Abstract: Improving the accuracy of photovoltaic power prediction is crucial for grid scheduling planning and is essential for the safe, stable, and economic operation of power systems. Based on the statistical characterization of the data, a two-stage PV power prediction model with error correction is developed. First, an Elman neural network model optimized by a small habitat genetic algorithm is introduced; subsequently, a more accurate model for the preliminary prediction error probability distribution is established, based on its distribution characteristics. This model aims to achieve error correction of the preliminary prediction results. The empirical results, derived from actual PV power curves and meteorological data, demonstrate the effectiveness of the proposed method.
Introduction: With the advancement of technology and reduction in costs, photovoltaic (PV) power generation has emerged as an important component of the renewable energy sector [1]. However, the power output from PV systems is significantly influenced by environmental factors, rendering its prediction highly uncertain and thereby impacting the stability and dispatch efficiency of the power grid. Current prediction methods comprise physical models [2], statistical analysis [3], and machine learning, with the latter gaining prominence in this field due to its ability to handle complex data and resist environmental disturbances [4]. The advantage of machine learning methods lies in their capability to achieve more reliable grid operation and efficient energy utilization through refined prediction methods. Strategies to improve the predictive performance of machine learning models include data preprocessing such as data cleaning and standardization [5], feature engineering [6], selecting appropriate models and optimizing hyperparameters [7, 8], and applying ensemble learning methods. For deep learning, this also includes adjusting network architecture, applying regularization techniques, and tuning the learning rate [9]. These strategies work together to enhance the model’s prediction accuracy and generalization ability. Despite significant research achievements in the field of PV power prediction, exploring new methods to further enhance predictive capability remains of great research value. This paper proposes a two-stage PV power prediction model aimed at further improving the accuracy of PV power generation predictions by combining advanced machine learning techniques with error correction mechanisms. Empirical analysis shows that this model brings an innovative research method to the field of PV power prediction, demonstrating its potential in improving prediction accuracy.
Adaptive Genetic Algorithm: Genetic algorithms derive their foundation from the core concepts of Darwin’s theory of evolution. They simulate real-world problems on a computer, mimicking the process of chromosome selection, crossover, and mutation in biological evolution, as illustrated in Figure 1.