Cost Efficient GPU Cluster Management for Training and Inference of Deep Learning

Expanding the scale of GPU-based deep learning (DL) clusters would bring not only accelerated AI services but also significant energy consumption costs. In this paper, we propose a cost efficient deep learning job allocation (CE-DLA) approach minimizing the energy consumption cost for the DL cluster operation while guaranteeing the performance requirements of user requests. To do this, we first categorize the DL jobs into two classes: training jobs and inference jobs. Through the architecture-agnostic modeling, our CE-DLA approach is able to conduct the delicate mapping of heterogeneous DL jobs to GPU computing nodes. Second, we design the electricity price-aware DL job allocation so as to minimize the energy consumption cost of the cluster. We show that our approach efficiently avoids the peak-rate time slots of the GPU computing nodes by using the sophisticated mixed-integer nonlinear problem (MINLP) formulation. We additionally integrate the dynamic right-sizing (DRS) method with our CE-DLA approach, so as to minimize the energy consumption of idle nodes having no running job. In order to investigate the realistic behavior of our approach, we measure the actual output from the NVIDIA-based GPU devices with well-known deep neural network (DNN) models. Given the real trace data of the electricity price, we show that the CE-DLA approach outperforms the competitors in views of both the energy consumption cost and the performance for DL job processing.

Study on Cold Chain Transportation Model of Fruit and Vegetable Fresh-Keeping in Low-Temperature Cold Storage Environment

Due to the limitation of later stage intelligent algorithms, the fruit and vegetable fresh-keeping cold chain transportation scheme did not meet the expectation and could not achieve the dual objectives of the shortest time and the lowest consumption at the same time. In order to solve the above problems, a cold chain transportation model of fruit and vegetable fresh-keeping in a low-temperature cold storage environment is proposed. The model is based on the topology of the cold chain transportation network. By setting the assumptions of the fruit and vegetable fresh-keeping cold chain transportation model, the objective model is composed of three parts: vehicle power fuel consumption cost, cold chain transportation refrigeration cost, and total fruit and vegetable loss cost. Under six constraints, the improved ant colony algorithm is used to find the optimal fruit and vegetable fresh-keeping cold chain transportation route. The experimental results show that compared with the methods based on ALNS, genetic algorithm, and quantum bacterial foraging optimization algorithm, the research method can bring the best comprehensive benefit by accomplishing the fruit and vegetable transportation task in the shortest time at the lowest cost, and the research goal is thus achieved.

Behaviour of Self-Compacting Concrete incorporating Natural Perlite used as Part of Cement and as Aggregates

Perlite, a natural glassy volcanic rock could be used as supplementary cementitious material to reduce environmental pollution and the consumption of precious natural resources in the concrete industries. The aim of this work is to assess natural perlite used as 50% aggregates substitution by volume (sand or gravel) and as 10%, 15%, 20% cement substitution in self-compacting concrete. Workability characteristics and mechanical properties were analysed. Results showed that replacing 50% of natural aggregates with 50% of perlite aggregates or substituting cement with 10% of perlite powder generated the best workability characteristics and improved compressive, flexural strength, and elastic modulus of concrete at 28 days. Moreover, the results were combined to develop correlations that prove to be good between mechanical properties of self-compacting. Using perlite as aggregates offers a new source of supply and saves natural aggregates. Also, perlite used as cement substitution helps to reduce PC consumption, cost, and CO2 emission.

Hierarchical dispatching model of power grid under electrochemical energy storage constraints

Aiming at the hierarchical and zoning operation control of active distribution network, focusing on electrochemical energy storage, theoretical analysis and simulation research are carried out. Firstly, the basic principle of electrochemical energy storage is analyzed, and the working characteristics such as state of charge, power loss and charge discharge power are studied. Secondly, from the perspective of economic indicators, the power consumption cost of electrochemical energy storage is studied, and the objective function of global optimal dispatching of active distribution network is constructed. Then, the energy constraints and power constraints of electrochemical energy storage are studied as an important supplement to the active distribution network optimal scheduling model. Thirdly, the application of electrochemical energy storage in regional autonomous control is analyzed, and the regulation strategy of power fluctuation is studied from the perspective of standby capacity and power deviation. Finally, combined with a specific application example, the simulation analysis is given. The simulation results show that electrochemical energy storage can effectively optimize the hierarchical scheduling of active distribution network.

