4 Presentations

• 
  01:30 PM - 01:55 PM

  Optimal design of decentralized adiabatic compressed air energy storage in urban building infrastructure for grid-peak shaving based on time of use pricing: a case study in Montreal-Canada

  • Elaheh Bazdar, presenter, Concordia University, Department of Building, Civil and Environmental Engineering Montreal, Quebec, Canada
  • Fuzhan Nasiri, Concordia University, Department of Building, Civil and Environmental Engineering Montreal, Quebec, Canada
  • Fariborz Haghighat, Concordia University, Department of Building, Civil and Environmental Engineering Montreal, Quebec, Canada

  In this study, a grid-connected photovoltaic (PV)/ adiabatic-compressed air energy storage system (A_CAES) is designed optimally to minimize the cost of energy subjected to the highest possible electrical load demand management. The optimal sizing is conducted within the simulation-optimization framework, including energy management operation strategy (EMOS), mathematical model as well as optimization model, and algorithm. The adopted EMOS plans the A-CAES system depending on the solar energy availability and the grid’s time-of-use electricity tariff. In such a way, A-CAES absorbs energy at the lowest electricity rates (e.g., during the off-peak hour or free solar energy) to satisfy the building load demand mainly during the expensive time of the day to avoid peak prices. Besides, a mixed integer nonlinear (MINL) optimization problem, including each component’s cost model as a function of the design parameters, is formulated, and solved using heuristic methods. Genetic algorithm (GA), particle swarm optimization (PSO) methods, and quantum particle swarm algorithm (QPSO) are applied and compared to solve the designing optimization problem. The outcomes verify that the performance of the PSO algorithm is better than GA and QPSO in terms of convergency speed and fitness function value. Hence, according to the case study specification, the optimized hybrid system configuration with a COE value of 0.071 $/kWh is achieved for satisfying the maximum 55.5% of the building load demand.
• 
  01:55 PM - 02:20 PM

  Optimization of clustering in wireless sensor network in terms of energy efficiency

  • amirmasoud soltanzadeh, presenter, P.hD. student

  Wireless sensor networks (WSNs) consist of an enormous number of tiny sensor nodes deployed in huge numbers which are able to sense, process and transmit environmental information between source node and destination node for a variety of applications.In wireless sensor networks, routing is a major challenge due to insufficient power supply.In addition, low-transmission bandwidth requires less memory space and handling limitations.These sensors are distributed randomly in nature and the environment. Sensor nodes gather data from the environment for further analysis before transmitting the information and data to the base station.Furthermore, energy efficiency is one of the primary concerns for maintaining WSN in operation.

  Wireless sensor networks (WSNs) are typically characterized by low-power, battery-operated devices that exist without fixed infrastructure, and whose purpose is to gather and send sensor data to a fixed location.It is crucial to study energy-saving and efficient communicationprotocolsl for WSN.To prolong the lifespan of a wireless sensor network and improve network throughput.It employs a hybrid clustering algorithm based on k means and DBSCAN, after defining cluster and cluster head we present a routing algorithm between cluster heads that provides a low end-to-end packet delay and minimal packet loss.
• 
  02:20 PM - 02:45 PM

  Rent-to-Own Contracts in Developing Economies

  • Elaheh Rashidinejad, presenter, Rotman School of Management, University of Toronto
  • Gonzalo Romero, Rotman School of Management, University of Toronto
  • Jose Guajardo, Haas School of Business, University of California Berkeley
  • Hosain Zaman, Data Scientist

  Access to electricity is a severe problem in many parts of the developing world. Several companies sell off-grid energy products using Rent-to-Own models, in which consumers make incremental payments over time to eventually become the owners of the product. Rent-to-Own models often imply giving degrees of payment flexibility to customers, which may increase the risk for companies given customers' income uncertainty. We study consumer’s payment behaviour under Rent-to-Own business models in developing economies where income uncertainty and hassle costs exists. We use a stochastic dynamic programming model to investigate how liquidity constraints and hassle costs affect customers’ payment behaviour. We show that customers may bundle their payment even without the presence of hassle costs. We further show that bundling payments happen closer to the ownership of the product. We examine different contract designs that firms selling off-grid energy products offer to their customers to minimize the expected time to ownership and improve social welfare, thereby leading to economic growth in developing areas.
• 
  02:45 PM - 03:10 PM

  Machine learning-aided mixed-integer quadratic programming for unit commitment ?

  • Philippe Maisonneuve, presenter, Department of Electrical Engineering, Polytechnique Montréal, Mila & GERAD
  • Antoine Lesage-Landry, Polytechnique Montréal

  The branch and bound (B&B) method is a cornerstone of integer programming. Leveraging highly effective heuristics for optimizing and pruning search trees, B&B-based solvers can address problems within limited timeframes. However, these heuristics are often generic and problem agnostic, leading to inefficiencies when applied to specific problem domains. In engineering contexts, similar problems are solved repetitively, such as the unit commitment problem in power engineering. By exploiting the problem's latent structure, we aim to significantly reduce computing time and enhance solver performance. Existing literature focuses primarily on linear problems which fall short when considering the quadratic power flow constraint at the core of power system decision-making. In this work, we propose an improved mixed-integer quadratic programming (MIQP) solver by integrating domain-specific knowledge to B&B algorithms through graph convolutional neural networks (GCNNs). To evaluate our approach, we conduct a series of numerical tests on an MIQP formulation of the unit commitment problem. We consider several electric network sizes to demonstrate the versatility and effectiveness of our method.