Power System Transient Stability Assessment Based On Artificial Intelligence

The safe and stable operation of power system relates to the stable development of modern society.With the scale-up of electric power system,the access of various new energy sources,new loads and the wide application of power electronic devices have increased the uncertainty and complexity of power grid operation,which is facing a severe test.The power system is often disturbed.It is necessary to study the transient stability of power system in case of various large disturbances such as short circuit,disconnection or switch trip without fault.With the rapid development of computer technology,artificial intelligence algorithm has been applied in the field of transient stability assessment.Big data and machine learning methods are more and more used in the field of power system.The popularization of wide area measurement system and the application of big data technology in power system also provide a new idea for transient stability analysis.It is considered that there is a relationship between the transient stability and the operation characteristics of the system.The historical data of power grid operation are trained offline through the establishment of the model and the unknown functional relationships are analyzed and extracted.Then the real-time online operation data are collected,the structure and parameters of the model are updated continuously,which can be used to judge the transient stability of power system.According to the viewpoint of artificial intelligence and the process of transient stability assessment,a hierarchical method for transient stability assessment of large-scale power system based on mutual information theory and artificial intelligence algorithm is proposed in this paper.The main contents are as follows(1)Based on the mechanism of power system transient stability assessment,fault data in different operation states are collected to form the collection of transient stability preliminary screening features.A set of complementary dynamic stability features is extracted one by one as the core feature set by using the maximum-relevance minimum-redundancy algorithm(MRMR).Several extreme learning machines(ELMs)are trained according to the generated feature data sets.In the training process,a confidence rule is used to indicate the credibility of individual results,so the credibility of the overall assessment results can be greatly improved.(2)In the practical application of power system,the assessment speed requires high level.In order to balance the contradiction between assessment speed and accuracy,the hierarchical assessment structure is adopted in the final assessment process.In order to meet the requirements of large power system for assessment speed,the number of core features and the number of limit learning machines(ELM)are used in the first layer.For the sample points far away from the stable boundary,the assessment results are given,and the indeterminate samples are sent to the next layer.In the subsequent layer,more core features and elm numbers are added to repeat the assessment process of the previous layer.If very few samples cannot get the confidence assessment results in the final layer,in order to prevent missing judgment,they will be directly judged as unstable.(3)The IEEE-39 bus system is used as the benchmark system in this paper to verify the practicability of the model.In the actual 1648 bus system provided by PSS/E,the optimal parameters are determined through a series of parameter influence experiments.Based on the optimal parameters,the performance of the model is tested.The experimental results show that,compared with other traditional methods,the hierarchical method proposed in this paper can give more accurate results in a shorter time.As an effective method,it is suitable for on-line transient stability assessment.