Marissa Hummon

This study performs a series of simulations of the grid in the western United States to determine how a CSP plant with TES might be dispatched to maximize its value when replacing conventional fossil fuel plants. The value of CSP with TES is compared to renewable generators without storage including PV.

India has ambitious goals for high utilization of variable renewable power from wind and solar, and deployment has been proceeding at a rapid pace. The western state of Gujarat currently has the largest amount of solar generation of any Indian state, with over 855 Megawatts direct current (MWDC). Combined with over 3,240 MW of wind, variable generation renewables comprise nearly 18% of the electric-generating capacity in the state. A new historic 10-kilometer (km) gridded solar radiation data… Show more

India has ambitious goals for high utilization of variable renewable power from wind and solar, and deployment has been proceeding at a rapid pace. The western state of Gujarat currently has the largest amount of solar generation of any Indian state, with over 855 Megawatts direct current (MWDC). Combined with over 3,240 MW of wind, variable generation renewables comprise nearly 18% of the electric-generating capacity in the state. A new historic 10-kilometer (km) gridded solar radiation data set capturing hourly insolation values for 2002-2011 is available for India. We apply an established method for downscaling hourly irradiance data to one-minute irradiance data at potential PV power production locations for one year, 2006.
The objective of this report is to characterize the intra-hour variability of existing and planned photovoltaic solar power generation in the state of Gujarat (a total of 1.9 gigawatts direct current (GWDC)), and of five possible expansion scenarios of solar generation that reflect a range of geographic diversity (each scenario totals 500-1,000 MW of additional solar capacity). The report statistically analyzes one year’s worth of power variability data, applied to both the baseline and expansion scenarios, to evaluate diurnal and seasonal power fluctuations, different timescales of variability (e.g., from one to 15 minutes), the magnitude of variability (both total megawatts and relative to installed solar capacity), and the extent to which the variability can be anticipated in advance. The paper also examines how Gujarat Energy Transmission Corporation (GETCO) and the Gujarat State Load Dispatch Centre (SLDC) could make use of the solar variability profiles in grid operations and planning. Show less

Production cost models are often used for planning by simulating power system operations over long time horizons. The simulation of a day-ahead energy market for a one-year horizon can take several weeks to compute. Tractability improvements are often made through model simplifications, such as reductions in transmission modeling detail, relaxation of commitment variable integrality, and reductions in cost modeling detail. One common simplification is to partition the simulation horizon so that… Show more

Production cost models are often used for planning by simulating power system operations over long time horizons. The simulation of a day-ahead energy market for a one-year horizon can take several weeks to compute. Tractability improvements are often made through model simplifications, such as reductions in transmission modeling detail, relaxation of commitment variable integrality, and reductions in cost modeling detail. One common simplification is to partition the simulation horizon so that weekly or monthly horizons can be simulated in parallel. However, horizon partitions are often executed with overlap periods of arbitrary and sometimes zero length. We calculate the time domain persistence of historical unit commitment decisions to inform time domain partitioning of production cost models. The results are implemented using PLEXOS production cost modeling software in a high performance computing environment to improve the computation time of simulations while maintaining solution integrity. Our results show that including two days of overlap on partitioned problems maximizes solution quality. Additionally, we achieve roughly 25-times speedup when optimizing 52 weeks in parallel, with overlap, versus a single annual optimization. Show less

Demand response (DR) has the potential to improve electric grid reliability and reduce system operation costs. However, including DR in grid modeling can be difficult due to its variable and non-traditional response characteristics, compared to traditional generation. Therefore, efforts to value the participation of DR in procurement of grid services have been limited. In this report, we present methods and tools for predicting demand response availability profiles, representing their… Show more

Demand response (DR) has the potential to improve electric grid reliability and reduce system operation costs. However, including DR in grid modeling can be difficult due to its variable and non-traditional response characteristics, compared to traditional generation. Therefore, efforts to value the participation of DR in procurement of grid services have been limited. In this report, we present methods and tools for predicting demand response availability profiles, representing their capability to participate in capacity, energy, and ancillary services. With the addition of response characteristics mimicking those of generation, the resulting profiles will help in the valuation of the participation of demand response through production cost modeling, which informs infrastructure and investment planning. We present an approach for predicting demand response availability profiles for thirteen end-use loads within the Western Interconnection for the year 2020. For each end-use, we estimate an annual hourly load profile for each of the 36 balancing authority areas. These load profiles are further evaluated and filtered to obtain an estimate of the amount of load available to participate in each of five products (three ancillary services, an energy product, and a capacity product) for each hour of the 2020 calendar year. We supplement these DR availability profiles with expectations on constraints for the duration and frequency of the load responses, in order to best represent the expected electricity customer system, such as their willingness and capability (e.g., equipment) to participate in demand response events. Finally, we discuss the projected theoretical availability for full load participation. Show less

