Full Research Paper – Page 4 – NPS Australia Submission System

DACA: A Distributed Algorithm for Task Partitioning and Offloading in Mobile Edge Computing Networks Supporting Transformer

Recent studies have explored collaborative Transformer-based inference in edge computing (EC), but they often overlook the mobility of users and edge devices, leading to potential reliability issues. This paper aims to minimize inference latency in mobile edge computing (MEC) by considering heterogeneity in mobility, computation, and communication. We propose a task partitioning model utilizing the GPipe scheme for Transformer-based inference. The task partitioning and offloading problem is then formulated with constraints on computation resources and mobility, decomposed into a bin-packing problem and an integer optimization problem. To solve these subproblems, we introduce the Distributed Aggregated Competition Algorithm (DACA). Extensive simulations and testbed experiments demonstrate the high performance of our proposed algorithm in minimizing inference latency across heterogeneous mobile edge devices and networks.

Predictive Analytics for Proactive Email Security Risk Management: A Systematic Review

In recent years, email has become an important communication tool for sharing private messages to crucial business message exchanges. However, its widespread use makes it a major target for cyber-attacks, including phishing, spam, and malware. These growing threats highlight the urgent need to investigate email security risk management to protect against attacks and maintain the integrity of communication systems. The study reviews the literature on the challenges of email security, risk management, and the role of predictive analysis in combating these threats. Using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), thirty-five (35) relevant peer-reviewed research articles were identified in various open research databases. This systematic literature review (SLR) also includes relevant case studies. The findings reveal that the integration of machine learning (ML), natural language processing (NLP), and real-time data analytics into email security frameworks improves threat detection and mitigation. Furthermore, these models often lack adaptability across languages and cultures. Additionally, they do not integrate well with human-centric security measures. Therefore, it is important to develop culturally adaptive predictive models, sector-specific solutions for industries such as finance and healthcare and incorporate behavioural analytics to enhance email threat detection and prevention. In other words, a comprehensive approach that combines technical advances with behavioural insights is crucial to strengthening email security and maintaining the integrity of global digital communications amid evolving cyber threats.

Sustainable Energy for Port Construction with Low Carbon Concrete from Industrial Symbiosis at WESTPORT Kwinana & BANTAENG Sulawesi

At Bantaeng in South Sulawesi and Kwinana in Western Australia new industrial scale ports will be built to serve the industrial precincts at these locations. At both these sites a 1-2Mtpa GPC plant is proposed for precast production of some 1,600 port modules as well as other infrastructure requiring some 750,000 cum of concrete and thereafter the plant can be repurposed for other products for local markets such as reef modules and wall panels. Geopolymer concrete can be the replacement for conventional concrete and be made from waste-derived materials while having a lower carbon footprint. The plant is designed to be operated by renewable energy and an energy audit estimated that a 1Mtpa geopolymer production plant needs up to 200 GWh pa to operate. This could be served by 6-10 on-land wind turbines combined with solar PV farm at a total cost $45-55 million USD. The electricity generated @ say $100/MWh was worth $12-20M pa that could result in a payback of 2-5 years. In Kwinana, planning is already underway for a large wind farm as part of the overall decarbonisation strategy for this industrial precinct. Feedstock materials can be harnessed for use in the geopolymer production plant by means of Circularity Hubs. These hubs can be established through the KIC4 and 6-Capitals models of Industrial Symbiosis and to optimize the proposed geopolymer plant within the industrial precincts at Bantaeng and Kwinana. Such an approach can contribute to Regenerative Development when both of the ports are built.

Low Carbon Concrete for Solid Gravity Energy Storage System and a Sustainable Electricity Grid.

Solid Gravity Energy Storage (SGES) Systems are an innovative way to store energy by using the force of gravity. These systems can use the excess energy from solar photovoltaic power systems to lift large blocks of concrete usually around mid-day and later as the sun sets and power demand is high, the blocks are released and generate gravitational energy which is converted to electricity. Colliecrete is a low-carbon, waste-derived, geopolymer concrete developed in 2021, from the Collie power plants’ flyash, by the Mudlark geopolymer lab at Murdoch University and geopolymer precursors can come from a number of waste-derived materials. Colliecrete can be used in the blocks for SGES. In Australia, most coal power plants will shut by 2030, while in Indonesia, the expectation is to achieve carbon-neutrality by 2060. There are many methods and pathways to achieve this goal with low-carbon geopolymer concrete one of them. Geopolymer precursor material is abundant with flyash available from 200 coal-fired power stations and slag from dozens of steel mills and nickel smelters. Rice husk is disposed of in millions of tonnes by farmers across the archipelago by burning and this ash can be converted to the geopolymer activator. All these make the possibility of an enormous new geopolymer concrete industry to at least partially replace the high-carbon, Portland cement industry. Geopolymer concrete blocks in the SGES system provide long-duration energy storage, assist firming the renewables and reduce carbon emissions while creating a new industry for the energy transition.

