Abstract:The intermittence and fluctuation of photovoltaic power generation seriously affect output power reliability, efficiency, fault detection of photovoltaic power grid, etc. The precise forecasting of photovoltaic power generation is the critical method to solve the above limitations. Current photovoltaic power generation forecasting methods generally usually adopt meteorological data and historical continuous photovoltaic power generation as inputs, but they do not take into account historical periodic photovoltaic power generation as inputs, which makes the existing methods inadequate in learning time correlation. Therefore, to further study the time correlation for improving the prediction accuracy, an LSTM-FC deep learning algorithm composed of long-term short-term memory (LSTM) and fully connected (FC) layers is proposed. The double-branch input of the model enables it not only to consider the impact of meteorological data on power generation but also to consider time continuity and periodic dependence, thereby improving the prediction accuracy to a certain extent. In this paper, meteorological data, historical continuous data, and historical periodic data are used as experimental data, and these three types of data are combined into different input forms to evaluate and compare LSTM-FC with other baseline models, including support vector machines (SVM), gradient boosting decision tree (GBDT), generalized regression neural network (GRNN), feedforward neural network (FFNN), and LSTM. The simulation results show that the accuracy of the models with meteorological data, continuous data, and periodic data as input is higher than that of other input forms, and the accuracy of LSTM-FC is the highest among these models, and its root mean square error (RMSE) is 11.79% higher than that of SVM.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the adverse effects of the intermittency and volatility of photovoltaic (PV) power generation on output power reliability, efficiency, and fault detection, etc. Specifically, accurately predicting the amount of photovoltaic power generation is a key method to address the above - mentioned limitations. Current photovoltaic power generation prediction methods usually use meteorological data and historical continuous photovoltaic power generation data as inputs, but do not take into account the influence of historical periodic photovoltaic power generation data, which makes existing methods deficient in learning time - relatedness. To solve these problems, the author proposes a deep - learning algorithm (LSTM - FC) based on long - short - term memory network (LSTM) and fully - connected layer (FC) to further study time - relatedness and improve prediction accuracy. The dual - branch input of this model not only considers the influence of meteorological data on power generation, but also takes into account time continuity and periodic dependence, thereby improving the prediction accuracy to a certain extent. The following are some key points involved in the paper: 1. **Problem Background**: - The intermittency and volatility of photovoltaic power generation seriously affect the reliability and efficiency of output power. - Accurately predicting the amount of photovoltaic power generation is an important means to improve the operating efficiency and reliability of the system. 2. **Limitations of Existing Methods**: - Existing prediction methods mainly rely on meteorological data and historical continuous data, ignoring the time - relatedness of periodic data. 3. **Proposed Solution**: - A deep - learning algorithm (LSTM - FC) based on LSTM and fully - connected layer (FC) is proposed. - The dual - branch input structure of this model can simultaneously consider meteorological data, historical continuous data, and historical periodic data, thereby better capturing time - relatedness. 4. **Experimental Results**: - The experimental results show that when using meteorological data, continuous data, and periodic data as inputs, the LSTM - FC model has the highest prediction accuracy, and its root - mean - square error (RMSE) is 11.79% lower than that of the support vector machine (SVM). Through these improvements, the LSTM - FC model can more accurately predict the amount of photovoltaic power generation under more complex conditions, thereby helping to improve the operating efficiency and reliability of the photovoltaic system and providing decision - support for power dispatching.

A Short-Term Photovoltaic Power Generation Forecast Method Based on LSTM

Ultra-short-term PV power forecasting based on LSTM with PeepHoles connections

SSA-LSTM: Short-Term Photovoltaic Power Prediction Based on Feature Matching

Photovoltaic Power Generation Forecasting with Hidden Markov Model and Long Short-Term Memory in MISO and SISO Configurations

A hybrid machine learning forecasting model for photovoltaic power

A Photovoltaic Power Prediction Approach Based on Data Decomposition and Stacked Deep Learning Model

Ultra-Short-Term Power Prediction of a Photovoltaic Power Station Based on the VMD-CEEMDAN-LSTM Model

Prediction of Solar Photovoltaic Power Generation Based on MLP and LSTM Neural Networks

Short-term photovoltaic power prediction based on modal reconstruction and hybrid deep learning model

PV power interval prediction based on EEMD-LSTM method

Short-term probabilistic photovoltaic power forecast based on deep convolutional long short-term memory network and kernel density estimation

Short-Term Prediction of PV Power Based on Combined Modal Decomposition and NARX-LSTM-LightGBM

Prediction method of photovoltaic power based on combination of CEEMDAN-SSA-DBN and LSTM

Short-term photovoltaic power forecasting based on multiple mode decomposition and parallel bidirectional long short term combined with convolutional neural networks

Comparative study of machine learning approaches for predicting short-term photovoltaic power output based on weather type classification

A hybrid model of CNN and LSTM autoencoder-based short-term PV power generation forecasting

Photovoltaic power prediction based on hybrid modeling of neural network and stochastic differential equation

A Combined Method of Two-model Based on Forecasting Meteorological Data for Photovoltaic Power Generation Forecasting

Improved multistep ahead photovoltaic power prediction model based on LSTM and self-attention with weather forecast data

Time series forecasting for hourly photovoltaic power using conditional generative adversarial network and Bi-LSTM