研究目的
To present an effective model for day-ahead forecasting PV power output of a plant based on deep belief network (DBN) combined with grey theory-based data preprocessor (GT-DBN).
研究成果
The proposed GT-DBN model provides superior forecast accuracy compared to other models, especially when the PV output varies more drastically. The model's integration of grey data preprocessor and DBN with a supervised learning scheme enhances its effectiveness for time series forecast with high accuracy.
研究不足
The study focuses on day-ahead forecasting and may not be directly applicable to longer-term forecasts. The model's performance is evaluated based on data from a single PV power plant, which may limit its generalizability to other locations or conditions.
1：Experimental Design and Method Selection:
The study employs a DBN combined with a grey theory-based data preprocessor for forecasting PV power output. The DBN learns high-level abstractions in historical PV output data using hierarchical architectures.
2：Sample Selection and Data Sources:
The study uses actual measured PV power output data from a 5-MW PV power plant in central Taiwan over 12 months in 2016, with a look-ahead time up to 24-hour at one-minute resolution.
3：List of Experimental Equipment and Materials:
The study utilizes a PC with Intel Core i7-2600 3.4-GHz processor and developed Matlab codes for simulations.
4：4-GHz processor and developed Matlab codes for simulations.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: The proposed model is trained with one day of measured PV power output data and validated with another day's data. The model's performance is compared with ARIMA, BPNN, RBFNN, SVR, and DBN alone.
5：Data Analysis Methods:
The forecast accuracy is evaluated using mean absolute percentage error (MAPE) and root mean squared percentage error (RMSPE).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
