate that the ANNs are able to extract relationships between recent historical wastewater volume patterns and weather conditions with future wastewater volume. By extension, it is expected that energy consumption by ACUA will similarly be related to recent historical energy consumption and wastewater volume data as well as recent historical and future projected weather conditions.

This water demand/wastewater volume capability can be used to minimize energy consumption costs by shifting load curves as possible to offset on-peak cost periods. In addition, maintenance operations can be scheduled to coincide with lower wastewater volume generation periods. Lastly, the forecasting capability can also be used to anticipate the volume of potable water that will be extracted for wastewater purposes.