Demand response (DR) resources present a potentially important source of grid flexibility particularly on future systems with high penetrations of variable wind an solar power generation. However, DR in grid models is limited by data availability and modeling complexity. This presentation focuses on the co-optimization of DR resources to provide energy and ancillary services in a production cost model of the Colorado “test system”. We assume each DR resource can provide energy services by… Show more

Demand response (DR) resources present a potentially important source of grid flexibility particularly on future systems with high penetrations of variable wind an solar power generation. However, DR in grid models is limited by data availability and modeling complexity. This presentation focuses on the co-optimization of DR resources to provide energy and ancillary services in a production cost model of the Colorado “test system”. We assume each DR resource can provide energy services by either shedding load or shifting its use between different times, as well as operating reserves: frequency regulation, contingency reserve, and flexibility (or ramping) reserve. There are significant variations in the availabilities of different types of DR resources, which affect both the operational savings as well as the revenue for each DR resource. The results presented include the system-wide avoided fuel and generator start-up costs as well as the composite revenue for each DR resource by energy and operating reserves. Show less

Most calculations of optimum photovoltaic (PV) performance focus on maximizing annual energy production. However, given the seasonally and daily time varying electricity demand and resulting variation in price, the PV orientation resulting in maximum energy yield may not yield the maximum economic benefit. With the use of historical solar irradiance and wholesale market prices for several locations in the USA, we evaluate the benefits of a variety of orientations for fixed and tracking PV… Show more

Most calculations of optimum photovoltaic (PV) performance focus on maximizing annual energy production. However, given the seasonally and daily time varying electricity demand and resulting variation in price, the PV orientation resulting in maximum energy yield may not yield the maximum economic benefit. With the use of historical solar irradiance and wholesale market prices for several locations in the USA, we evaluate the benefits of a variety of orientations for fixed and tracking PV arrays. We find that orienting fixed arrays slightly to the west of due south generally increases their economic value in the simulated systems because the reduced generation on an annual basis is more than offset by increased generation in high-value hours in late summer afternoons. However, this effect is small, typically providing an increase in value from 1% to 5%. The economic value of adjusting the orientation semi-annually (May 1st and September 1st) and monthly shows a modest increase in value from 3% to 5%. Several other implications of this analysis are also discussed. Show less

High penetration renewable integration studies require solar power data with high spatial and temporal accuracy in order to understand the impact of high frequency solar power ramps on the operation of the system. Our previous work concentrated on down-scaling solar power simulation from one hour to one minute. This method used clearness classifications to categorize temporal and spatial variability, and iterative methods to simulate intra-hour clearness variability. Solar power ramp… Show more

High penetration renewable integration studies require solar power data with high spatial and temporal accuracy in order to understand the impact of high frequency solar power ramps on the operation of the system. Our previous work concentrated on down-scaling solar power simulation from one hour to one minute. This method used clearness classifications to categorize temporal and spatial variability, and iterative methods to simulate intra-hour clearness variability. Solar power ramp correlations between sites decrease with distance and the duration of the ramp, starting at around 0.6 for 30-minute ramps between sites that are less than 20 km apart. The sub-hour irradiance algorithm has a noise floor that causes the correlations to approach ~0.005. Below one minute, the majority of the correlations of solar power ramps between sites less than 20 km apart are zero, and thus a new method to simulate intra-minute variability is needed. Intra-minute solar power ramps can be simulated using several methods. We evaluate three methods: cubic spline fit to the one-minute solar power data; projection of the power spectral density toward higher frequency domain; and development of a “template” for high frequency power spectral density. Each method either under- or over-estimates the variability of intra-minute solar power ramps. We show that an optimized weighted linear sum of methods, dependent on the classification of temporal variability of the segment of one-minute solar power data, yields time series and ramp distributions similar to measured high-resolution solar irradiance data. This method has been applied to sites in California to model automatic generator control in response to high frequency solar power ramps. Show less