The Advent of Metal Additive Manufacturing Technologies – Alternative Options for Small Medium Enterprises

This study aims to explore the latest development made in Metal Additive
Manufacturing Technologies including alternative options which can be readily implemented by Small Medium Enterprises. This research paper also provides a brief overview on the consideration of fused filament fabrication technique based on material extrusion for metal 3D-Printing. Other indispensable processes in this technique, namely debinding and sintering processes are further discussed.

Stacking LLM Models’ Predictions for Feature Selection in Anomaly Classification

Large language models (LLMs) are increasingly being integrated into machine learning (ML) pipelines, particularly for tasks like feature selection in supervised classification. With the growing diversity of available LLMs, their predictions often complement one another, making ensembles of LLMs a promising approach for solving various ML challenges. In this paper, we propose using stacking methods to combine the predictions of multiple LLMs. The focus of the ML task is anomaly detection, specifically identifying whether an anomaly has occurred in a system and classifying its type. The ensemble’s base models are built on feature sets selected by six different LLMs. We demonstrate that stacking LLM predictions can enhance the accuracy of individual classifiers and advocate for the use of stacking as a simple yet effective method for integrating traditional classifiers with LLMs. Additionally, we assess the impact of various base classifiers and meta-classifiers on the performance of the proposed approach.

Energy trading in a decentralized blockchain based energy network

Application of new technology (Blockchain) in peer-to-peer energy sharing in smart grids.

Machine Learning-based Active Power Loss Forecasting in Distribution Systems

This paper introduces an optimal model that utilizes machine learning algorithms to predict the
active power losses during the allocation and sizing of distributed generation (DG) units in power distribution networks. The model incorporates the technique of Gradient Boosting Machine Regression (GBMR). This study estimates DG location, bus voltages, DG size, and active
losses without conventional power flow calculations.
The results demonstrate that the suggested estimations of power losses and DG sizing method are effective, practical, and adaptable. The accuracy of the proposed estimation methods has been validated using R-squared and mean absolute percentage error (MAPE) metrics. In the case of fixed load, the GBMR outperforms with a very low (MAPE) (0.9281%), a root mean square error (RMSE) of 1.748, and 0.999 accuracy in predicting
active power losses. When the normalized load variation (NLV), the R-squared 0.9991 value with low MAPE (1.9815%). This approach enables grid operators to effectively manage DG unit integration by providing precise estimates and forecasts of power loss. The effectiveness of the proposed strategy is validated in the IEEE 33 bus test system using MATLAB software.

Improved Energy Management for Hybrid Systems via Dual Predator Optimization

A novel hybrid renewable energy microgrid optimization algorithm supported with wind, solar, and backup diesel generators is suggested in this research work. The proposed Dual Predator Optimization (DPO) algorithm combines the Whale Optimization Algorithm (WOA), and Grey Wolf Optimizer (GWO). This algorithm integrates with a hybrid microgrid to optimize the use of renewable resources, reduce reliance on fossil fuel, and increase the cost-effectiveness by adjusting these parameters over time. The DPO is flexible and more suitable for hybrid energy management taking into consideration the exploration (exploitation) of system-level energy behaviors simultaneously in large-scale problems. The results show that the DPO is efficient in handling hybrid systems by significantly reducing electricity costs and decreasing probabilities of non-supply. It was determined that in comparison to the current GWO and WOA, the Cost of Energy (COE) of the proposed DPO algorithm is decreased to an average of 20%, while Loss of Power Supply Probability (LPSP) increases to an average of 7.5%.

Strengthening Fault Tolerance of Private/Consortium Blockchain with Trusted Execution Environment

Consensus is one of the key components of Blockchain. Common public blockchains use Proof-Of-Work or Proof-Of-Stake as their consensus protocols. In contrast, private or consortium blockchains often use Raft, which is only crash fault tolerant. It means that strong trust on node holders in private or consortium blockchains is required. To relax the strong trust requirement, we propose by taking raft as foundation and modification on raft and leveraging threshold signature and trusted execution environment to improve security. We have implemented and integrated our proposed consensus algorithm with ConsenSys Quorum. Our experiment shows that our work has slight performance degradation on blockchain compared to original Raft.

Paper Type: Full Research Paper