CPS-Solar-Project

Latest Updates for 2021

The aim of this collaborative project is to increase the efficiency of utility scale solar arrays using sensors, machine learning and signal processing methods to detect faults and optimize power. New cyber-computing strategies, that rely on sensor data and imaging methods to predict solar panel shading, are used to improve efficiency. A programmable 18kW testbed that consists of 104 panels equipped with sensors, actuators and cameras is used to validate all theoretical results and test new approaches for using solar analytics to optimize power generation. Machine learning and dynamic image modeling algorithms are used to control each individual panel and change connection topologies to optimize power for different cloud, load, and fault conditions.

Outcomes of the CPS project include advances in: a) cloud movement modeling and shading prediction using computer vision algorithms, b) PV fault detection and optimization methods that will switch array topologies dynamically while limiting PV inverter transients, d) experimental (testbed) validation of all array monitoring methods, and e) secure wireless sensor and data fusion. Theoretical and experimental research which enables real-time analytics and remote connection topology control may influence PV array standards and smart grid initiatives. The project tasks also include: education activities, outreach at high schools, and engagement with several organizations including minority and HBCU institutions to enhance diversity.

Image-based measures of sky-clarity, an attribute useful for predicting shading. This metric was created from dynamical models of image texture, with a manifold-based metric on dynamical model parameters. Sample images at various times show how the index separates ‘clear skies’ and ‘hazy/cloudy skies’. Using a small network of horizon-viewing cameras it is possible to develop early warning systems.

Spatio-temporal modeling of sky videos using GIST and largest Lyapunov exponents; with time stamps for clear sky, transition from clear-cloudy sky and cloudy sky. These transitions can be used for prediction.

Clustering using GMM. Training forms clusters of normal and simulated fault PV data.

A novel distributed algorithm for estimating the maximum of the node initial state values in a network, in the presence of additive communication noise is proposed. Conventionally, the maximum is estimated locally at each node by updating the node state value with the largest received measurements in every iteration. However, due to the additive channel noise, the estimate of the maximum at each node drifts at each iteration and this results in nodes diverging from the true max value. Max-plus algebra is used as a tool to study this ergodic process. The subadditive ergodic theorem is invoked to establish a constant growth rate for the state values due to noise, which is studied by analyzing the max-plus Lyapunov exponent of the product of noise matrices in a max-plus semiring. The growth rate of the state values is upper bounded by a constant which depends on the spectral radius of the network and the noise variance. Upper and lower bounds are derived for both fixed and random graphs. Finally, a two-run
algorithm robust to additive noise in the network is proposed and its variance is analyzed using concentration inequalities. Simulation results supporting the theory are also presented.

We describe a Cyber-Physical system approach to fault detection in Photovoltaic (PV) arrays. More specifically,
we explore customized neural network algorithms for fault detection from monitoring devices that sense data
and actuate at each individual panel. We develop a framework for the use of feedforward neural networks for
fault detection and identification. Our approach promises to improve efficiency by detecting and identifying
eight different faults and commonly occurring conditions that affect power output in utility-scale PV arrays.

A neural network algorithm was deployed to detect more faults and shading conditions including ground fault, arc faults, varying temperature conditions, varying shading patters, short circuits and degraded modules. Identifying these faults helps improve mean time to repair at a faster rate.

We also show the limitations of traditional clustering algorithms for a multi-class classification problem. Faults have similar maximum power points and clustering algorithms are unable to classify the eight faults and shading cases considered. A neural network overcomes the limitations of the traditional clustering algorithms and classifies the eight cases with a high level
of accuracy.