Shahar Ben-Menahem

Neural network estimation of photovoltaic I–V curves under partially shaded conditions

In this paper, we present a neural network algorithm to estimate the I-V curve of a photovoltaic (PV) module under non-uniform temperature and shading distributions. We first present a novel photovoltaic simulation model which includes the interaction of (1) heat transfer including conduction, convection, and radiation (long and short wavelength), (2) an electro-optical two diode model including ohmic heat dissipation, and (3) environmental influences including shading, irradiance, and wind dependencies. The neural network trains on inputs which consist of shading and temperature patterns of each cell of the module, and predicts the current versus voltage and power versus voltage landscapes. This information can be used for maximum power point tracking under non-uniform conditions. The neural network was validated on the simulation model and on data collected from our rooftop PV lab.

Online photovoltaic array hot-spot Bayesian diagnostics from streaming string-level electric data

We report on a physics-based suite of machine-learning algorithms which performs online, real-time photovoltaic-array incipient hot-spot diagnostics and prognostics, using as input panel-string electrical sensor data. These data consist of module-string currents and voltages and their instantaneous (filtered) rates of change, acquired via a combination of analog and digital sensors. Our diagnostics algorithms comprise thermal and electrical System ID stages, augmented by a Bayesian MAP (Maximum A posteriori Probability) inference engine utilizing a simplified stochastic, spline-based integrated plant model in the optical, electrical and thermal domains. Our (path-integral, physics-based) MAP inference engine detects incipient hot-spots; attributes them to the most likely proximate- and underlying causes (such as shading or dust-deposition patterns, direct heating, inherent mismatches or circuit failures); and renders quantitative estimates of the most likely culpable environmental factor (e.g. a time-varying shading pattern). The same simplified plant model is periodically used to prognosticate the likely future evolution of the incipient hot spot. Our diagnostics and prognostics algorithms can be applied to simulated and/or empirical data (the latter from our roof-top PV lab at Carnegie Mellon). The algorithms leverage changes in estimated conductances in different regions of the electrical state-space, and at different granularities (cells, panels, strings and full array). It take into account temperature coefficients, ohmic- and optical heating, and (optionally) inherent electrical mismatches. Our algorithm suite requires as input neither temperature sensors, nor sub-panel electrical data, nor load sweeps; even pyranometer data is optional. The only necessary sensor data are string electrical data at the actual instantaneous in-operation external load (either with or without a perturb-and-observe MPPT control procedure). Digitized measurements of the low-pass-filtered time derivatives of these currents are useful, but if not available they can be replaced with numerical differentiation. However, occasional voltage sweeps - augmented with pyranometer, thermal and intra-panel (bypass diode voltage) data from our heavily instrumented PV array - along with online astronomical and meteorological data and artificially induced shading patterns - are used to validate our algorithms and to improve the baseline System ID parameters estimation.

Rollable nano-etched diffractive low-concentration PV sheets for small satelites

This paper discuses a novel, rollable, mass fabricable, low-concentration photovoltaic sheets for Cubesats providing them with efficient photoelectric conversion of sunlight and secondary diffuse light. The wrap consists of three thin (of order a millimeter or less), cheap plastic-sheet layers, which can be rolled together in a spiral wrapping configuration when stowed. Preliminary simulation based on the above modeling approaches show that the designs achieve comparable photovoltaic power (area for area) and (b) result in a at angular response curve which remains at from normal incidence of over 35 degrees to the normal. The simulation were performed using a ray tracing simulator built in Matlab. In addition, we have constructed a demonstrator using quartz wafers based on the optimized design to show the technology. Details of its fabrication are also provided.

LPFG sensing network for distributed shape control

In this paper, we discuss various aspects of the control and sensing in a flexible wing aircraft using embedded LPFG (Long Period Fiber Grating). Driven by the need to improve aerodynamic efficiency and reduce fuel burn, interest in light-weight structures for next generation aircraft has been on the rise. However, in order to fully exploit novel lightweight structures, there is a critical need for distributed sensing along the entire wing span and its integration with closed-loop control systems. A model of an LPFG sensor string embedded in an Euler-Bernoulli beam is proposed along with an associated control algorithm.

Optimization of a low concentration photovoltaic system

In this paper, we discuss optimization of a novel low-concentration photovoltaic system with the following properties: (1) static concentration without the need for tracking (2) thermal uniformity via Diffraction Efficiency Modulation (DEM), and (3) mass-fabricability and rollability. The approach leverages a unique combination of waveoptics modeling, multi-objective thermal-electro-optical optimization, and mass-fabricable, nano-manufacturing technology. We discuss various aspects of the optimization including a novel Helmholtz FD solver and thermal and electrical considerations.

Rollable, efficient, low concentration PV for powering small satellites via diffractive modulation

We present a novel LP-CPV (Low Power Concentrating Photo-Voltaics) technology well suited for Low Earth Orbit small satellites. The LP-CPV consists of three layers: the first two layers are optical redirection layers which implement non-tracking concentrating functionality and the third one is a support layer for single crystal silicon lamellar strips covering 10% of the overall area. The optical layers are diffractive transparent plastic layers embossed with micro-structures via roll-to-roll embossing. Instead of using spectral dispersion, we use diffraction efficiency modulation to reduce the amount of IR light concentrated on the crystal silicon strips, therefore allowing maximum conversion efficiency even at 10X concentration over silicon. All three layers (concentrating optics and lamellar silicon PV strips) comprise, when deployed, a uniaxial PV system periodic in the other (non-axis) principal in-plane direction. The layers can be rolled together in one compact cylindrical roll to minimize the payload volume. The entire LP-CPV system can then be deployed/unrolled in space and held in place, but can also be rolled back in order to prevent damage from solar flares, micro-meteorites, etc.

Control on landscapes with local minima and flat regions: A simulated annealing and gain scheduling approach

Landscapes containing local minima and ¿flat¿ regions are frequently encountered in multi-layer neural networks that employ sigmoid-like activation function in the hidden layers. Numerous techniques in the neural network community have been proposed to address these issues. In this note, we extend these ideas to the neural network control of nonlinear systems. We propose a solution which employs simulated annealing and a gain scheduled learning rate.