Model data hybrid driven power grid reliability rapid calculation method and device

Abstract

The invention discloses a model data hybrid driven power grid reliability rapid calculation method and a device, and the method comprises the steps: calculating a reliability index of each state through optimal power flow modeling in a pre-sampling stage, and providing a label and a data sample for improved SDAE network training; in an actual sampling stage, adopting a trained improved SDAE network for carrying out direct mapping from a system operation state to a reliability index, and improving a noise adding link of an improved SDAE network model, so that the improved SDAE network model adapts to change fluctuation of power flow data of a power system, and data driving based on deep learning is formed; mining the power flow characteristics in the system through a deep stacked SDAE neural network, establishing the mapping relation between the system operation state parameters and the minimum load reduction, so that optimal power flow calculation with the minimum load reduction amount as the target is achieved. The device comprises a processor and a memory. According to the method, the solving time and precision meet online application requirements.

Description

technical field [0001] The present invention relates to the field of electric power systems, in particular to a method and device for fast calculation of power grid reliability by means of a model-data hybrid driving method for massive random scenarios generated by uncertain source load levels. The calculation of reliability aims to measure the ability of the power system to continuously supply power. It is a key link in the planning of the power system and provides design reference for the planning of the power system. Background technique [0002] High penetration of renewable energy in the electricity system mitigates carbon emissions. On the other hand, it inevitably brings massive system states that require additional analysis to reliability calculations, and data-driven research methods emerge as the times require. However, for the power grid reliability calculation task, there is a lack of corresponding historical data as support in the data-driven model training sta...

AI Technical Summary

Problems solved by technology

The traditional method has its inherent limitations in the first two processes: 1. The proportion of failure states that play a leading role in the calculation of indicators in the state space is low; 2. The calculation of the state analysis process is cumbersome

The second problem stems from the iterative method to solve the optimal power flow (OPF) algorithm to obtain the minimum load shedding for a given state, which is computationally intensive

[0008] Most of the above studies focus on classification tasks, while the great potential of data-driven methods in reliability calculations is still untapped, but data-driven applications are limited in many ways

First, the machine learning method requires prior knowledge and hyperparameter settings to severely limit the final performance of the algorithm, and manual screening and processing of input features may cause fallacies or manual omissions caused by experience

Second, the use of pure data-driven in the field of reliability will face the problem of insufficient historical data. Reliability calculation is aimed at power systems that are still in the planning stage or expansion planning stage, and there is a lack of historical data of actual operation to provide training samples.

Images