A method for remote online detection of measurement errors based on reinforcement learning algorithm

Abstract

The invention belongs to the technical field of substation metering, and relates to a metering error remote online detection method of a reinforcement learning algorithm. In the error detection and evaluation process, error comparison is carried out by utilizing a metered power acquisition sample and a standard ammeter value, and a standardized error expert database value is optimized; a Q-Learning function reinforcement learning algorithm is used to construct a mapping model of electric quantity imbalance; the numerical value of the expert database is updated to guide the initial value definition of the Q iterative network; the electric quantity value of the next moment is accurately predicted through the change process of the electric quantity data collected in real time, the electric quantity value prediction function is achieved, the average prediction error value is obtained by comparing and calculating the electric quantity value with the standard value in the expert database, and when the error range exceeds the error confidence interval, early warning information is sent out immediately. The method has the advantages of being capable of remotely testing the overall error ofthe metering system, high in self-adaptive adjustment capacity, accurate in evaluation conclusion and easy to implement, and meeting the dynamic characteristics of error generation.

Description

technical field [0001] The invention relates to a remote on-line detection method of a measurement error based on a reinforcement learning algorithm, and belongs to the field of detection and evaluation of the operating state of electric power measurement equipment. Background technique [0002] The smart substation metering system belongs to the category of high-voltage metering. The accuracy of the metering directly affects the settlement results of the electricity bill of the large power grid. The error margin is higher than that of the household electric energy metering system by a certain percentage, and the resulting power error is also considerable. Therefore, corresponding detection methods should be adopted for substation measurement errors, so as to detect and compensate the operation errors in time. [0003] The smart substation metering system will involve different metering devices, including current / voltage transformers, power characteristic acquisition modules...

AI Technical Summary

Problems solved by technology

This method has two disadvantages. One is that real-time error adjustment cannot be realized, and it will lead to power settlement errors and fund loss during the long time interval that causes errors.

If the error caused by the equipment itself is not maintained in time, it may lead to infinite amplification of the impact of the error; second, in the process of information collection, a large amount of measurement data will be generated to be stored and transmitted, and optical fiber resources are required to transmit the data to the measurement management background , on-site optical fiber resources are scarce, and it is not easy to lay, and a lot of construction and maintenance costs are invested to transmit metering data

A storage server needs to be deployed in the background to provide redundant storage space for on-site data. On the whole, the investment cost is very low, and it does not meet the needs of later expansion and construction of substations.

Images