We present an algorithm that estimates a clear sky performance signal from the measured power of a PV system. The algorithm uses only observed power output, and assumes no knowledge of weather, irradiance data, or system configuration metadata. This is a novel approach to understanding the clear sky behavior of an installed PV system, that does not rely on traditional atmospheric and geometric modeling techniques.

Photovoltaic (PV) systems are increasing in distribution systems, but utilities lack visibility of the generation of this distributed PV. This paper presents a set of methods for disaggregating the photovoltaic (PV) generation downstream of a measurement device that measures net load using only readily available measurements. We propose two strategies in which we use measurements from the substation as well as a proxy solar irradiance measurement. Using these two measurement points, we first propose a multiple linear regression strategy. We estimate a relationship between the measured reactive power and the load active power consumption, which are then used in real-time disaggregation. Then, we expand this strategy to reconstruct the errors in the estimators, thus separating the solar and load signals from their aggregate. We show that it is possible to disaggregate the generation of a 7.5 megawatt photovoltaic site with a root-mean-squared error of ≈ 450 kilowatts.

Real-time photovoltaic (PV) generation information is crucial for distribution system operations such as switching, state-estimation, and voltage management. However, most behind-the-meter solar installations are not monitored. Typically, the only information available to the distribution system operator is the installed capacity of solar behind each meter; though in some cases even the presence of solar may be unknown. We present a method for disaggreagating behind-the-meter solar generation using only information that is already available in most distribution systems: advanced metering infrastructure, substation monitoring, and generation monitoring at a few PV systems nearby the circuit. The proposed method accurately predicts which homes have solar in over 90% of cases, and recovers the 15-min resolution PV generation signals with root mean square errors between 20% and 50% of average daily PV generation both historically and real-time. A sensitivity analysis shows the method to be robust to the number of buildings and time span of data used to fit. However including more than 3 solar proxies can cause false positive of PV systems behind meters. We find that the proposed method performs better at homes that export electricity to the grid more often.

The demand for vehicle charging will require large investments in power distribution, transmission, and generation. However, this demand is often also flexible in time, and can be actively managed to reduce the needed investments, and to better integrate renewable electricity. Harnessing this flexibility requires forecasting and controlling electric vehicle (EV) charging at thousands of stations. This paper addresses the problem of forecasting and management of the aggregate flexible demand from tens to thousands of EV supply equipment (EVSEs). First, it presents an equivalent time-variant storage model for flexible demand at an aggregation of EVSEs. The proposed model is generalizable to different markets, and also to different flexible loads. Model parameters representing multiple EVSEs can be easily aggregated by summation, and forecasted using autoregressive models. The forecastability of uncontrolled demand and storage parameters is evaluated using data from 1341 nonresidential EVSEs located in Northern California. The median coefficient of variation is as low as 24% for the forecast of uncontrolled demand at the highest aggregation and 10-15% for the storage parameters. The benefits of aggregation and forecastability are demonstrated using an energy arbitrage scenario. Purchasing energy day ahead is less expensive than in the real-time market, but relies on a uncertain forecast of charging availability. The results show that the forecastability significantly improves for larger aggregations. This helps the aggregator make a better forecast, and decreases the cost of charging in comparison to an uncontrolled case by 60% with respect to an oracle scenario.

The Internet of Things (IoT) is creating a major shift from the traditional approaches to operating and maintaining all energy systems. With the proliferation of cheap sensors, power grids are generating vast amounts of data and there is an increasing interest around data-driven decision making, or more broadly, energy analytics. Due to the inherent nature of device and data ownership in the power grids, these analytics efforts have been siloed. Energy analytics developers create databases for monolithic applications that lack the infrastructure to quickly transform data between computational resources and persistent data storage, easy dissemination and validation of results, and support integration. We propose a data architecture that supports broader use of data technologies and systems integration guided by publish and subscribe architecture, and polyglot persistence. We provide use cases to discuss our design choices using the electric power grid as the main application domain. However, we believe that the energy industry as a whole can benefit from the proposed architecture.

Thermal-sensitive electricity usage accounts for the majority of residential electricity use worldwide. Moreover, it has substantial impacts on the power system and the society during the extreme weathers. We propose a novel data-driven method to identify the thermal-sensitive electricity usage and quantify the associated thermal sensitivity directly from widely deployed smart meter data, where the Multivariate Adaptive Regression Splines model provides both the flexibility and interpretability. We apply the proposed method to analyze the thermal sensitivity of individual residential electricity use for two representative urban areas in the United States and China, San Francisco Bay Area and Shanghai, respectively. More than 120,000 households' smart meter data sampled from Bay Area and Shanghai is used for the analysis. To best of our knowledge, we for the first time reveal the heterogeneous patterns of thermal sensitivity of households from USA and China at the individual level. The difference of thermal sensitivity serves as an important factor for personalized techniques and policies towards mitigating the impact of extreme weathers to power systems and improving the buildings' energy efficiency.

