Evaluation of Missing Data Imputation Methods for an Enhanced Distributed PV Generation Prediction

Sundararajan, Aditya

doi:10.1007/978-3-030-32520-6_43

Cited by 16 publications

(8 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Data from the PV system are subject to exploratory analyses and cleaning as outlined in the authors' prior work for similarly sized PV systems [34]- [38]. These include imputing missing values [39] and statistical curve fitting [40]- [42]. Time-series data on X = {I, W S, AT , RH, AP }, and Y = Gen from the month of September 2019 were used for this study.…”

Section: A Data Preparationmentioning

confidence: 99%

Regression and Generalized Additive Model to Enhance the Performance of Photovoltaic Power Ensemble Predictors

Sundararajan

Ollis

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

Photovoltaic (PV) power prediction has a constantly evolving solutions landscape with a myriad of data-driven techniques. Each technique leverages a self-adaptive algorithm that must retrain in intervals, be it each day, week, or season, to avoid the model generalizing poorly because of overfitting, underfitting, or concept drift. This paper aims to improve the generalization capability of PV power predictors such as autoencoders used widely in the industry by introducing feature-enhanced ensemble learning (FEEL) after the feature selection step. This framework uses a combination of nonparametric regression and generalized additive models, and an ensemble of weak regularized multilayer perceptron models. Once trained, the framework can reliably generalize on test data across long time periods without any significant degradation in performance. The proposed framework was validated against the baseline autoencoder-based feature enhancement model on a real PV system from a smart neighborhood in Alabama for September 2019. The FEEL framework performed three times better than the baseline, but when applied to the baseline, its performance improved by two times on average. Furthermore, the framework generalized consistently better than five other feature enhancement strategies. Despite fluctuations in weather, the FEEL framework's R-square score had a range of 8.1%, whereas that of the baseline was 48.3%. The mutual information and Minkowski distance scores attempted to quantify concept and model drift, respectively. These scores show that the FEEL framework generalized the ensemble learning models at least two times better than the baseline across the different test days. These results form the first step toward decentralized intelligence for smart grid applications that could free up resources for other expensive analytics in the field. INDEX TERMSFeature enhancement, nonparametric regression, ensemble learning, generalized additive model, concept drift, photovoltaic power prediction. Nomenclature Unless otherwise specified, the following nomenclature is applied throughout this paper. Abbreviations AE Autoencoder, also denotes one of the feature enhancement strategies ARIMA Autoregressive integrated moving average ARIMAX ARIMA with extragenous inputs ERM Ensemble of regularized multilayer perceptrons FEEL Feature-enhanced ensemble learning GAM Generalized additive model

show abstract

Section: A Data Preparationmentioning

confidence: 99%

Regression and Generalized Additive Model to Enhance the Performance of Photovoltaic Power Ensemble Predictors

Sundararajan

Ollis

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…Preprocessing (PrP): In centralised computing, data are transmitted directly to the utility's repository or the cloud, but in decentralised edge or fog computing, preprocessing is pushed closer to data sources [2][3][4]. Preprocessing steps include, but are not limited to, checking for completeness [16], consistency, and integrity [17], dimensionality reductions to better ingest and store data [18][19][20][21], and applying restructuring techniques such as splicing, dicing, splitting, and aggregation.…”

Section: Big Data Lifecycle Stagesmentioning

confidence: 99%

Adapting big data standards, maturity models to smart grid distributed generation: critical review

2020

Self Cite

View full text Add to dashboard Cite

Big data standards and capability maturity models (CMMs) help developers build applications with reduced coupling and increased breadth of deployment. In smart grids, stakeholders currently work with data management techniques that are unique and customised to their own goals, thereby posing challenges for grid-wide integration and deployment. Although big data standards and CMMs exist for other domains, no work in the literature considers adapting them to smart grids, which will benefit from both. Further, existing smart grid standards and CMMs do not fully account for big data challenges. This study bridges the gap by analysing the role of big data in smart grids, and explores if and how big data standards and CMMs can be adapted specifically to ten distributed generation (DG) use-cases that use big data. In doing so, this work provides a useful starting point for researchers and industry members developing standards and CMM assessments for smart grid DG. 3 Emerging and existing big data standards Key big data standards are described below, their relevance to attributes and lifecycle stages summarised in Table 2.

show abstract

“…Locally installed weather stations measure GHI, module temperature and ambient temperature with up to a resolution of 1 minute. A cloud-based DAS is used to access the raw data for further processing and analysis, as described in the authors' previous works [36][37][38]. Systems A and B are connected to feeders A and B, respectively, which are radially distributed from their substations located in Miami and Daytona, respectively.…”

Section: The Eclipse Of August 21 2017mentioning

confidence: 99%

Case study on the effects of partial solar eclipse on distributed PV systems and management areas

Sundararajan¹,

Olowu²,

Wei³

et al. 2019

IET Smart Grid

Self Cite

View full text Add to dashboard Cite

Photovoltaic (PV) systems depend on irradiance, ambient temperature and module temperature. A solar eclipse causes significant changes in these parameters, thereby impacting PV generation profile, performance, and power quality of larger grid where they connect to. This paper presents a case study to evaluate the impacts of the solar eclipse of August 21, 2017 on two real-world grid-tied PV systems (1.4MW and 355kW) in Miami and Daytona, Florida, the feeders they are connected to, and the management areas they belong to. Four types of analyses are conducted to obtain a comprehensive picture of the impacts using 1-minute PV generation data, hourly weather data, real feeder parameters, and daily reliability data. These analyses include: individual PV system performance measurement using power performance index; power quality analysis at the point of interconnection; a study on the operation of voltage regulating devices on the feeders during eclipse peak using an IEEE 8500 test case distribution feeder; and reliability study involving a multilayer perceptron framework for forecasting system reliability of the management areas. Results from this study provide a unique insight into how solar eclipses impact the behavior of PV systems and the grid, which would be of concern to electric utilities in future high penetration scenarios.

show abstract

Evaluation of Missing Data Imputation Methods for an Enhanced Distributed PV Generation Prediction

Cited by 16 publications

References 37 publications

Regression and Generalized Additive Model to Enhance the Performance of Photovoltaic Power Ensemble Predictors

Regression and Generalized Additive Model to Enhance the Performance of Photovoltaic Power Ensemble Predictors

Adapting big data standards, maturity models to smart grid distributed generation: critical review

Case study on the effects of partial solar eclipse on distributed PV systems and management areas

Contact Info

Product

Resources

About