Risk-aware deep learning-driven extreme transmission capacity adjustment method

Abstract

The invention discloses a risk-aware deep learning-driven extreme transmission capacity adjustment method. The method comprises the steps of embedding an extreme transmission capacity predictor into an adjustment model to replace the most complex and time-consuming calculation part in the adjustment model to obtain a deep belief network agent-assisted double-layer model; constructing a predictioninterval based on a deep belief network, adjusting the prediction interval according to the coverage probability of the prediction interval, the normalized average bandwidth of the prediction intervaland the accumulated bandwidth deviation, and obtaining an optimal prediction interval through ensemble learning training; obtaining a probability R that the TTC value falls in the interval based on the trained prediction interval, and evaluating a risk probability of prevention and control failure according to the probability R; and introducing the probability R into a target function of a double-layer model assisted by a deep belief network agent, and adjusting the limit transmission capacity value. According to the invention, the balance between adjustment cost and control risk can be realized.

Description

technical field [0001] The present invention relates to, in particular, a risk-aware deep learning-driven limit transmission capacity adjustment method. Background technique [0002] The prevention and control of TTC requires not only rapid perception of TTC, but also the formulation of TTC preventive adjustment strategies within minutes. Some scholars proposed to use sensitivity technology to adjust the power flow state of the section, but the calculation of sensitivity is too rough and multi-prediction accident sets were not considered in this study. In general, there are few security economic studies on TTC preventive dispatch, so a new method is urgently needed to solve this problem. In recent years, some scholars have proposed a proxy-assisted (Surrogate-assisted, SA) optimization strategy. This strategy uses machine learning agent rules to replace the high-dimensional nonlinear or complex differential equation constraints in the optimization model, thereby greatly re...

AI Technical Summary

Problems solved by technology

If some studies propose to add a certain margin to the control model to make the operating point of the system farther away from the critical safety point, however, this scheme is difficult to ensure the economy of regulation, and will lead to a decrease in the convergence of optimization in practical applications; some scholars also propose Use ensemble learning to quantify the prediction error with probability. However, this scheme is only for classification decision trees, and the compatibility of machine learning methods and regression problems still needs to be studied.

Images