| State estimation is a very important part of the energy management system. In smart grid environment, the power system state estimation will experience the diversity of test and measurement equipment. The smart grid state estimation related to the traditional supervisory control and data acquisition and wide-area monitoring system based on Global Positioning System. This paper analyzes the characteristics of the two measurement systems data and use the technology of measurement transformation to standardize the measurement data. In this paper, the state estimation system research is mainly in data analysis method and the regression algorithm and data dimension reduction algorithm is applied to the state estimation of electric power system. Because the relevance vector machine algorithm is easy to obtain a sparse model and greatly reduce the calculation amount of the kernel function, this paper put forward state estimation model the based relevance vector machine. Due to the complexity of power grid, measurement system that generated the data dimension is very high. On the state estimation model based relevance vector machine the model of data dimensionality reduction is added. Autoencoder neural network can find the nonlinear low-dimensional structures embedded in high-dimensional data space and is suited to reduce the dimensionality of the measurement data. In this paper, autoencoder is used in the data dimensionality reduction module. On the platform of MATLAB, IEEE14node system is simulated, a lot of comparison indicate that module increased the status of data dimensionality reduction is good accuracy, better real-time and promotional. And smart grid state estimation system based browser/server is implemented. In the system design process, following the software engineering principle that is high cohesion and low coupling, using Java program language and using hibernate, spring, and Flex technology. The system has a good ease of use, maintainability and scalability. |
