Wai Kean Yap

Emissions predictive modelling by investigating various neural network models

Highlights? ANN model accurately predicts exhaust emissions with minimal inputs. ? ANN optimization-layer-by-layer model proves to be the most accurate. ? Models act as virtual emission sensor without the need of additional equipment. ? A single ECU is just needed to predict the engine parameters and emission. ? The generic two-stage ANN model is applicable to any ICE vehicle applications. This paper presents a two-stage emissions predictive model developed by investigating common feedforward neural network models. The first stage model involves predicting engine parameters power and tractive forces and the predicted parameters are used as inputs to the second stage model to predict the vehicle emissions. The following gasses were predicted from the tailpipe emissions for a scooter application; CO, CO2, HC and O2. Three feedforward neural network models were investigated and compared in this study; backpropagation, optimization layer-by-layer and radial basis function networks. Based on the experimental setup, the neural network models were trained and tested to accurately predict the effect of the engine operating conditions on the emissions by varying the number of hidden nodes. The selected optimization layer-by-layer network proved to be the most accurate and reliable predictive tool with prediction errors of ?5%. The effect of the engine operating conditions on the tailpipe emissions for a scooter is shown to display similar qualitative and quantitative trends between the simulated and the experimental data. This study provides a better understanding in effects of engine process parameters on tailpipe emissions for the scooter as well as for general vehicular applications.

Comparative Study in Predicting the Global Solar Radiation for Darwin, Australia

This paper presents a comparative study in predicting the monthly average solar radiation for Darwin, Australia (latitude 12.46 deg S longitude 130.84 deg E). The city of Darwin, Northern Territory (NT), has the highest and most consistent sunshine duration among all the other Australian states. This unique climate presents an opportunity for photovoltaic (PV) applications. Reliable and accurate predictions of solar radiation enable potential site locations, which exhibit high solar radiations and sunshine hours, to be identified for PV installation. Three predictive models were investigated in this study—the linear regression (LR), Angstrom–Prescott–Page (APP), and the artificial neural network (ANN) models. The mean global solar radiation coupled with the climate data (mean minimum and maximum temperatures, mean rainfall, mean evaporation, and sunshine fraction) obtained from the Australian Bureau of Meteorology (BoM) formed the basis of the dataset. Using simple and easily obtainable climate data presents an added advantage by reducing model complexity. Predictive results showed the root mean square errors (RMSEs) obtained were 6.72%, 13.29%, and 8.11% for the LR, APP, and ANN models, respectively. The predicted solar exposure from the LR model was then compared with the satellite-derived data to assess the accuracy of the LR method.

Neural network-based active power curtailment for overvoltage prevention in low voltage feeders

As non-controllable and intermittent power sources, grid-connected photovoltaic (PV) systems can contribute to overvoltage in low voltage (LV) distribution feeders during periods of high solar generation and low load where there exists a possibility of reverse power flow. Overvoltage is usually prevented by conservatively limiting the penetration level of PV, even if these critical periods rarely occur. This is the current policy implemented in the Northern Territory, Australia, where a modest system limit of 4.5kW/house was imposed. This paper presents an active power curtailment (APC) strategy utilizing artificial neural networks techniques. The inverter active power is optimized to prevent any overvoltage conditions on the LV feeder. A residential street located in Alice Springs was identified as a case study for this paper. Simulation results demonstrated that overvoltage conditions can be eliminated and made to comply with the Australian Standards AS60038 and AS4777 by incorporating the proposed predictive APC control. In addition, the inverter downtime due to overvoltage trips was eliminated to further reduce the total power losses in the system.

Performance modelling and simulation of a hybrid electric scooter

This paper presents a model developed on the Matlab/Simulink platform for the operation and optimisation of a scooters hybrid powertrain. The model, which is a parallel hybrid structure, is powered by a two-stroke Internal Combustion Engine (ICE) coupled with an electric hub motor to make up the scooters propulsion unit. The model was built and the simulation results are shown. The developed multi-mode Matlab/Simulink hybrid electric scooter model demonstrates that the petrol and energy costs are significantly reduced compared to traditional petrol and electric scooters. The costs were reduced per every 100 km compared to traditional petrol and electric scooters. Research showed that to achieve higher distances with lower petrol consumption, a hybrid scooter is more reliable. The qualitative trends showed that the multi-mode approach minimises petrol consumption by minimising the usage of the ICE. This was examined on the Standard European Driving Cycle (ECE-15) driving cycle. In this paper, it is argued that this multimode hybrid approach reduces cost compared with petrol scooters and increases the distance travelled compared with pure electric scooters.