Injected fuel flow forecasting with Online Sequential Extreme Learning Machine

Abstract

This study deals with Online Sequential Extreme Learning Machine (OS-ELM) modeling of a gasoline engine to predict the injected fuel flow of the engine. The single hidden layer feedforward networks (SLFN) trained by OS-ELM algorithm was selected as a black box model for forecasting purposes. The algorithm is used to train a SLFN using a set of data consists of the running gasoline engine features such as speed, revolution, fuel volume, current fuel consumption, gear, distance to empty in volume, distance to empty in kilometer, current distance, and battery voltage. A total of 700 data were used in forecasting process. The effectiveness of the method has been demonstrated through analysis of the performance error of the fitted network using a mean square error (MSE) expressed in decibel (dB), the best learning mode, optimum number of hidden nodes and forecasting time. Promising result of maximum speed of forecasting has been achieve with – 62.00 dB of mean square error and 1-by-1 learning mode for the OS-ELM using sinusoidal activation function.