Timothy M. Hansen

Heuristic Optimization for an Aggregator-Based Resource Allocation in the Smart Grid

We utilize a for-profit aggregator-based residential demand response (DR) approach to the smart grid resource allocation problem. The aggregator entity, using a given set of schedulable residential customer assets (e.g., smart appliances), must set a schedule to optimize for a given objective. Here, we consider optimizing for the profit of the aggregator. To encourage customer participation in the residential DR program, a new pricing structure named customer incentive pricing (CIP) is proposed. The aggregator profit is optimized using a proposed heuristic framework, implemented in the form of a genetic algorithm, that must determine a schedule of customer assets and the CIP. To validate our heuristic framework, we simulate the optimization of a large-scale system consisting of 5555 residential customer households and 56 642 schedulable assets using real-pricing data over a period of 24-h. We show that by optimizing purely for economic reasons, the aggregator can enact a beneficial change on the load profile of the overall power system.

A multi-agent model and strategy for residential demand response coordination

This paper proposes a multi-agent model and strategy for aggregator-based residential demand response, and details how elements in the system interact to solve an issue requiring load to be temporarily decreased. The system uses assets such as plug-in hybrid electric vehicles, air conditioning units, and electric water heaters to achieve this goal. Simulation results, based on probabilistic models and run on bus 5 of the RBTS test system, show that the system is capable of meeting the design objectives by shifting or shedding load so that the aggregate load remains under a given threshold. Results at the customer level also show that the impact on the comfort of customers is limited.

Experiences integrating transmission and distribution simulations for DERs with the Integrated Grid Modeling System (IGMS)

This paper discusses the development of, approaches for, experiences with, and some results from a large-scale, high-performance-computer-based (HPC-based) co-simulation of electric power transmission and distribution systems using the Integrated Grid Modeling System (IGMS). IGMS was developed at the National Renewable Energy Laboratory (NREL) as a novel Independent System Operator (ISO)-to-appliance scale electric power system modeling platform that combines off-the-shelf tools to simultaneously model 100s to 1000s of distribution systems in co-simulation with detailed ISO markets, transmission power flows, and AGC-level reserve deployment. Lessons learned from the co-simulation architecture development are shared, along with a case study that explores the reactive power impacts of PV inverter voltage support on the bulk power system.

A Partially Observable Markov Decision Process Approach to Residential Home Energy Management

Real-time pricing (RTP) is a utility-offered dynamic pricing program to incentivize customers to make changes in their energy usage. A home energy management system (HEMS) automates the energy usage in a smart home in response to utility pricing signals. We present three new HEMS techniques—one myopic approach and two non-myopic partially observable Markov decision process (POMDP) approaches—for minimizing the household electricity bill in such a RTP market. In a simulation study, we compare the performance of the new HEMS methods with a mathematical lower bound and the status quo. We show that the non-myopic POMDP approach can provide a 10%–30% saving over the status quo.

Spatial-temporal stochasticity of electric vehicles in an integrated traffic and power system

Penetration of a large number of electric vehicles in the grid can have a negative impact to the grid. To prevent a negative effect to the grid, the behavior of electric vehicles must be accurately modeled and their charging schedules must be coordinated. Therefore, it is necessary to determine where and how much charge is available in electric vehicles in the distribution system. In this paper, a state transition algorithm is designed to determine a stochastic model of electric vehicles to simulate electric vehicle movement in an integrated traffic and power network. Dijkstras algorithm is used to determine the shortest distance between end-user residential and office areas. A naïve charging technique is used to charge electric vehicles at different time intervals at different charging stations based on their driving patterns. The probability of finding electric vehicles at different charging stations and available charge using information on driving behavior is determined.

