Short-Term Photovoltaic Power Interval Prediction Based on VMD and GOA-KELM Algorithms

Affected by environmental factors such as irradiance and atmospheric temperature, the output power of photovoltaic power generation shows a high degree of randomness, intermittency and fluctuation. When large-scale photovoltaic energy is connected to the power grid, it will pose a challenge to the security and stability of the power grid. To solve the problem that the prediction accuracy of current photovoltaic power time series is not high and there is no reference interval for the prediction results, a photovoltaic short-term power interval prediction model based on variational mode decomposition (VMD) and improved grasshopper algorithm (GOA)-Kernel Extreme Learning Machine (KELM) is presented. The prediction model first uses VMD algorithm to decompose the original data and get several subsequences with lower complexity; then uses grey correlation method to select the subsequences with higher degree of association with the original data and divide them into low-frequency, medium-frequency and high-frequency sequences; Then establishes a kernel limit learning machine prediction model optimized by grasshopper algorithm, and makes the subsequences with higher degree of association from the original sequence. The training of mid-frequency and high-frequency sequences obtains the set of prediction errors from the predicted and actual values and the model. Finally, the density curve of error function is calculated using the kernel density estimation method (KDE). Based on this curve, the prediction intervals with 85% and 90% confidence are established to realize the interval prediction of photovoltaic power