10th International Conference on Computational Management
HEC Montréal, 1 — 3 May 2013
10th International Conference on Computational Management
HEC Montréal, 1 — 3 May 2013
WA2 Smart Grids
May 1, 2013 10:30 AM – 12:30 PM
Location: St-Hubert
Chaired by Miguel F. Anjos
4 Presentations
-
10:30 AM - 11:00 AM
Strategies for the Fast Identification of Umbrella Constraints in Security-Constrained Optimal Power Flow Problems
Security-constrained optimal power flow (SCOPF) problems are essential to the economic and reliable operation of large power systems. Typically, they attempt to optimize the cost of electricity generation subject to the balance of supply and demand and to sets of inequalities bounding power flows on transmission lines, both in pre- and post-outage states. Empirical evidence and industry practices indicate that very few of those line flow constraints are ever potentially binding.
In previous work, we developed a linear optimization approach to identify those constraints, which we also call "umbrella constraints". Umbrella constraints are those both necessary and sufficient to describe the feasible space of an SCOPF problem. We have shown that for SCOPF instances based on standard IEEE test networks, we can generally bring down the number of constraints in a SCOPF to under 1% of the original constraint count. In turn the solution times and memory requirements of SCOPF problems subjected to their umbrella constraint are significantly smaller. For example, an SCOPF for the IEEE 118 node test system (with 65,730 constraints originally), when stripped down to its 451 umbrella constraints, runs in 0.57% of the time needed to solve the SCOPF subjected to its original set of constraints.
The Umbrella Constraint Discovery (UCD) problem, in spite of being a linear program, is unfortunately larger than the original SCOPF, both in terms of variable and constraint numbers. In this talk, we discuss a number of strategies which exploit the structure of the SCOPF to speed up the solution of the UCD.
First, we propose a divide-and-conquer approach to the solution of UCD. This method generates a number of smaller UCD problems that can be solved in parallel, recombined and solved until the complete umbrella constraint set is identified. The success of this approach is closely related to the partitioning of the original set of SCOPF constraints. In that matter, we demonstrate how line-based partitioning is superior to outage-based partitioning.
Second, we discuss the sensitivity of UCD outcomes in terms of SCOPF parametrizations and show how those results are useful in warm starting the UCD solution process.
Finally, we present the partial UCD solution approach which further exploits the structure of the underlying SCOPF and leads to significant reductions in UCD run times (circa 50% lower than the best divide-and-conquer approach).
This work is of significance to improving the operation and planning of electricity grids. It should contribute to simplify the computational burden associated with planning preventive and corrective control resources in large power systems. Also, it could contribute to streamline the solution processes of large electricity market clearing algorithms now found in the US and Europe.
-
11:00 AM - 11:30 AM
Demand Side Load Management in the Smart Grid: a Smart Consumer Perspective
Energy efficiency is one of the central topics in the Smart Grid, which is concerned with a large range of technologies and systems. Although the reduction of line losses, improvement of active and reactive load, distribution optimization, and exploitation of renewable energy sources are essential to energy efficiency enhancement, load management also has a significant impact. One of the distinguished characteristics of the Smart Grid is that the two-way flow of both electricity and data enables the active collaboration of consumers and allows improving energy efficiency by consumption scheduling, load forecasting, peak shaving, etc. on the consumer side. This is likely one of the most efficient and cost-effective solutions leading to truly smart grids.
In this talk, we present an architecture for demand side load management in the Smart Grid. Being of a layered structure composed of three main modules for admission control, load balancing, and demand response management, this architecture encapsulates the main system functionality, assure the interoperability between various components, allow the integration of different energy sources, and ease maintenance and upgrading. Hence it is capable of handling autonomous energy consumption management for systems with heterogeneous dynamics in multiple time-scales and allows seamless integration of diverse techniques for online operation control, optimal scheduling, and dynamic pricing. We will present in particular solutions inspired by real-time computing system scheduling and optimal control techniques for admission control and illustrate some design issues via the control of HVAC (Heat Ventilation Air Conditioning) in a smart building.
-
11:30 AM - 12:00 PM
Value of Large-Scale Energy Storage in Québec
Advanced energy storage has been a key technology for the rapid expansion of portable electronics and the transportation industry. Moreover, electrical energy storage technologies, when properly designed and integrated into the electricity grid system, could serve several roles in the system. For example, energy storage systems could reduce fluctuation by load leveling and peak shaving; reduce the risks of power blackout by operating independently to supply emergence backup power; support intermittent renewable energy generation. Indeed, energy storage technologies from the large-scale pumped hydro systems to the small-scale battery systems are used worldwide. It is believed that combined with a smart grid, they will see a great leap forward in the next 30 years.
In this presentation we will focus our study on large-scale energy storage by batteries in Quebec. The study assesses quantitatively the possibility of investing in a new energy storage project, which aims to make the grid more efficient, stable and reliable. The increasing demand for electrical power has put into evidence the big difference of electric power demand between peak and off-peak periods, such as day and night, weekday and weekend, different seasons. If a large-scale energy storage system were available, it could provide value via energy arbitrage by filling the storage device during the off-peak periods when electricity is abundant and cheap, then discharging the energy stocked into the grid during peak periods when electricity is more valuable. This could not only match electricity supply to load demand, but also could reduce the need for peak-only power stations or standby plants. Furthermore, Hydro-Quebec’s policy of developing overcapacity enables us to take the opportunity to expand exports. A linear programming optimization model is used to simulate the operation of the energy storage system combined with the power system, which helps us to find the optimum size of the storage system while maximizing the benefit of the project.
-
12:00 PM - 12:30 PM
Optimization of Wind, Diesel and Battery Systems for Remote Areas
Imagine traveling to a mine in the territory of Nunavut (Canada), which is 1000 km away from any electrical grid. You will rapidly wonder where the mine’s energy comes from. The high extension cost of the grid leaves only one alternative: to produce the energy locally and autonomously. As mining increases in remote areas in wide countries such as India or China, the number of such isolated sites will rise. Experts predict an energy market of $10B for this type of site by 2020. Traditionally electricity was supplied by diesel generation. This technology is easy to implement but is also very expensive because of fuel price and transportation. Moreover it is in opposition to the objective of increased use of green energy. The use of wind turbines in remote areas in order to reduce fuel consumption by wind was proposed in 1990s. After some years of adjustment, this technology matured and is now widely used in Alaska off-grid sites. Batteries have recently improved greatly and are a solution against wind intermittency. Their use may further reduce the use of diesel generators. Thus a lot of models composed of wind, diesel and batteries have been developed to determine their potential and feasibility. These hybrid systems can work only because of an operation strategy. This strategy is established by deciders and states the rules of the system: when and how dispatching energy from the 3 elements. The challenge in hybrid systems optimisation is to find both optimal sizing and optimal strategy, which are linked and influence each other. Additionally the optimization is made through stochastic wind and load data. The novelty of our model is the use of integer linear programming in order to eliminate the need to set a strategy. Because we give what the wind and load would be for a whole year, the model finds the optimal sizing and optimal scenario for this year. It is not a strategy but what we would have done at each hour in the year to have the perfect dispatch. It represents the reference to test all strategies, and it is also very useful for deciders to know the maximum possibilities.