Ultra short term PV power forecast based on scarab algorithm – pv magazine International

30 October 2024
Lior Kahana

scientists from china Hubei University of Technology They proposed a new deep learning model for ultra-short-term PV power forecasting.

The new technique combines personal attention temporal convolutional networks (SATCN), bidirectional long short-term memory networks (BiLSTM) and scarab optimization (DBO) and is therefore called DBO-SATCN-BiLSTM.

“Current photovoltaic energy production prediction methods have some shortcomings. Meta-heuristic learning methods such as neural networks are complex and time-consuming to train and often rely on empirical parameter tuning, which cannot achieve the best results.” explained the group. “Swarm intelligence algorithms are expected to improve optimal parameters, but currently used optimization algorithms still have certain limitations. For example, some optimization algorithms are relatively simple and easy to fall into local optimality. To address the above issues, this paper proposes a DBO-SATCN-BiLSTM.”

SATCN includes a specialized convolutional neural network to analyze time series data to reduce gradient vanishing or bursting. The proposed model also uses the self-attention mechanism to capture dependencies between different time steps in time series data. However, the BiLSTM layer uses bidirectional data processing capabilities to achieve forward and reverse data exchange features in the array data.

The DBO part of the model operates at the external level by defining and optimizing the hyperparameters of the hybrid SATCN-BiLSTM model. Unlike parameters learned during training, hyperparameters are external configurations that direct how a model works. The model takes its name from dung beetles because it mimics the way they roll dung balls into their nests; It uses scent and the position of the sun or moon to navigate. In this way, optimization searches for optimal solutions based on some external guidance.

The model consists of an input layer, three layers of SATCN residual blocks, two layers of BiLSTM, and one fully connected layer. The SATCN structure enables the extraction of temporal features from photovoltaic power data, while BiLSTM further captures the temporal correlation between forward and backward features.

The model was trained and tested using annual information from a 30 MW PV power plant in Shaanxi Province, China. Using parameters such as temperature, humidity, barometric pressure and solar radiation every 15 minutes throughout this year, the model was asked to predict the power in the next 15-minute step. He was also asked to make a 45-minute multi-pitch prediction that meant three points in the future.

The results were compared with those of seven reference models: Convolutional neural network (CNN); BiLSTM alone; temporal convolutional network (TCN); Hybrid of CNN-BiLSTM; TCN-BiLSTM, SATCN-BiLSTM hybrid with particle swarm optimization (PSO-SATCN-BiLSTM); and SATCN-BiLSTM (SSA-SATCN-BiLSTM) with salp swarm algorithm.

In one-shot estimation, DBO-SATCN-BiLSTM was found to achieve root mean square error (RMSE) of 0.357 when predicting one-year PV power.

“This indicates a significant reduction of 33.1%, 23.4% and 18.1% compared to CNN, TCN and BiLSTM,” the academics emphasized. “CNN-BiLSTM outperforms TCN-BiLSTM, PSOSATCN-BiLSTM and SSA-SATCN-BiLSTM with margins of 17.5%, 10%, 2.9% and 2.4%,” the group said. “In terms of multi-step (3-step) forecasting, errors increase for all models as the forecast range increases. DBO-SATCN-BiLSTM is 0.437, indicating significant improvements of 52.3%, 32.4%, 32.9%, 31.5%, 31.1%, 9.5% and 4.7% compared to the other seven models. It stands out with its RMSE of .

Their findings were presented as follows:Hybrid deep learning model based on scarab optimization algorithm for ultra-short-term PV power forecasting.published in iScience.

Popular content