By Andy Tomaswick
March 29, 2025
The decision on how to supply a Cubesat with electricity is one of the greatest challenges in the design of a modular spaceship. Compromises in the size of the solar panel, battery sizes and power consumption are important considerations when choosing parts and mission architecture. In order to help with these design decisions, a paper from researchers in Ethiopia and Korea describes a new algorithm for machine learning, with which Cubesat designer can optimize their electricity consumption and ensure that these small satellites have a better chance of fulfilling their purpose.
The performance situation for Cubesats is complicated. As a rule, they are driven by solar collectors that have to be used from the “U” structure to which all cubesats are designed. Even if you are used successfully, you will be exposed to wide areas in the sun's rays and temperature, which leads to dramatic fluctuations of your overall performance.
According to the authors, power supply errors cause about a quarter of all mission failures by Cubesate. Several design options, such as MIMO converter (multi-input multi-output), can already alleviate this. However, managing this type of power distribution system is also at costs, since it is designed in such a way that a function called maximum power point tracking (MPPT) is carried out.
Fraser discusses space with Dr. Stephen Sweeney
MPPT is simply a control algorithm that tries to achieve the highest possible performance from the system that is an environment in which it is located. Suppose the incidence of solar radiation dipped slightly because the cubeeat has a not optimal alignment of the sun. In this case, the MPPT algorithm asks him to realize so that the highest solar radiation strength falls on its solar panels.
Several control algorithms are designed in such a way that they optimize the MPPT of a cubesate. This includes creative algorithms with the name Pusturb and observation (P&O), incremental conductivity (INC) and particle swarm optimization (PSO). While all of these are relatively effective in the optimization of MPPT and range from 88 to 94% efficiency, they all have the same weakness sie, and their parameters must be defined before the start of the Cubesat.
Enter one of the most popular algorithms for artificial intelligence – deep learning. The authors describe the development of a deep feedforward Neural Network (DFFNN), which is connected to a standard proportional integral controller that exceeds all other MPPT algorithms. Its efficiency, which they calculated on 97%, based on experiments with simulated data of a one -year mission, also increases the total performance efficiency of the entire system and lowers the “Power Ripple” changes in the performance delivered by the system, the “transient” or temporary changes in the voltage and flow that can damage the potentially damage components.
Switching on a cubesat is an important consideration of design, as in this video series to create one.
Credit – Creating a Cubesat YouTube Canal
The algorithm has some disadvantages because it is very computing -intensive, like all mechanical learning. To solve this problem, the new algorithm uses a technique called Linear Tangents and Neville interpretation. This mathematical technique divides polynomial problems into much easier to solve equations and simplifies the calculation of the desired trajectory of the CubeAT.
Every little piece counts when it comes to improving the Cubesate performance, and this paper contributes to this effort. An improvement of 3% does not seem to be much, but when thousands of hours of engineering and testing are at stake, even small improvements can be life -changing.
Learn more:
Abagero et al. -Deep learning-based MPPT approach to improve the Cubesat current generation
UT – A 3U Cubesat could collect data during asteroid flying
UT-a cubesat mission will recognize X-rays from GRBS and black hole fusions
Ut – the first cubesat with a Hall effect
