Cycle 5 (2015 Deadline)
Seamless solar PV integration in Moroccan buildings
PI: Mounir Ghogho (email@example.com), International University of Rabat
U.S. Partner: Paul Flikkema, Northern Arizona University
A high density of installed photovoltaic (PV) systems poses grid integration challenges. Excessive ramps and peaks of the injected PV power into low-voltage grids are some of the most important challenges that may destabilize the grid and even cause blackouts. The penetration of PV in buildings in Morocco is still very low. The potential for roof-mounted PV power generation in Morocco has been estimated to each 10TWh, representing 40% of the country's total electricity consumption. Despite this high potential, there are two main obstacles to the uptake of PV in buildings, the first being the capital investment cost, as no financial incentives are offered yet by the government, and the second being the conservative attitude of the national electricity utility company toward PV injection in the low-voltage grid.
To optimally design a battery-supported PV system, realistic household load profiles in Morocco must be used. The general objectives of this project are (1) to build a statistically significant dataset of household electricity consumption profiles in Morocco and make it available as open data to the scientific community; (2) to analyze the data using machine learning to assess and classify power quality at the household level; and (3) to develop scientific tools to investigate issues of sizing and operation of residential battery-supported PV systems in Moroccan settings from both the energetic and economic perspectives. Forecast-based control mechanisms will be devised for different scenarios. The results of the project should be of interest to the Moroccan Ministry of Energy and the national electric utility company. The International University of Rabat encourages through its technology transfer unit joint ventures with industry to turn research results into commercial products. The research team will explore this avenue at the end of the project. Overall, through their planned open data platform and outreach activities, the project team aims to contribute to the promotion of clean energy adoption in buildings and the modernization of the Moroccan electric grid.
Summary of Recent Events
During the quarter ending June 2017, the team worked on four main tasks:
They classified sky images into two weather condition classes: clear or not clear sky, using statistical features of images and specific machine learning algorithms. This is the first module in the process of short term PV power forecasting and it yielded very good results, with a classification error around 0.5%.This Classifier was built using Matlab and its machine learning toolbox. Subsequently, they investigated the problem of finding a good model mapping the sky image and the solar irradiance, which is the second module in the process of short term PV forecasting. Different modeling methods, based on statistical features and different regression techniques (support vector machines, neural networks) have been tried and so far the modeling error in predicting the solar irradiance from the sky image is still not low enough (around 50W/m²). They are currently exploring other methods to improve performance.
Study of the literature on the estimation of active and reactive powers of electrical signals was also done. Since, no real data was available yet they simulated electrical signals in different conditions using LabVIEW and implemented different estimation methods. The obtained estimation performance was good however, this will be investigated again when real signals are available.
The team developed the National Instruments (NI)-based acquisition system prototype. This was done by debugging the code and adding new features to the system so it could be simple to deploy (start acquisition when turned on), easy to detect a malfunction in case of an error (color-changing LED indicating Acquisition ON, Acquisition OFF and error states) and reliable to collect and save data to a USB.
One of the most popular and simple models for the PV cell found in the literature studied is the single diode model (SDM). This model describes the PV panel as a current source with a nonlinear diode. Their electrical parameters can be described as functions of the irradiance and the panel temperature (see De Soto's model). In the literature studied, the team also found out that aging degrades irreversibly the performance of the solar panel and hence also modifies the parameters of the mathematical model.
The PI and his team studied the main types of batteries, which are lead-acid battery, nickel based batteries and lithium ion batteries. According to the literature, nickel based batteries are pollutants so their use in their project will most likely be discarded. Lead-acid batteries are cheap but their specific energy (i.e. energy per weight) is very low and so if they choose this type of batteries they may require a large number to be purchased. Lithium-ion batteries have a high specific energy but they are more expensive. Therefore a final economic and engineering analysis will have to be made to select the type of battery to be used. They also found a simple but relatively accurate mathematical model, which describes the current-voltage behavior of the Lithium-ion batteries during the charge and discharge processes.
Lastly, they explored the main classical techniques to analyze and model this type of circuits using nonlinear differential equations. They have also explored an oversimplified model which can be useful for the analysis of our system at large scale.
In the next 3-6 months, the team plans to purchase equipment to duplicate the prototype system, including current sensor, IO modules, acquisition card and USBs used, in order to install them in multiple and different Moroccan households for a one month period. Toward the objective of PV power forecasting, research work will be undertaken to find a more accurate representation of the relationship between solar irradiance and sky images (second module). They will estimate active and reactive powers for non-sinusoidal conditions and non-linear loads, which are characterized by harmonics and other distortions in the waveforms. Once data is collected, they will investigate the problems of load forecasting, as well as the analysis and classification of the electrical power quality. Lastly, they will study the principles behind the operation of the inverter as well as its different mathematical models. This would complete the mathematical modeling of the main components of the system. Afterwards, they will look into architecture of the switching system which will feed the electrical energy produced by the PV cell and stored by the battery either into the electrical grid or into the appliances.
Back to PEER Cycle 5 Grant Recipients