Concentrating solar power with thermal energy storage (CSP-TES) can provide multiple benefits to the grid, including low marginal cost energy and the ability to levelize load, provide operating reserves, and provide firm capacity. It is challenging to properly value the integration of CSP because of the complicated nature of this technology. Unlike completely dispatchable fossil sources, CSP is a limited energy resource, depending on the hourly and daily supply of solar energy. To optimize the… Show more

Concentrating solar power with thermal energy storage (CSP-TES) can provide multiple benefits to the grid, including low marginal cost energy and the ability to levelize load, provide operating reserves, and provide firm capacity. It is challenging to properly value the integration of CSP because of the complicated nature of this technology. Unlike completely dispatchable fossil sources, CSP is a limited energy resource, depending on the hourly and daily supply of solar energy. To optimize the use of this limited energy, CSP-TES must be implemented in a production cost model with multiple decision variables for the operation of the CSP-TES plant. We develop and implement a CSP-TES plant in a production cost model that accurately characterizes the three main components of the plant: solar field, storage tank, and power block. We show the effect of various modelling simplifications on the value of CSP, including: scheduled versus optimized dispatch from the storage tank and energy-only operation versus co-optimization with ancillary services. Show less

Operating reserves impose a cost on the electric power system by forcing system operators to keep partially loaded spinning generators available to respond to system contingencies and random variation in demand. In many regions of the United States, thermal and hydropower plants provide a large fraction of the operating reserve requirement. Alternative sources of operating reserves, such as demand response and energy storage, may provide these services at lower cost. However, to estimate the… Show more

Operating reserves impose a cost on the electric power system by forcing system operators to keep partially loaded spinning generators available to respond to system contingencies and random variation in demand. In many regions of the United States, thermal and hydropower plants provide a large fraction of the operating reserve requirement. Alternative sources of operating reserves, such as demand response and energy storage, may provide these services at lower cost. However, to estimate the potential value of these services, the cost of reserve services under various grid conditions must first be established.
The analysis demonstrates that the price of operating reserves depends greatly on many assumptions regarding the operational flexibility of the generation fleet, including ramp rates and the fraction of the fleet available to provide reserves. While variable renewables increase the total reserve requirements, the additional operational cost of these reserves appears modest in the evaluated system. Wind and solar generation tend to free up generation capacity in proportion to its production, largely canceling out the net cost of the additional operating reserves. Further work is needed to address issues, such as down reserves and implementation of fast-response regulation, which were not included in this study. Finally, this analysis points to the need to consider how the operation of the power system and composition of the conventional generation fleet may evolve if wind and solar power reach high penetration levels. Show less

High penetration renewable integration studies need high quality solar power data with spatial-temporal correlations that are representative of a real system. For instance, as additional solar power sites are added, the relative amount of variability should decrease due to spatial averaging
of localized irradiance fluctuations. This presentation will sum­marize the research relating sequential point-source sub-hour global horizontal irradiance (GHI) values to static, spatially distributed… Show more

High penetration renewable integration studies need high quality solar power data with spatial-temporal correlations that are representative of a real system. For instance, as additional solar power sites are added, the relative amount of variability should decrease due to spatial averaging
of localized irradiance fluctuations. This presentation will sum­marize the research relating sequential point-source sub-hour global horizontal irradiance (GHI) values to static, spatially distributed GHI values. This research led to the development of an algorithm for generating coherent sub-hour datasets that span distances ranging from 10 km to 4,000 km. The algorithm,
in brief, generates synthetic GHI values at an interval of one minute, for a specific location, using SUNY/Clean Power Research, satellite-derived, hourly irradiance values for the
nearest grid cell to that location and grid cells within 40 km. During each hour, the observed GHI value for the grid cell of interest and the surrounding grid cells is related, via probability distributions, to one of five temporal cloud coverage classifications (class I, II, III, IV, V). Synthesis algorithms are used to select one-minute time step GHI values based on the classification of the grid cell of interest in a particular hour. Three primary statistical measures of the dataset are
demonstrated: reduction in ramps as a function of aggregation; coherence of GHI values across sites ranging from 6 to 400 km apart over time scales from one minute to three hours; and
ramp magnitude and duration distributions as a function of time of day and day of year. Show less