Energi surya merupakan salah satu sumber energi terbarukan dengan potensi besar di Indonesia karena letak geografisnya yang berada di daerah tropis dengan intensitas cahaya matahari tinggi sepanjang tahun. Namun, panel surya konvensional masih menghadapi kendala, seperti fluktuasi cuaca, perubahan posisi matahari, serta suhu berlebih pada sel surya yang menurunkan efisiensi konversi energi. Mengatasi hal ini, penelitian ini mengusulkan sistem panel surya terapung berbasis Arduino Uno. Metode Logistic Regression (LR) digunakan untuk memprediksi performa konversi energi dengan mempertimbangkan intensitas cahaya matahari dan suhu sel surya. Hasil evaluasi menggunakan Metode Logistic Regression (LR) dalam memprediksi suhu permukaan sel surya menunjukkan akurasi model sebesar 93,3%. Berdasarkan confusion matrix, pengujian diperoleh data precision 97% dan recall 95% dengan f1-score 0,96. Nilai macro average (precision 0,88, recall 0,91, f1-score 0,90) dan weighted average (precision 0,94, recall 0,93, f1-score 0,93), model bekerja sangat baik meskipun masih perlu perbaikan dalam mengenali kelas minoritas. Kemudian data pengujian konversi energi pada grafik daya panel menunjukkan fluktuasi seiring perubahan intensitas cahaya antara pukul 11:00 hingga 13:00 WIB, dengan daya puncak 114,2 Wp. Secara keseluruhan, sistem terbukti mampu menghasilkan daya listrik stabil dan optimal, sehingga potensial diterapkan untuk mendukung penyediaan energi berkelanjutan. Solar energy is a renewable energy source with great potential in Indonesia due to its geographical location in tropical areas with high sunlight intensity throughout the year. However, conventional solar panels still face obstacles, such as weather fluctuations, changes in the sun's position, and excessive temperatures on solar cells that reduce energy conversion efficiency. To overcome this, this study proposes a floating solar panel system based on Arduino Uno. The Logistic Regression (LR) method is used to predict energy conversion performance by considering sunlight intensity and solar cell temperature. The evaluation results using the Logistic Regression (LR) method in predicting solar cell surface temperature show a model accuracy of 93.3%. Based on the confusion matrix, the test obtained 97% precision and 95% recall data with an f1-score of 0.96. The macro average value (precision 0.88, recall 0.91, f1-score 0.90) and weighted average (precision 0.94, recall 0.93, f1-score 0.93), the model works very well although it still needs improvement in recognizing minority classes. Energy conversion test data on the panel power graph then showed fluctuations with changes in light intensity between 11:00 AM and 1:00 PM WIB, with a peak power of 114.2 Wp. Overall, the system has proven capable of producing stable and optimal electrical power, making it potentially applicable to support sustainable energy provision.