Performance validation of battery management system under prediction error for photovoltaic based distribution system

Abstract

An effective reduction in power prediction error profile and an improved battery management system design for photovoltaic (PV) based microgrid application are presented in this study, where battery life and power loss are considered to be effectiveness measures. For local energy management the prediction error has a direct influence on distributed generator (DG) control reference calculation and thus in system stability. The silent effect of prediction error in battery energy storage life deterioration is highlighted in terms of battery temperature and power losses. The PV power prediction challenge (null versus positive volatility nature) is addressed with effective error reduction by kernel-based feature mapping function. To obtain fast prediction (operational references to DG primary control) in an online manner, a new fast reduced Morlet kernel-based online sequential extreme learning machine is proposed in this study. The battery (lithium-ion) temperature effect is addressed by introducing a new secondary controller, which comprises battery temperature reference model (model reference) along with rule-based temperature tolerance switching of stacks. The effectiveness of the proposed design is presented by rigorous case studies (MATLAB and TMS320 C6713), where extreme performance is achieved by simultaneous prediction error and local uncertainty.

Islanding and power quality disturbance monitoring in microgrid using adaptive cross variational mode decomposition and reduced kernel ridge regression

“…The nonlinear KRR transforms the original data samples into a higher dimensional feature space using nonlinear kernel functions governed by Mercer's condition . Using a data set comprising N number of training pairs ( x 1 , y 1 ), ( x 2 , y 2 ), …, ( x N , y N ), the linear regression model is formulated as y i = ϕ ( x i ) β , i = 1, 2, ……… , N , where the ϕ ( x i ) is a nonlinear mapping function that transforms the data to a greater dimensional space.…”

“…The main drawback of the simple Kernel based techniques in handling large data sets is the large training time . Thus, this limitation can be improved by generating a reduced kernel matrix of size of ( N × P) from size ( N × N ) by randomly selecting P number of patterns from the training data (here P < N ).…”

Comprehensive design and control methodology for DC‐powered satellite electrical subsystem based on PV and battery

“…Considering the mission lifetime, critical PV cell parameters such as power, voltage and current are derived at EOL. Energy requirement is divided into two parts: E req-op , operational energy requirement calculated in (14), and E req_total , total energy requirement calculated in (15).…”

“…One of the key differences between celestial and terrestrial electrical systems is the nature of load variations, which can be respectively considered as deterministic and stochastic. The design for terrestrial applications is well documented in literature [11][12][13][14][15]. Additionally, with availability of substantial time-series data for terrestrial systems statistical approaches can be exploited to design and reliability integrate power generation sources, specially renewables such as PV and wind, into the grid [16].…”

Distribution Level Aggregator Platform for DSO Support—Integration of Storage, Demand Response, and Renewables

“…Under the umbrella of energy storage, use of battery is the first priority, and the most common and conventional battery technology is based on Lithium-ion cells (Yang et al, 2018). The applications exist in all areas of the modern era (Stroe et al, 2017), and the application of Lithium-ion batteries for distribution system operation is focused here (Satapathy et al, 2018).…”