Scholarlink

Overcoming most problems in PV, a monitoring system including data acquisition and data display was created in real-time, and a prediction model for PV power in the next few hours was developed. The highest value of efficiency is when the PV module is configured at a tilled angle of 30°. The input predictions are processed by the stored model. The model used variations of k-NN, k-NN- BPNN, and k-NN-D-BPNN. The model has a MAPE yield of 0.52% for k-NN, 0.95% for k-NN-BPNN and 33.47% for k-NN-D-BPNN, and MSE of 59.84 W2 for k-NN, 225.94 W2 for k-NN-BPNN and 17.701 W2 for k-NN-D-BPNN so that the model is a very good and feasible prediction. The resulting accuracy decreases when the prediction time is added. Therefore, predictions need to be limited to the next 3 h.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a Real Time Monitoring and Power Prediction System for Solar Power Plants Using Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ridho Hantoro,
  • Gunawan Nugroho,
  • Erna Septyaningrum,
  • Iwan Cony Setiadi,
  • Rasyid Yuniarto Kusuma,
  • Mochammad Arief Febrianto

摘要

Overcoming most problems in PV, a monitoring system including data acquisition and data display was created in real-time, and a prediction model for PV power in the next few hours was developed. The highest value of efficiency is when the PV module is configured at a tilled angle of 30°. The input predictions are processed by the stored model. The model used variations of k-NN, k-NN- BPNN, and k-NN-D-BPNN. The model has a MAPE yield of 0.52% for k-NN, 0.95% for k-NN-BPNN and 33.47% for k-NN-D-BPNN, and MSE of 59.84 W2 for k-NN, 225.94 W2 for k-NN-BPNN and 17.701 W2 for k-NN-D-BPNN so that the model is a very good and feasible prediction. The resulting accuracy decreases when the prediction time is added. Therefore, predictions need to be limited to the next 3 h.