Real-time PV generation information is crucial for distribution system operations, in particular for switching operations, state-estimation, and management of the voltage at the point of provision. However, most behind-the-meter solar installations are not monitored. Typically, the only information available to the distribution system operator is the installed capacity. Our main purpose in this study is to estimate behind-the-meter PV generation using data from smart meters.

This paper presents a set of methods for estimating the renewable energy generation downstream of a measurement device using real-world measurements. First, we present a generation disaggregation scheme where the only information available for estimation is the micro-synchrophasor measurements obtained at the substation or feeder head. We then propose two strategies in which we use measurements from the substation as well as a proxy solar irradiance measurement. Using these two measurement points, we first propose a multiple linear regression strategy, in which we estimate a relationship between the measured reactive power and the load active power consumption, which are then used in disaggregation. Finally, we expand this strategy to strategically manage the reconstruction errors in the estimators. We simultaneously disaggregate the solar generation and load. We show that it is possible to disaggragate the generation of a 7.5 megawatt photovoltaic site with a root-mean-squared error of approximately 450 kilowatts.

Locations of photovoltaic (PV) systems affect the variability and uncertainty of their power generation, and as a result the amount of flexible resources needed to balance supply and demand. However, studies on the integration of renewable electricity into power systems focus on the total amount of renewable generation, and not their locations. This paper uses a hidden state, spatial-statistical model to simulate how locational arrangements and balancing policies affect the need for reserves load following and regulation in California's power system when including 12 GW of photovoltaic generators and 9.5 GW of wind.Our results show that locations of utility-scale PV systems significantly affect on the amount of reserves needed for balancing their variability and uncertainty. When PV is geographically dispersed the additional load following and regulation reserve needs are small; on average <1.2% and <0.05% of installed PV capacity respectively. These rise to 5.6% and 0.2% in centralized scenarios. Most the benefits of this dispersion can be achieved with relatively few, 25, systems. These are sized at roughly 500 MW each, which is about the size of the largest systems in California today. Almost all of the load following reserve need is driven by errors in the hourly forecasts of solar generation. These can be mitigated either by better forecasts, or dispersing plants.Siting policies for PV must weigh system flexibility against other locational concerns, such as the energy and capacity value of the solar resource in an area. We find a small trade off between energy and reserves; where dispersed systems require less reserves but also have lower capacity factors than more centralized systems. However, we find a much greater trade-off between energy and capacity value in California; where the regions that produce the most energy on average in the Mojave desert tend to be cloudy during current peak demand hours, which occur during Summer afternoons. (C) 2016 Elsevier Ltd. All rights reserved.

In this paper, we construct, fit, and validate a hidden Markov model for predicting variability and uncertainty in generation from distributed (PV) systems. The model is unique in that it: 1) predicts metrics that are directly related to operational reserves, 2) accounts for the effects of stochastic volatility and geographic autocorrelation, and 3) conditions on latent variables referred to as "volatility states." We fit and validate the model using 1-min resolution generation data from approximately 100 PV systems in the California Central Valley or the Los Angeles coastal area, and condition the volatility state of each system at each time on 15-min resolution generation data from nearby PV systems (which are available from over 6000 PV systems in our data set). We find that PV variability distributions are roughly Gaussian after conditioning on hidden states. We also propose a method for simulating hidden states that results in a very good upper bound for the probability of extreme events. Therefore, the model can be used as a tool for planning additional reserve capacity requirements to balance solar variability over large and small spatial areas.

Power systems are undergoing a paradigm shift due to the influx of variable renewable generation to the supply side. The resulting increased uncertainty has system operators looking to new resources, enabled by smart grid technologies, on the demand side to maintain the balance between supply and demand. This study uses a unique data set to estimate and validate models of demand response from residential thermostatically controlled loads (TCLs)---specifically, HVAC units---and quantifies the extent to which a population of TCLs can provide demand response (DR). We use measured temperature setpoints, internal temperatures, compressor cycling ratio and metered energy data collected from over 4200 homes in Texas during the summer of 2012. Using autoregressive moving average (ARMA) models for individual households, we investigate the instantaneous power shed, the duration of the power shed, steady state energy savings and total energy savings. Specifically, we provide insight into the dependency of household DR availability to the temperature setpoint schedule, outdoor air temperature and time of the day.

This paper presents a new method for estimating the demand response potential of residential air conditioning (A/C), using hourly electricity consumption data (“smart meter” data) from 30,000 customer accounts in Northern California. We apply linear regression and unsupervised classification methods to hourly, whole-home consumption and outdoor air temperature data to determine the hours, if any, that each home׳s A/C is active, and the temperature dependence of consumption when it is active. When results from our sample are scaled up to the total population, we find a maximum of 270–360MW (95% c.i.) of demand response potential over a 1-h duration with a 4°F setpoint change, and up to 3.2–3.8GW of short-term curtailment potential. The estimated resource correlates well with the evening decline of solar production on hot, summer afternoons, suggesting that demand response could potentially act as reserves for the grid during these periods in the near future with expected higher adoption rates of solar energy. Additionally, the top 5% of homes in the sample represent 40% of the total MW-hours of DR resource, suggesting that policies and programs to take advantage of this resource should target these high users to maximize cost-effectiveness.

