| 摘要: |
| 传统的神经网络算法在电价变化剧烈的情况下，精度较低并且所耗费的时间较长，难以满足电力市场发展的需求。为解决该问题，提出了一种基于回声状态网络(ESN)的短期电价预测方法。所提方法介绍了基于回声状态网络的预测原理，提出了电力市场短期电价的预测机制，包括参数选取、采样数据预处理和ESN训练及预测过程；并分别采用回声状态网络和反向传播算法(BP)神经网络进行短期电价预测。经过仿真验证，所提出的基于回声状态网络的电价预测具有较好的准确率和可行性。 |
| 关键词: 电力市场 电价预测 回声状态网络 储备池运算 |
| DOI：10.7667/PSPC160564 |
|
| 基金项目:国家863项目基金资助(2012AA050804) |
|
| Echo-state-network based electricity price forecasting in electric power market |
| REN Yuan |
| (State Grid Shanxi Electric Power Company, Taiyuan 030002, China) |
| Abstract: |
| Traditional neural network based electricity price forecasting algorithm fails to meet current demands by future electric power market, with low accuracy and long computation time when the electric power price changes greatly. Using the method based on Echo-State-Network (ESN), an electricity power price short-term forecasting approach is proposed. Firstly, the principle of ESN is introduced and discussed. On this basis, the electricity power price short-term forecasting approach is proposed, including parameter selection, sampling data pre-processing and ESN training and forecast process. Then, the short-term electricity price forecasting is performed by ESN and BP neural network. The simulation results show that using ESN the short-term electricity price can be forecasted more quickly and steadily. This work is supported by National High-tech R & D Program of China (No. 2012AA050804). |
| Key words: electricity market electricity price forecasting echo state network reservoir computation |
