Team course project (4 people)

Photovoltaic MPPT — P&O and its improved variant

Université de Technologie de Compiègne · BA04 Renewable Energy Conversion and Management (spring 2024)

Modelled a Kyocera KC200GT module in MATLAB/Simulink and compared P&O against the improved P&OA over 366 days of per-minute irradiance and temperature measured by UTC's STELLA platform, in terms of annual energy and efficiency.

Background

A photovoltaic module’s I-V curve has a single maximum-power point (MPP), and its position moves continuously with irradiance and temperature. If the load pins the operating point, the module can drift far from the MPP and lose 20-30% of its production. The job of MPPT is to keep the operating point on the MPP with a DC-DC converter.

This was the BA04 “Renewable Energy Conversion and Management” course project at UTC, a team of four, defended in June 2024. The object was a Kyocera KC200GT (200 Wp polycrystalline module). We built the PV module, a SEPIC converter and the controller in Simulink, then drove the model with one full year of measurements from UTC’s STELLA platform. The report compares four algorithms: P&O, incremental conductance (INC), an improved P&O (P&OA) and fuzzy logic.

Data and simulation

  • Data: 1 April 2023 to 31 March 2024, 366 days; one sample per minute, 1440 points per day; irradiance, ambient temperature and cell temperature.
  • Model: the module is described by a parameterised exponential model; the solver step is 2 s, and daily runs are concatenated into a full year.
  • Analysis: instantaneous power is integrated into daily energy (kWh) and divided by the incident solar energy of the same day (kWh/m²):
E_day = Σ P(t) · Δt / 3.6e6      [kWh]
η     = Σ E_day(PV) / Σ E_day(G)

The code has three layers: loadYearData / loadDayData read the Excel files and rebuild the time axis, runSimulation / runYearSimulation run the model, and runAnalysis computes energy, efficiency, CSV files and the comparison figure. Every path is relative to the repository root, so the project can be moved freely.

My contribution

In the four-person split I owned P&O and its improved version P&OA:

  • implementing both duty-cycle perturbation schemes in Simulink
  • injecting the per-minute irradiance and temperature series into the From Workspace blocks, one day at a time
  • batching all 366 days and concatenating the instantaneous power into a yearly series
  • computing daily and annual energy, efficiency, and the comparison plots
  • rewriting and tidying the code after the course: function split, relative paths, README and reproduction steps

The other group methods (incremental conductance, fuzzy logic) are kept in archive/other_methods/; the report and defence material are in docs/.

Results

  • Annual energy: 180.28 kWh with P&O, 179.36 kWh with P&OA — a difference of about 0.5%.
  • Over a full year the two power curves almost overlap. P&OA’s advantage shows up mainly during irradiance transients and in steady-state oscillation when the step is badly tuned.
  • The gain sweep in the report (0.1 / 0.5 / 1) shows the first two settings behaving similarly, while a gain of 1 noticeably lowers the recovered power — too large a step amplifies the oscillation.

An honest conclusion: with one-minute data over a whole year, “annual energy” smooths away the transient differences between the algorithms. A fairer comparison of P&O and P&OA uses settling time after a step, steady-state power ripple, and tracking efficiency at different step sizes.

Source, Simulink models, raw data and report: https://github.com/yicongl1/MPPT-direct

Technical notes

01 The PV maximum power point and P&O: why perturb, why oscillate, where the improvement lies I-V curves and the MPP, the P&O decision rule and steady-state oscillation, variable-step P&OA, and why annual energy differs by only 0.5%. 02 Feeding a full year of measurements into Simulink: day loops, time series and energy integration Reshaping a 1440-column Excel sheet, injecting time series, concatenating daily runs, integrating kWh at a 2 s step and choosing the efficiency denominator. 03 Why SEPIC suits PV: converter choice and four MPPT algorithms Buck / Boost / Buck-Boost / Cuk / SEPIC trade-offs, the two-phase operation, and fixed-step versus variable-step algorithms.

Related skills

  • Simulink
  • MATLAB
  • Academic writing