System And Neural Net Based Methods For Signal Validation Of Power Plant Measurement Data

Abstract: An artificial neural network-based method for reconciliation of data in a process plant, in particular thermal power plant carried out by a data validation and reconciliation module (900) by input data processing and ANN-Data Reconciler (400), said method comprising the steps of receiving data from each sensor from a plurality of sensors by the input processing module (200) through at least one of a distributed control system (120) or a plant historian (130); validating the sensor data by assessing and segregating the process states between steady state and transient state; validating the sensor data by assessing and segregating with missing data and data without missing data; segregation of the missing data from a measurement vector and updating configuration of the thermal power plant based on determination of operational status of the thermal power plant, by the input processing module (200); predicting the missing data vector, by the plurality of auto-encoders contained in a missing value predictor (300), wherein each of the plurality of auto-encoders generate a predicted sub-vector of the missing data which is further convolved to generate a predicted full vector of the missing data; filtering, by the ANN-Data Reconciler (400), the predicted full vector of the data to generate reconciled data and further storing the reconciled data (700) in the reconciled data store (700); generating data based on equipment artifacts configured and forming a set of constraints to minimize the error by means of a data reconciliation module RNO (500); identifying faulty sensors based on the reconciled data.

Patent Information

Applicants

Inventors

Specification

Documents

Application Documents

Search Strategy

ERegister / Renewals Under Extension Period