Improving the performance of data-driven techniques through data pre-processing for modelling daily reservoir inflow

Abstract

The effect of data pre-processing while developing artificial intelligence (AI) -based data-driven techniques, such as artificial neural networks (ANN), model trees (MT) and linear genetic programming (LGP), is studied for Pawana Reservoir in Maharashtra, India. The daily one-step-ahead inflow forecasts are compared with flows generated from a univariate autoregressive integrated moving average (ARIMA) model. For the full-year data series, a large error is found mainly due to the occurrence of zero values, since the reservoir is located in an intermittent river. Hence, all the techniques are evaluated using two data series: 18 years of daily full-year inflow data (from 1 January to 31 December); and 18 years of daily monsoon season inflow data (from 1 June to 31 October) to take into account the intermittent nature of the data. The relevant range of inputs for each category is selected based on autocorrelation and partial autocorrelation analyses of the inflow series. Conventional preprocessing methods, such as transformation and/or normalization of data, do not perform well because of the large variation in magnitudes, as well as the many zero values (65% of the full-year data series). Therefore, the input data are pre-processed into un-weighted moving average (MA) series of 3 days, 5 days and 7 days. The 3-day MA series performs better, maintaining the peak inflow pattern as in the actual data series, while the coarser-scale (5-day and 7-day) MA series reduce the peak inflow pattern, leading to more errors in peak inflow prediction. The results indicate that AI methods are powerful tools for modelling the daily flow time series with appropriate data preprocessing, in spite of the presence of many zero values. The time-lagged recurrent network (TLRN) ANN modelling technique applied in this study maps the inflow forecasting in a better way than the standard multilayer perceptron (MLP) neural networks, especially in the case of the seasonal data series. The MT technique performs equally well for low and medium inflows, but fails to predict the peak inflows. However, LGP outperforms the other AI models, and also the ARIMA model, for all inflow magnitudes. In the LGP model, the daily full-year data series with more zero inflow values performs better than the daily seasonal models.Key words data pre-processing; time-lagged recurrent network; model tree; linear genetic programming; moving average; transformation; full-year inflow data; seasonal inflow; Pawana Reservoir, India Amélioration de la performance de techniques conditionnées par les données par pré-traitement pour la modélisation de l'apport journalier d'un réservoir Résumé L'effet du pré-traitement des données à travers le développement des techniques d'intelligence artificielle (IA) conditionnées par les données, telles que les réseaux de neurones artificiels (RNA), les arbres modèles (MT) et la programmation génétique linéaire (PGL), est étudié pour le Réservoir Pawana dans le Maharashtra, Inde. Les prévisions à un jour des apports journaliers so...

Forecasting Reference Evapotranspiration Using Time Lagged Recurrent Neural Network

“…Then the output error is computed, and the error is backpropagated across the network. An analytical discussion of BPTT is available in Werbos [27], Jothiprakash and Kote [28].…”

Modeling Reservoir Water Release Decision Using Adaptive Neuro Fuzzy Inference System

“…One of the well-known computational intelligence techniques used for modelling reservoir water release decision and forecasting is the Artificial Neural Network (ANN) (Nazri et al, 2013;Mokhtar et al, 2014;Wan Ishak et al, 2015). However, this technique suffers from poor interpretability, since it is difficult for humans to explain the practicality and logical meaning behind the learned weights of the model (Jothiprakash & Kote, 2011;Kajornrit et al, 2013). This problem can be solved by the Adaptive Neuro Fuzzy Inference System (ANFIS).…”

Training of Artificial Neural Networks Using Information-Rich Data

“…They found that annual runoff forecast was improved using the particle swarm optimizing method based on empirical model decomposition. Jothiprakash and Kote [23] used data pre-processing for modelling daily reservoir inflow using a data-driven technique, and they found that intermittent inflow during the monsoon period alone could be modelled well, using the full year data, but model prediction accuracy increases when only the seasonal data set for the monsoon period is used. This simply implies that the information content in a series plays a big role in the model training, hence the prediction accuracy.…”

“…One can eliminate irrelevant data to produce faster learning, due to smaller data sets and due to the reduction of confusion caused by irrelevant data [25]. Most of the conventional pre-processing techniques, such as transformation and/or normalization of data, do not perform well, because of the large variation in magnitude and scale, as well as the presence of many zero values in data series [23]. Data from the real world are never perfect; it can be an incomplete record with missing information, occurrence of zeros, improper types, erroneous records, etc.…”