Computational Intelligence in Smart Power Grid Management and Energy Feasibility Studies
| Today, big data is a significant matter! This seminar will discuss research in predictive modelling with artificial intelligence. It will describe briefly big data principles, with less technical detail but a greater focus on applications and result. In application space it will provide case studies in recent papers demonstrating the merits of advanced data analytic models in real-life, particularly in energy management systems. Models considered will include, but not limited to, deep learning, extreme learning machines（极限学习机器）, artificial neural network（人工神经网络）, support vector machines（支持向量机器）, multivariate adaptive regression spline （多元自适应回归样条）and M5 Tree, whereas model optimisation tools will include the results obtained by applying meta-heuristic feature selection （元启发式特征选择）(or ‘search’) algorithms, feature weight optimisation (or ‘add-in’) algorithms and multi-resolution tools such as empirical mode decomposition applied to model data to improve the prediction. In particular, feature selections are required to screen optimal inputs, improve the accuracy and reduce the computational burden, whereas add-in algorithms can help extract most, if not all of the predictive features from a large pool of carefully screened input variables. Empirical mode decompositions, can assist in identifying the frequency components in model inputs and addressing issues of non-stationarity, trends, jumps and periodicities present in model design data. This seminar will reveal the importance of ancillary tools in predictive modelling with applications of artificial intelligence models in energy demand management and solar energy simulations. The seminar will discuss and expect to exchange ideas and future challenges that we as, researchers, face in predictive modelling that must be considered in practical energy management models that are used in real-life simulations to design decision systems for energy management with big data analytics.