Optimization of Electric Vehicles Based on Frank-Copula-GlueCVaR Combined Wind and Photovoltaic Output Scheduling Research

Improving the efficiency of renewable energy and electricity utilization is an urgent problem for China under the objectives of carbon peaking and carbon neutralization. This paper proposes an optimization scheduling method of electric vehicles (EV) combined with wind and photovoltaic power based on the Frank-Copula-GlueCVaR. First, a joint output model based on copula theory was built to describe the correlation between wind and photovoltaic power output. Second, the Frank-Copula-GlueCVaR index was introduced in a novel way. Operators can now predetermine the future wind–photovoltaic joint output range based on this index and according to their risk preferences. Third, an optimal scheduling model aimed at reducing the group charging cost of EVs was proposed, thereby encouraging EV owners to participate in the demand response. Fourth, this paper: proposes the application of a Variant Roth–Serve algorithm; regards the EV group as a multi-intelligent group; and finds the Pareto optimal strategy of the EV group through continuous learning. Finally, case study results are shown to effectively absorb more renewable energy, reduce the consumption cost of the EV group, and suppress the load fluctuation of the whole EV group, which has a practical significance and theoretical value.

Green Inventory Routing Problem using Hybrid Genetic Algorithm

Carbon dioxide (CO2) is known as one of the largest sources of global warming. One of the ways to curb CO2 emissions is by considering the environmental aspect in the supply chain management. This paper analyses the influence of carbon emissions on the Inventory Routing Problem (IRP). The IRP network consists of a depot, an assembly plant and multiple suppliers. The deterministic demands vary and are determined by the assembly plant. Fixed transportation cost, fuel consumption cost and inventory holding cost are used to evaluate the system’s total cost in which fuel consumption cost is determined by fuel consumption rate, distance, and fuel price. Backordering and split pick-up are not allowed. The main purpose of this study is to analyze the distribution network especially the overall costs of the supply chain by considering the CO2 emissions as well. The problem is known as Green Inventory Routing Problem (GIRP). The mixed-integer linear programming of this problem is adopted from Cheng et al. wherein this study a different Hybrid Genetic Algorithm is proposed at mutation operator. As predicted, GIRP has a higher total cost as it considered fuel consumption cost together with the transportation and inventory costs. The results showed the algorithm led to different sequences of routings considering the carbon dioxide emission in the objective function.

KEY APPROACHES OF LIFE-CYCLE COST IN GREEN GOVERNMENT PROCUREMENT (GGP) FOR GREEN PROJECTS

Sustainability has emerged as a critical concern in any viable physical planning and development. Hence, the Malaysian government has promoted the concept of green procurement also known as Government Green Procurement (GGP) that aims to minimize environmental degradation. In GGP, life cycle perspective thinking is introduced where life-cycle cost (LCC) tools act as decision-making in controlling the initial and future value of building ownership. Despite the increasing importance of green procurement and LCC in the planning phase of green projects, the viability and implementation of LCC is still lacking. Many have stated the benefits of LCC in green procurement for green building projects, however the criteria of LCC are not clearly determined. The study aims to determine the important level of LCC components relating to the green project planning phase. Questionnaire survey was distributed to 50 respondents composed of project team members that were involved in the selected green government projects. 32 respondents returned their responses to the survey. The results revealed that the highest rank of LCC components in green procurement is energy consumption cost, greenhouse gas (GHG) savings cost, acquisition cost, energy simulation cost and utilities cost. These results would elevate the use of LCC in the green procurement adoption and viability of green projects.