A Combined Dual-stage Framework for Robust Scheduling of Scientific Applications in Heterogeneous Environments with Uncertain Availability

Scheduling parallel applications on existing or emerging computing platforms is challenging, and, among other attributes, must be efficient and robust. A dual-stage framework is proposed in this paper to evaluate the robustness of efficient resource allocation and dynamic load balancing of scientific applications in heterogeneous computing environments with uncertain availability. The first stage employs robust resource allocation heuristics, while the second stage incorporates robust dynamic loop scheduling techniques. The combined dual-stage framework constitutes a comprehensive framework that enables and provides guarantees for the robust execution of scientific applications in computing systems where uncertainty is caused by various unpredictable perturbations. The paper reports on studies for determining the best techniques to be used for each stage that: (a) maximize the probability that the system make span satisfies a deadline, and (b) minimize the system make span for every given availability level in the system. The usefulness and benefits of the proposed framework are demonstrated via a small scale example.

Economic analysis of a data center virtual power plant participating in demand response

Data centers represent a large load for the grid, and the number of data centers are increasing at a high rate. The transmission network in the grid is becoming congested as a result of the load growth. Data centers have underutilized energy resources, such as backup generators and battery storage, which can be used for demand response (DR) to benefit both the electric power system and the data center. Therefore, in this paper, data center energy resources, including renewable energy, are aggregated and controlled using an energy management system (EMS) to operate as a virtual power plant (VPP). The data center as a VPP participates in a day-ahead DR program to relieve network congestion and improve market efficiency. A case study is conducted in which the data center is connected to bus 8 of the modified IEEE 30-bus test system to evaluate the potential economic savings by participating in the DR program, coordinated by the Independent System Operator (ISO). We show that the savings of the data center operating as a VPP and participating in the DR program far outweighs the expense due to operating its own generators.

Energy storage system operation: Case studies in deterministic and stochastic environments

As more renewable energy sources (solar, wind, hydro, etc.) are being incorporated into the smart grid, the problem of balancing generation and load demand gets more attention in the field of power system optimal economic operation. As a solution, grid-level energy storage systems (ESSs) have been used to increase the efficiency of the power supply by smoothing load fluctuations. In this paper, monotone adaptive dynamic programming (MADP) is investigated for the optimal operation of ESSs. MADP is capable of solving optimal policies for time-dependent and finite-horizon energy storage problems. Six types of real-world ESSs, with different storage capacities, charging/discharging rates, and efficiencies are compared under energy storage optimal benchmark problems. Simulation results discuss the impact of ESS parameters on the total revenue. The robust performance of the algorithm is also validated for both deterministic and stochastic energy storage benchmark problems.

Final Technical Report: Integrated Distribution-Transmission Analysis for Very High Penetration Solar PV

Transmission and distribution simulations have historically been conducted separately, echoing their division in grid operations and planning while avoiding inherent computational challenges. Today, however, rapid growth in distributed energy resources (DERs)--including distributed generation from solar photovoltaics (DGPV)--requires understanding the unprecedented interactions between distribution and transmission. To capture these interactions, especially for high-penetration DGPV scenarios, this research project developed a first-of-its-kind, high performance computer (HPC) based, integrated transmission-distribution tool, the Integrated Grid Modeling System (IGMS). The tool was then used in initial explorations of system-wide operational interactions of high-penetration DGPV.

Reliability analysis of 380V DC distribution in data centers

Data center downtime causes losses of millions of dollars. Maintaining high availability at all times is very critical to data centers. The distribution system with higher number of components is more likely to fail, resulting in increased downtime. This study presents a 380V DC powering option for data center distribution system, as the number of power conversion stages is less compared to the AC system. This study aims at comparing reliabilities of typical 480V AC distribution architecture against 380V DC architecture. Reliability assessment was done for both AC and DC sytems complying with Tier IV standard, as most of the data center uses Tier IV standard in their distribution system. The analysis was done for different level of redundancy (for eg. N, N+1, N+2) in the UPS system for both AC and DC systems. The reliability data was obtained from IEEE 493 Gold Book. Monte carlo simulation method was used to perform the reliability calculations. The simulation results showed that the 380V DC distribution system has higher level of reliability than conventional 480V ac distribution system in data centers but only up to certain level of redundancy in the UPS system. The reliability level of AC system will approach the reliability level of a DC system when a very high level of redundancy in the UPS system is considered, but this will increase the overall cost of the data center.