Selecting decision trees for power system security assessment

Power systems transport an increasing amount of electricity,and in the future,involve more distributed renewables and dynamic interactions of the equipment.The system response to disturbances must be secure and predictable to avoid power blackouts.The system response can be simulated in the time domain.However,this dynamic security assessment(DSA)is not computationally tractable in real-time.Particularly promising is to train decision trees(DTs)from machine learning as interpretable classifiers to predict whether the systemwide responses to disturbances are secure.In most research,selecting the best DT model focuses on predictive accuracy.However,it is insufficient to focus solely on predictive accuracy.Missed alarms and false alarms have drastically different costs,and as security assessment is a critical task,interpretability is crucial for operators.In this work,the multiple objectives of interpretability,varying costs,and accuracies are considered for DT model selection.We propose a rigorous workflow to select the best classifier.In addition,we present two graphical approaches for visual inspection to illustrate the selection sensitivity to probability and impacts of disturbances.We propose cost curves to inspect selection combining all three objectives for the first time.Case studies on the IEEE 68 bus system and the French system show that the proposed approach allows for better DT-selections,with an 80%increase in interpretability,5%reduction in expected operating cost,while making almost zero accuracy compromises.The proposed approach scales well with larger systems and can be used for models beyond DTs.Hence,this work provides insights into criteria for model selection in a promising application for methods from artificial intelligence(AI).

PREDICTIVE STUDY ON THE ECONOMIC VALUE OF RADIO SPECTRUM IN CHINA

Increasing popularity of spectrum-based services brings the striking contradictions between the limited spectrum resource and its increasing demands.This paper puts forward an approach to forecast the future spectrum demand and its economic value,so as to offer a scientific basis for spectrum regulators to resolve this contradiction effectively and make a long-term spectrum-use plan.Specifically,this paper analyzes the driving factors of spectrum demand firstly,based on which a forecasting model is constructed to predict the spectrum demand and its deficit/surplus in the next few years.Then,a forecasting model to measure the economic value of spectrum is proposed based on marginal opportunity cost theory,and the indifference curve is introduced to show the economic value generated by additional spectrum.Additionally,an empirical study is conducted to forecast the spectrum demand and its economic value for China in the next 10 years according to the proposed method.The results of this study show that spectrum deficit is a trend in future and releasing additional spectrum will bring China huge economic benefits.