As a special form of the Internet of Things, Smart Grid is an internet of both power and information, in which energy management is critical for making the best use of the power from renewable energy resources such as solar and wind, while efficient energy management is hinged upon precise forecasting of power generation from renewable energy resources. In this paper, we propose a novel least absolute shrinkage and selection operator (LASSO) and long short term memory (LSTM) integrated forecasting model for precise short-term prediction of solar intensity based on meteorological data. It is a fusion of a basic time series model, data clustering, a statistical model and machine learning. The proposed scheme first clusters data using k-means++. For each cluster, a distinctive forecasting model is then constructed by applying LSTM, which learns the non-linear relationships, and LASSO, which captures the linear relationship within the data. Simulation results with open-source datasets demonstrate the effectiveness and accuracy of the proposed model in short-term forecasting of solar intensity.

In this paper, a hierarchical online distributed algorithm (HODA) is developed to achieve optimal energy management in inter-connected Microgrids (IMG). The energy management objectives include maximizing users’ utility, optimizing the output power of controllable generators, and keeping the system operating in an economic manner. We formulate the problem as an online least absolute shrinkage and selection operator (LASSO) problem, considering both reactive power and system operation characteristics. We then employ Averaging Fixed Horizon Control (AFHC) to solve the formulated problem under some mild assumptions on the uncertainties in renewable power generation and load demand. The alternating direction method of multipliers (ADMM) is adopted to decouple the coupled constraints. The proposed online algorithm is asymptotically optimal, since its solution converges to the offline optimal solution. The performance of the proposed algorithm is validated using data traces obtained from a real-world IMG system.

Due to the existence of predicting errors in the power systems, such as solar power, wind power and load demand, the economic performance of power systems can be weakened accordingly. In this paper, we propose an adaptive solar power forecasting (ASPF) method for precise solar power forecasting, which captures the characteristics of forecasting errors and revises the predictions accordingly by combining data clustering, variable selection, and neural network. The proposed ASPF is thus quite general, and does not require any specific original forecasting method. We first propose the framework of ASPF, featuring the data identification and data updating. We then present the applied improved k-means clustering, the least angular regression algorithm, and BPNN, followed by the realization of ASPF, which is shown to improve as more data collected. Simulation results show the effectiveness of the proposed ASPF based on the trace-driven data.

The optimal power energy scheduling of a microgrid (MG) is not only related to the active power of distributed generators but also dependent on the reactive power and system operational constraints. It is essential to manage the active and reactive power simultaneously in optimal energy distribution. This paper is focused on developing an online algorithm to optimize the real-time power energy distribution in the MG, considering both reactive power and system operational constraints. The goal is to provide high quality electricity usage for users in the MG and maximize users' utility. The output power of controllable generators are also optimized. The proposed online algorithm is asymptotically optimal, since its solution converges to the offline optimal solution. The effectiveness of the proposed algorithm is validated using data traces obtained from a real-world MG.