Energy Harvesting for the Internet of Things
In the world of IoT, we are conditioned by the limitations of batteries, which have a limited capacity. Even if the consumption of IoT devices is becoming less significant each time, sometimes we find use cases where regular batteries are not enough to power our devices. In this blog, we are going to describe a prototype that has been developed for a project called Refer, which we named The Energy Harvesting Loadsensing.
Loadsensing
For those who are not familiar with Loadsensing, this is one of the main Worldsensing products. It is a datalogger that wirelessly monitors infrastructures in remote locations. It is widely used over the world by construction and mining companies, as well as operators of bridges, tunnels, dams, railways, etc.
The way Loadsensing works is the following. A great variety of sensors can be connected to the wireless node (datalogger) and after a simple configuration via cellphone app, the node is ready to communicate wirelessly to a gateway using the LoRa protocol, which allows a long range communication radio. Data is then forwarded to the client’s servers, who analyze it through different metrics.
This means that the wireless node communicates sensed values periodically to the gateway. The process of transmitting a packet wirelessly is quite energy consuming. The consumption depends, of course, on the number of bytes transmitted, the distance from the node to the gateway and transmission power. Even if the sampling process consumption is less significant than the transmission process, it also implies some energy consumption for the device. In consequence, we will see that the more frequently we require a measured value (the higher sampling rate or lower sampling delay) the more consumption the node will have to face.
In general, Loadsensing dataloggers already last for several years even in low sampling periods (i.e. once every 5 minutes). However, there are a few specific cases where high consuming digital sensors (or chains of sensors) are desired to be used. So what if clients need to lower the sampling delay to only a few minutes when having a very consuming device or chain connected to the datalogger? How long would the node last? And in case this duration is not enough, is there a way to extend the datalogger’s battery lifetime? We will see that there is and we will present the prototype right away!
Energy harvesting prototype
As it was already mentioned, the Energy Harvesting Loadsensing was developed specifically for a project called Refer. The project is about reducing energy consumption on different types of buildings by improving energy harvesting technologies, as well as improving the energy efficiency of those buildings. Worldsensing’s collaboration on the project is to provide a wireless node, equipped with a dirt sensor in order to qualitatively estimate the level of dirt of photovoltaic panels on a building’s rooftop. As dirty panels are less energy efficient, their cells are periodically cleaned as a maintenance procedure. Sometimes the panels are cleaned even when they are not that dirty and sometimes it is the opposite case. Thus, having a way to estimate the level of dirtiness, allows a more efficient management of the cleaning process.
On the other hand, another Worldsensing task is to validate a new photovoltaic panel technology developed by a partner in the project. This is why we chose to create an Energy Harvesting Loadsensing, which would connect a digital dirt sensor prototype and integrate a photovoltaic panel and rechargeable batteries to the regular Loadsensing that we all know. This is how it looks like, compared to the original Loadsensing:
We can see that we have added the dirt sensor on the side and a photovoltaic panel on top. It also has a charger and rechargeable batteries inside and a support on the back so that the panel can be pointed on the most suitable angle to receive solar radiation during the day.
As we mentioned, the goal was to place these dataloggers on the rooftop of a building, in order to monitor the level of dirt of the photovoltaic panels installed there. Here we can see the prototypes attached to the panels:
Lifetime estimation
Measuring dirtiness of a panel does not require more than a few samples per day. This means that, for this specific case, our brand new Loadsensing prototype could have been a regular one too and it would have lasted for years. Nevertheless, any other digital sensor can be connected to the Energy Harvesting Loadsensing and if the sampling delay needed is low enough, then it makes a lot of sense to have such a solution, specially taking into account the very high amount of energy consumed by this sensor.
Furthermore, we also need to estimate how long the node will last and it would be interesting to compare it with the original Loadsensing duration. This is why we decided to create a software that estimates both the lifetime of a datalogger with no harvesting capabilities and compare it to the one with harvesting equipment.
Both variants share the same mean consumption calculation algorithm. It receives the following parameters: Loadsensing model (digital, tiltmeter, voltage, etc), model region (EU or FCC), spreading factor, sampling delay. Once the consumption is known, both the harvesting and non-harvesting lifetime algorithms require a temperature profile of the location where the node is installed. This is because battery capacity varies significantly with temperature. Knowing how this varies, allows us to calculate lifetime more precisely.
It is important to mention that, in order to have a fair comparison, the batteries modeled in both harvesting and non-harvesting cases are rechargeable ones. These are not the original non-rechargeable batteries used by commercialized Loadsensing dataloggers, which would have a higher performance.
Modeling battery discharge is not easy, but when you also have the possibility of charging the battery, then things get even more complicated. Moreover, we should also estimate how much power we will receive from the photovoltaic panel, provided by sun radiation, which also varies a lot depending on several factors, such as the weather, time of the year and geographic location!
Then, we need to consider firstly, the power output generated by a solar panel given a time and location, and secondly, how this energy input will affect the state of charge of the battery, considering that there is a mean consumption, which we have already estimated.
Considering output power calculation, there is a very interesting paper that describes how to estimate the received solar radiation, given the panel’s dimensions, orientation and efficiency, as well as, geographical location, time of the year, etc. It also considers a few correction parameters in order to be adjusted to particular cases. This, in combination with a solar power output model (which also depends on the temperature profile of the location), provides the estimated load to the battery from the solar panel.
The final step is to model the State of Charge (SoC) of the battery, taking into account both input and output power. It is important to consider that the battery capacity varies over time with the number of charge and discharge cycles and aging. Here we can find a paper describing how SoC can be calculated.
Now that we can model the SoC of the Energy Harvesting Loadsensing and the Non-Harvesting one, we can run the software and plot the results over time. For this purpose, we developed a friendly interface that allows us to choose among the different input parameters. If we take into account a fixed location in the city center of Barcelona and choose the Refer Node (Loadsensing with the digital dirt sensor connected), EU region, spreading factor 9, sampling delay of 3 minutes, starting the 1st of January, this it what we obtain as an output:
What we can see is that the Non-Harvesting Loadsensing will last about 15 days, while the Harvesting one will last more than 60. If we observe the curve, the SoC decreases at a slower rate as the number of days increase. This is because spring starts and the solar radiation is more powerful, which provides more charge to the batteries.
It is interesting to compare these results with the same set up but starting the 1st of June instead. This is the output obtained:
In this case, we can see that in the beginning, we have a full SoC for at least 100 days. The reason is summer, where the solar radiation is quite powerful and does not let the sensor discharge much. After that, as the solar radiation decreases during autumn and winter, the SoC will also gradually decrease more and more until the node is fully discharged, which is estimated to be after 187 days.
What’s next?
Despite there are still several aspects of the lifetime software to validate and improve, this is a good starting point to consider the actual implementation of a new version of Loadsensing, enhanced by energy harvesting capabilities for cases where the regular solution would not be enough in terms of power requirements. If integrating energy harvesting is the future of IoT, then we can think of the Energy Harvesting Loadsensing as the evolution of our datalogger.
If you found this post interesting, we are always hiring and interested in meeting all types of engineers, regardless of your skills or what tools you use day-to-day. Your intelligence, creativity, energy and enthusiasm are much more important to us than your experience with our stack.
Check out our careers page in here — https://worldsensing.wpengine.com/engineering/
Disclaimer
Aquest projecte ha estat cofinançat per la Unió Europea a través del Fons Europeu de Desenvolupament Regional (FEDER)
