Scholarlink

Energy supply planning is crucial in modern energy systems, especially with the integration of renewable sources like solar and wind, which introduce significant stochasticity into the production process. Distributed photovoltaic generation, due to its intermittent nature, further complicates grid stability. Advanced machine learning models, particularly Long Short-Term Memory (LSTM) networks, have been applied to address these challenges, improving forecasting accuracy for energy production and enabling better optimization and control of Smart Grid systems. Understanding how distributed photovoltaic energy impacts the grid is key to developing strategies for maintaining stability and integrating renewable sources effectively. The meteorological data for this study were sourced from the weather station at the Kamianka photovoltaic plant, which provided real-time measurements of key parameters. These data were then used to simulate the photovoltaic power output using a physical model of the photovoltaic system, taking into account the specific characteristics of the panels. This study focused on long-term forecasting of photovoltaic power generation for six months, using the LSTM model. The model demonstrated high accuracy, with an R2 value of 0.92 on the test set and 0.88 on the validation set, indicating a strong correlation between predicted and actual values. The RMSE for the test set was 1908.47 kW (6.12%), while for the validation set, it was slightly higher at 2107.35 kW (6.58%). The MAE was 1004.23 kW (3.24%) on the test set and 1152.89 kW (3.82%) on the validation set. These error values arise primarily due to the inherent stochasticity of photovoltaic energy, influenced by varying meteorological conditions such as cloud cover and sunlight intensity. The unpredictable nature of solar energy generation complicates the forecasting process and results in fluctuations in energy supply, challenging the ability to meet energy demand reliably. Consequently, this variability also affects energy demand, as grid operators must adjust consumption patterns or deploy backup generation to compensate for sudden dips or surges in photovoltaic output.

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

Long-Term Photovoltaic Power Generation Forecasting Using Long Short-Term Memory Network: Insights into Stochastic Dynamics

  • 章节

  • 发表日期：

    2026年

查看全文
  • Dmytro Matushkin,
  • Artur Zaporozhets,
  • Valentyna Stanytsina,
  • Volodymyr Artemchuk

摘要

Energy supply planning is crucial in modern energy systems, especially with the integration of renewable sources like solar and wind, which introduce significant stochasticity into the production process. Distributed photovoltaic generation, due to its intermittent nature, further complicates grid stability. Advanced machine learning models, particularly Long Short-Term Memory (LSTM) networks, have been applied to address these challenges, improving forecasting accuracy for energy production and enabling better optimization and control of Smart Grid systems. Understanding how distributed photovoltaic energy impacts the grid is key to developing strategies for maintaining stability and integrating renewable sources effectively. The meteorological data for this study were sourced from the weather station at the Kamianka photovoltaic plant, which provided real-time measurements of key parameters. These data were then used to simulate the photovoltaic power output using a physical model of the photovoltaic system, taking into account the specific characteristics of the panels. This study focused on long-term forecasting of photovoltaic power generation for six months, using the LSTM model. The model demonstrated high accuracy, with an R2 value of 0.92 on the test set and 0.88 on the validation set, indicating a strong correlation between predicted and actual values. The RMSE for the test set was 1908.47 kW (6.12%), while for the validation set, it was slightly higher at 2107.35 kW (6.58%). The MAE was 1004.23 kW (3.24%) on the test set and 1152.89 kW (3.82%) on the validation set. These error values arise primarily due to the inherent stochasticity of photovoltaic energy, influenced by varying meteorological conditions such as cloud cover and sunlight intensity. The unpredictable nature of solar energy generation complicates the forecasting process and results in fluctuations in energy supply, challenging the ability to meet energy demand reliably. Consequently, this variability also affects energy demand, as grid operators must adjust consumption patterns or deploy backup generation to compensate for sudden dips or surges in photovoltaic output.