Forecasting of municipal solid waste generation using non-linear autoregressive (NAR) neural models

Abstract

Municipal solid waste (MSW) generation is a multi-variable dependent process and hence its quantification is relatively not easy. The estimations for monthly MSW generation are required to provide theoretical guidelines for understanding and designing the disposal system. These estimations help in budgetary planning for the handling of future waste with optimized waste management system. This study forecasts the monthly MSW generation in Nagpur (India) for the year 2023 using non-linear autoregressive (NAR) models. The classical multiplicative decomposition model with simple linear regression in time series was constructed with maximum absolute error of 6.34% to overcome the problem of data availability. It was observed that NAR neural models were able to predict short-term monthly MSW generation with absolute maximum error of 6.45% (Model A) and 3.05% (Model B) for the observation period. It was also concluded that the variation in MSW generation was best captured when yearly lagged values were used for the construction of NAR model and coefficient of efficiency (E) was 0.99 and 0.97 during testing and validation, respectively. It was found that in the year 2023, the city will record minimum waste generation in the month of February and maximum in the month of December. For the year 2023, it had been estimated that the maximum 48504 ± 1569 tons of waste in December and minimum 39682 ± 471 tons in February will be generated. It had also been estimated that the minimum waste generation from the year 2017 to 2023 will increase by approximately 5345 tons.