This paper parameterizes the distribution of fluctuations of generation from over 100 solar photovoltaic (PV) systems in the areas of San Jose, Los Angeles, and the Central Valley of California using a Hidden Markov Model (HMM) with Gaussian emissions. Emissions from the hidden states of the HMM, referred to as volatility states, have similar means and different variances, thus accounting for the high kurtosis of the general distribution. The resulting Gaussian shape of emissions from the HMM allows for simple prediction of the distribution shape of the sum of fluctuations from many systems. Geographic auto-correlation among fluctuations from neighboring systems is also assessed and is found to be highest for the volatility state with the second highest emission variance. Preliminary evidence shows that volatility states are dependent on cloud cover observations at a nearby ground station, indicating that distributions of these states may be predictable with commonly observed weather data.

T et al. 1 present a useful template for evaluating polymers from cradle to resin. It is necessary to point out one error and one significant caution associated with this methodology. In Table 2, overall Atom Economy associated with polyvinyl(chloride) (PVC) is given without reference as 55%. This is incorrect. The process of making PVC involves (1) addition of chlorine to ethylene over an iron catalyst to produce 1,2dichloroethane (EDC), (2) dehydrochlorination of EDC to vinyl chloride, (3) recycling coproduct HCl with ethylene and oxygen over a copper catalyst to generate more EDC, and (4) free-radical polymerization of vinyl chloride with recycling of any unpolymerized material. The atom efficiency of processes 1, 2, and 4 are 99%, 98%, and 99þ%, respectively. Process 2 is carried to 50 60% conversion to avoid side product formation; however, unreacted EDC is purified and recycled in the cracking process. Side product formation is about 2%, and conversion on a chlorine basis of EDC to VCM is about 97% for ultimate atom economy of about 94%. Omission of the recycling process in the calculation may be the source of the error. Treatingmaterial from cradle to resin, while a useful step in life cycle assessment, can be misleading. Differences in physical properties can distort the assumption that equal volumes of materials will be used to produce fungible articles. As an example, a significantly larger mass of high-density polyethylene (HDPE) than PVC must be used to manufacture pipe with the same pressure specifications. This difference is significant and may impact the decision on the “greenest” raw material for an application, but by design is not captured by the methodology in this article. This comment should not detract from what is a good attempt to put green chemistry metrics in the context of real product choices. At the same time, the error and caution must be taken into account if correct choices are to be made on that basis.

Proper pattern organization and reorganization are central problems facing many biological networks which thrive in fluctuating environments. However, in many cases the mechanisms that organize system activity oppose those that support behavioral flexibility. Thus, a balance between pattern organization and pattern flexibility is critically important for overall biological fitness. We study this balance in the foraging strategies of ant colonies exploiting food in dynamic environments. We present discrete time and space simulations of colony activity that uses a pheromone-based recruitment strategy biasing foraging towards a food source. After food relocation, the pheromone must evaporate sufficiently before foraging can shift colony attention to a new food source. The amount of food consumed within the dynamic environment depends non-monotonically on the pheromone evaporation time constant—with maximal consumption occurring at a time constant which balances trail formation and trail flexibility. A deterministic, 'mean field' model of pheromone and foragers on trails mimics our colony simulations. This reduced framework captures the essence of the flexibility-organization balance, and relates optimal pheromone evaporation to the timescale of the dynamic environment. We expect that the principles exposed in our study will generalize and motivate novel analysis across a broad range systems biology.

This study evaluates the efficacy of green design principles such as the "12 Principles of Green Chemistry," and the "12 Principles of Green Engineering" with respect to environmental impacts found using life cycle assessment (LCA) methodology. A case study of 12 polymers is presented, seven derived from petroleum, four derived from biological sources, and one derived from both. The environmental impacts of each polymer's production are assessed using LCA methodology standardized by the International Organization for Standardization (ISO). Each polymer is also assessed for its adherence to green design principles using metrics generated specifically for this paper. Metrics include atom economy, mass from renewable sources, biodegradability, percent recycled, distance of furthest feedstock, price, life cycle health hazards and life cycle energy use. A decision matrix is used to generate single value metrics for each polymer evaluating either adherence to green design principles or life-cycle environmental impacts. Results from this study show a qualified positive correlation between adherence to green design principles and a reduction of the environmental impacts of production. The qualification results from a disparity between biopolymers and petroleum polymers. While biopolymers rank highly in terms of green design, they exhibit relatively large environmental impacts from production. Biopolymers rank 1, 2, 3, and 4 based on green design metrics; however they rank in the middle of the LCA rankings. Polyolefins rank 1, 2, and 3 in the LCA rankings, whereas complex polymers, such as PET, PVC, and PC place at the bottom of both ranking systems.