Solar photovoltaic power generation prediction method based on TCN-LSTM

Abstract

The invention discloses a solar photovoltaic power generation prediction method based on TCN-LSTM. The solar photovoltaic power generation prediction method comprises the following steps: data preprocessing: processing photovoltaic data of one year into photovoltaic data of four seasons of spring, summer, autumn and winter, removing useless photovoltaic data, arranging and concluding key featuresaffecting photovoltaic power generation in a photovoltaic data set, and carrying out the normalization processing; building and predicting a prediction model: building a hybrid model based on a time convolutional neural network and a long short-term memory network, and training the prediction model by adopting the training set data; inputting the training sample into a TCN-LSTM model; carrying outfeature extraction by a TCN; inputting the time sequence data into a two-layer extended causal convolution, adding the time sequence data with a one-dimensional convolution output result to obtain anoutput result, and inputting the output result into an LSTM model; and performing information abstraction of high-level features on the features of the output result of the TCN by the LSTM model, processing the features into a one-dimensional vector and inputs the one-dimensional vector into a full connection layer of the LSTM model, and directly outputting a photovoltaic power generation power prediction value at the next moment by the full connection layer.

Description

technical field [0001] The invention relates to the technical field of computer forecasting, in particular to a TCN-LSTM-based solar photovoltaic power generation forecasting method. Background technique [0002] For time series prediction analysis, traditional methods mainly use statistical knowledge, mainly including: time series method, regression analysis method, Kalman filter method, and Markov chain method. In recent years, with the rapid development of intelligent algorithms, researchers have successfully applied intelligent algorithms to the study of time series, such as artificial neural network method, SVM support vector machine method, gray system prediction method, expert system method, etc. Several modern forecasting methods are described below. [0003] Artificial neural network is a calculation model modeled after the biological nervous system, simulating the process of brain processing information, suitable for time series prediction, especially with strong ...

AI Technical Summary

Problems solved by technology

[0018] In view of this, the present invention provides a solar photovoltaic power generation prediction method based on TCN-LSTM for the problem that the above-mentioned prior art cannot realize accurate prediction of photovoltaic power generation, and at the same time reduces the error of prediction

Images