Short-Term Demand Forecasting Method in Power Markets Based on the KSVM-TCN-GBRT.

With the consumption of new energy and the variability of user activity, accurate and fast demand forecasting plays a crucial role in modern power markets. This paper considers the correlation between temperature, wind speed, and real-time electricity demand and proposes a novel method for forecasting short-term demand in the power market. Kernel Support Vector Machine is first used to classify real-time demand in combination with temperature and wind speed, and then the temporal convolutional network (TCN) is used to extract the temporal relationships and implied information of day-ahead demand. Finally, the Gradient Boosting Regression Tree is used to forecast daily and weekly real-time demand based on electrical, meteorological, and data characteristics. The validity of the method was verified using a dataset from the ISO-NE (New England Electricity Market). Comparative experiments with existing methods showed that the method could provide more accurate demand forecasting results.

Deep Reinforcement Learning-Based Trading Strategy for Load Aggregators on Price-Responsive Demand.

With the development of the Internet of things and smart grid technologies, modern electricity markets seamlessly connect demand response to the spot market through price-responsive loads, in which the trading strategy of load aggregators plays a crucial role in profit capture. In this study, we propose a deep reinforcement learning-based strategy for purchasing and selling electricity based on real-time electricity prices and real-time demand data in the spot market, which maximizes the revenue of load aggregators. The deep deterministic policy gradient (DDPG) is applied through a bidirectional long- and short-term memory (BiLSTM) network to extract the market state features that are used to make trading decisions. The effectiveness of the method is validated using datasets from the New England electricity market and Australian electricity market by introducing a bidirectional LSTM structure into the actor-critic network structure to learn hidden states in partially observable Markov states through memory inference. Comparative experiments of the method show that the method can provide greater yield results.