研究目的
To develop an optimal load dispatch model for a grid-connected community microgrid by forecasting solar power and load using deep learning, considering uncertainties to promote supply-demand balance and reduce costs.
研究成果
The DRNN-LSTM model outperforms MLP and SVM in forecasting residential power load and PV power output, with MAPE of 7.43% for load forecasting. The optimal load dispatch model, considering uncertainties, shows that ESS and coordinated EV charging can reduce daily costs by 8.97%, shift peak load, and improve system reliability.
研究不足
The forecasting model and optimization model are separate; future work should explore dynamic mechanisms between them. The study uses hourly data; real-time optimization requires finer-grained data.
1：Experimental Design and Method Selection:
The study uses a deep recurrent neural network with long short-term memory units (DRNN-LSTM) for forecasting residential power load and PV power output. An optimal load dispatch model is established for a microgrid including residential load, PV arrays, EVs, and ESS under three scheduling scenarios. Particle swarm optimization (PSO) is used for optimization.
2：Sample Selection and Data Sources:
Residential power load data from 40 buildings in Austin, Texas, obtained from Dataport website (January 1, 2018 to February 1, 2018). PV power output data from Yulara PV plant in Australia (same period). Weather data from MesoWest website for KATT station, Austin.
3：8). PV power output data from Yulara PV plant in Australia (same period). Weather data from MesoWest website for KATT station, Austin.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: Software: Python with Keras API on TensorFlow backend for implementing DRNN-LSTM model. Data sources as mentioned.
4：Experimental Procedures and Operational Workflow:
Data preprocessing including min-max normalization. Training and testing of DRNN-LSTM model with hyperparameter optimization using Hyperopt. Forecasting results used as input for PSO-based load dispatch optimization under different scenarios.
5：Data Analysis Methods:
Evaluation metrics: RMSE, MAE, MAPE, PCC for forecasting performance. PSO algorithm for optimizing load dispatch costs.
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
