August 9, 2012
The weeks here have been zipping by a little too quickly! I can’t believe I’m already in my last week at Bosch Solar Energy. During my ninth week, I experienced the beautiful valleys and long tunnels (with large ventilation propellers) along the autobahn from Thüringen to Bayern as I rode with Andre, my supervisor, to another Bosch location in Bamberg. The location has a large PV installation on the roof, divided into two parts, each connected to a small field computer to collect data. The computers were old, and one of them had stopped working, so our mission was to install new ones. It was a pretty straightforward job, and once they were in, we walked past each row of modules to check that the inverters were working properly. Andre was a little perplexed that they all were, because he had been seeing problems with the data collected, but everything seemed to be in order. Back in Arnstadt, we climbed up to the roof to stick metal supports under the measurement cabinet we had installed some weeks earlier, as well as to check the temperature sensors and wiring.
Things progressed steadily back at my desk. My user interface for the optimization program has all its desired functionality, and performs the optimization reasonably well. I’ve also written up documentation for all three of my projects and tried to make sure that the tools I’ve developed will be accessible for anyone in the future who wants to use them or develop them further.
The following week, my department head, Wolfgang Bronner, organized an after-work team event at ‘Kletterwald’ (climbing forest), a ropes course, which was followed by a barbeque. At work, I typically only see most of my coworkers during lunch, so this was a great chance for me to get to know them in an informal setting, especially before they left for vacation. This was the last week of July, and in Germany, August is a prime vacation time. Of the eight employees in the EPS division (Engineering Photovoltaic Systems), only two are here this week.
Last Friday I presented the results of my projects to an audience of three. It was good for me to actually try to explain what I did, and they brought up some important questions that I hadn’t paid attention to before. When running the self-consumption simulations, I had entered in some parameters like number of modules, and let the simulation go (since it had been written by someone else). My colleagues asked about the inclination of the panels, and why the 5kW peak power was never reached, questions which hadn’t occurred to me.
In addition to preparing my presentation, I spent more time on the roof last week to prepare for testing out a new inverter on one third of our test system. We had to connect DC-DC converters to each of the modules and lay down new cables to the inverter, as well as drilling holes into the metal rack to hang up the inverter. When the inverter was connected and turned on the following day, the installer had some questions about whether the three inverters were considered one system or three separate ones. Apparently there is a law that for systems larger than 10kW peak power, the power company must be able to control the ‘reactive power’. I still am not entirely sure how reactive power works, but in the end, it means some of the power generated by the PV system doesn’t make it out.
This week I’ve mostly been wrapping things up, organizing my files so that they’ll be easy to find in the future, etc. Some of my coworkers have an interesting project designing a system for shutting off a module/array. They have some prototype junction boxes to test, and over the past couple of days I helped mount the boxes onto the test modules and solder the connections. Unfortunately, I won’t get to watch them run the tests, but I’m glad I got to learn about the project.
Overall, my internship here has been a great experience. I’ve gotten real-world project experience as well as hands-on experience in the solar energy industry, learned loads of new things about solar installations and power systems, improved my German, and met many friendly, helpful, and generally great people. Thank you Cal Energy Corps and thank you Bosch Solar Energy, I can’t imagine a better way to have spent my summer.
Mit freundlichen Grüßen / Best regards
July 4, 2012
After working through some errors and many hours of simulation, I have finally finished the battery capacity simulations using conditions from several countries. The quantity of interest in these simulations is what percentage of the PV-generated power would be consumed on site (as opposed to being fed into the grid). If the size of the PV system is too small (as was the case for the industrial scale simulation) all of the power is consumed even without a battery. If the system size is too large, all consumption can be supplied by solar energy, with excess that needs to be fed into the grid.
The battery capacity has a notable influence when the system and consumption are balanced. I was surprised to find that even with properly sized systems and large batteries, around half of the energy produced would have to be fed back into the grid. It turned out this was a seasonal effect: during the winter, less power is generated, and the battery is drained every day; however, during the summer, the battery quickly reaches capacity and can’t store a large portion of the energy generated. These percentages should help system designers in choosing what capacity battery to attach to a residential system for best results.
The next assignment I’ve started on involves fitting curves to data. As different types of cells and modules are produced, we need to be able to predict how they will behave over time. The standard two-diode model of a solar cell (which I have become quite familiar with) is an equation with several parameters, representing a small resistor in series with a parallel combination of a large resistor, two diodes, and a current source.
Various parameters in the equation can be fit to match data from a particular cell. I was given a program that does the necessary optimization (with eight parameters), but is inconvenient to run, because it requires editing and running a script file, which means it can’t just be sent out to whoever needs to use it (since they probably don’t want to deal with MATLAB). My assignment is to write a graphical user interface that can load the data from a simple csv file and compute the parameters with the click of a button, and then display the values along with a graph of the IV-curves to show how well the model fits the data. It turns out that the program doesn’t do a very good job of fitting the parameters, so I am also working on tweaking it to give more accurate results.
June 11, 2012
As I’m just starting my fourth week at Bosch Solar Energy, my project involves using MATLAB and Simulink to create models of photovoltaic systems, both at the level of electronic circuits as well as overall systems. These models can be used to determine the optimal settings for operating systems effectively. Though I learned a lot about photovoltaic materials at Berkeley, I’ve seen for the first time how the systems work when everything is connected, and it’s not as simple as I’d imagined.
During my first week, I was shown a complicated Simulink model representing the electrical components that extract energy from a PV installation and feed it into the grid. The system is designed to draw exactly the right amount of current from the PV modules to maximize the power output. If all cells receive the same amount of light, it’s not too complicated; but, if some of the cells are shaded, it can completely change the optimal current. For example, a few days ago I was on the roof with coworkers to set up a measurement cabinet. As an experiment, we placed a phone on one of the cells. Looking at the panel through an infrared camera, we could clearly see one square get warmer (taking away power from the system) because cells around it were generating more energy than it could carry.
My initial task was to create a block that would simulate a PV module (a single solar panel) which could be easily substituted into the larger simulation. I had to adapt code to use the right components and add a nice interface. I proceeded to test the model I’d created to make sure it worked as desired. After encountering and tackling numerous errors, the model worked. In the simplest cases, allowing different amounts of light to hit each solar cell caused the numerical solver to insist the equations were impossible. When I inserted the block into the larger model, the solver again refused to do any work regardless of the parameters I gave it. So my first two weeks of work came to a road block.
I’ve put this project aside for the time being to run numbers in a more general simulation, for which the model has already been created. Rather than simulating all of the electronics, this model just uses the size of a PV installation and the connected storage capacity (battery. However, it also incorporates the geographical location and time of year. I’ve actual power consumption data at the level of a household, commercial building and an industrial building, which this model uses to determine how useful the battery is to the system (to minimize feeding power into the grid or drawing power from it).