Photovoltaic power interval prediction method combining neural network and parameter estimation

Abstract

The invention discloses a photovoltaic power interval prediction method combining a deep cycle neural network and parameter estimation, belonging to the technical field of photovoltaic power prediction. The method of photovoltaic power forecasting based on a long-term and short-term memory network firstly chooses the data of product day, ambient temperature, ambient humidity, wind speed and solarirradiance as the original data of photovoltaic power forecasting. The data of product day, ambient temperature, ambient humidity, wind speed and solar irradiance are selected as the original data ofphotovoltaic power forecasting. The confidence intervals of PV power values and predicted values corresponding to 24 hourly hours of the predicted day are outputted from the predicted model to 24 hourly hours of the predicted day for 365 days of the year. This method establishes a relationship between the current photovoltaic power change and the previous photovoltaic power change, realizes the dynamic modeling of the time series data, and can reflect the change law of photovoltaic power more fully, and realizes more accurate photovoltaic power prediction. The method is easy to operate, is high in practicability and has a high promotion value.

Description

technical field [0001] The invention belongs to the technical field of photovoltaic power prediction technology, and in particular relates to a photovoltaic power interval prediction method combined with a deep cyclic neural network and parameter estimation. Background technique [0002] Photovoltaic system power generation is fluctuating and periodic due to the influence of external environmental factors such as weather conditions, day and night alternation, and seasonal changes. The large-scale connection of photovoltaic power generation systems to the power grid will bring greater impact on the safe and stable operation of the power system. , Periodic shocks. Accurate photovoltaic power prediction is the premise to ensure the safe and stable operation of photovoltaic grid-connected power generation, and it is also an important basis for the reasonable allocation and scheduling of photovoltaic system power generation. At present, the forecasting of photovoltaic power is m...

AI Technical Summary

Problems solved by technology

Although the quantile regression method avoids the assumption of the error distribution function, it still needs to build a complex error prediction model, and there is a high-dimensional nonlinear relationship between the variables. The error of the model itself will increase the error of the result and cannot reflect the prediction situation.

The Bootstrap method is a resampling statistical method, and there is no need to make assumptions about the distribution, but this method only uses the original sample data to replicate the observation information for statistical inference, and the accuracy of parameter estimation is not high enough.

Moreover, due to a large number of repeated sampling, the amount of calculation is relatively large compared to the method of parameter estimation.

Although the extreme learning machine runs efficiently, because it only has a single-layer structure, the output of the model is prone to random fluctuations, and the stability and generalization ability are not strong.

Images