FIELD OF INVENTION:-
The present invention relates to data reconciliation and more particularly to systems and methods for data reconciliation, missing value prediction and parameter estimation considering both random and gross errors using non-linear optimization.
BACKGROUND OF THE INVENTION: -
In power plants, generally field data is measured using instrumentation devices which are coupled to one or more centralized or decentralized control system. Field devices may be sensors (temperature, pressure flow rate, voltage, current, power, etc.), switches, levels, actuator positioners, transmitters, etc. The control system receives data in form of analog or digital signals pertaining to the field devices. This data is used for process monitoring, process control, process, optimization and diagnostics which are predominantly based on process models.
The measurement values, however, are either erroneous or missing. The p rocess control, monitoring and optimization based on this erroneous data is unlikely to lead to efficient handling of the plant. Mass flow sensors in a power plant are fewer in quantity as compared to temperature and pressure sensors, and are unreliable due to turbulent flow at higher velocities. This may cause erroneous process control feedback. Data validation and reconciliation thus is significant for process models based applications.
Data reconciliation through classical approach involves describing the process in terms of functions viz. mass, energy and momentum balances expressed in terms of process variables and process characteristics. These functions are used to test the accuracy of measured values and further iteratively improve the measured values. Data reconciliation can be attempted in different ways for example some approaches involve reconciling the mass flows initially and use them for reconciling pressures and temperatures. The difference between expected function values and the ones obtained using measured values of process variables are

called the residuals. The process variables are adjusted to minimize these residual terms resulting in reconciled values for measured process variables which give a more accurate representation of the actual process.
Australian Patent AU2909001by Aspen Technology describes a method for detection identification and classification of faulty sensors using first principle based process models. The identification of faulty sensors is done using statistical tests.
US patent US8892382B2 by General Electric Company describes a system and method for condition-based power sensor calibration where random errors in sensor measurements are reconciled based on sensor calibration history and trends of sensor. It includes an anomaly detection by comparison of performance model and actual power plant performance which may be due to equipment degradation or sensor issue.
US patent US20140316754A1 by Invensys Systems Inc. describes a method for monitoring performance of a process and the condition of equipment units by performing data reconciliation using mass and energy balance and tracking the maintenance parameters of individuals units. This method focuses on process monitoring and does not provide estimates for unmeasured process variables. The presence of faulty sensors is done by determining the variance of each sensor and whether the adjustment of a variable goes beyond its variance.
US patent US8271104B2 by Tendrup Consult & Associates BV describes a method for data reconciliation and parameter estimation by minimizing the weighted sum of squared model error plus the weighted sum of squared deviations from the measured flows and qualities. This amounts to an unconstrained optimization. Considering the lesser availability of mass flow sensors and sufficient availability of

pressure sensors in power plants, pressure flow model has been included in the
embodiment of the invention, which adds to the observability of involved process
variables. The invention presented here specifically caters to power plants.
The embodiment of the invention includes a Gross Error Detection (GED) module which detects and identifies faulty sensors on account of complete failure, sensor bias, sensor drift or precision degradation. It further estimates the value indicated by the faulty sensors. This is accomplished by statistical tests to initially detect presence of gross error and then identification of one or more faulty sensors. This process is repeated at every instance of data reconciliation making detection of faulty sensors more efficient.
OBJECT OF THE INVENTION:-
An object of the invention is to provide a method and system for validating and reconciling measured process variables considering both random and gross errors capable of handling sensor redundancy and dynamic process configuration, dynamically arriving at redundancy and observability of the process and trending the health of the sensors in power plants at steady state.
Another object of the invention is to perform non-liner constrained optimization.
Another object of the invention is to provide steady state data reconciliation of measured values in presence of both random and gross errors.
Another object of the invention is to simultaneously estimate the unmeasured variables for overspecified as well as underspecified systems.
Yet another object of the invention is handling redundant sensors as well as dynamically changing flow paths, variation in process inputs , etc.

A still further object of the invention is to dynamically arrive at both overall redundancy and observability of the process.
SUMMARY OF THE INVENTION:-
A method for validating and reconciling measured process variables and estimating the unmeasured process variables simultaneously, detecting, identifying and estimating for faulty sensors in power plants at steady state. A system including a computer processor along with a storage device capable of storing field data from power plant as also computer-executable set of instructions suited to the following operations: configuring the process using steady state process models, receive field data from instrument devices from a central control system or historian, validating the raw data by preprocessing, configuring the process dynamically based on equipment status, arriving at redundancy and observability of the process, reconciling the measured process variables using constrained non-linear optimization, estimating the unmeasured process variables, detecting presence of faulty sensors, identifying faulty sensors and estimating for variables indicated by them, trending health of the sensors.
BRIEF DESCRIPTION OF THE ACCOMPANYING DRAWINGS:
It is to be noted, however, that the appended drawings illustrate only typical embodiments of the present subject matter and are therefore not to be considered for limiting of its scope, for the invention may admit to other equally effective embodiments. The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The same numbers are used throughout the figures to reference like features and components. Some embodiments of system or methods in accordance with

embodiments of the present subject matter are now described, by way of example, and with reference to the accompanying figures, in which:
Fig. 1 illustratively depicts the network architecture where an embodiment of the invention may be incorporated.
Fig. 2 illustratively depicts the modules which constitute an embodiment of the invention.
Fig. 3 is a flow diagram illustrating details of one method for implementing data reconciliation and gross error detection, as per an embodiment of the invention.
Fig. 4 illustratively represents a process configuration with equipment model and process streams.
Fig. 5 shows a mathematical form for data reconciliation in accordance with an embodiment of the present invention.
Fig. 6 is a flow diagram which illustrates a method by which gross errors may be detected and identified in accordance with an embodiment of the present invention.
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION:-
Illustrative embodiments of the invention shall now be described in detail with reference to the accompanying drawings. The invention may be embodied in

different forms and is not limited to the embodiments described here. Illustrative embodiments of the invention are directed to systems and methods for data reconciliation with gross error detection in power plant. These may be aimed at configuring process using steady sate process models, receiving field data from sensors from a central control system, validating the raw data in pre-processing by range check, updating process configuration based on equipment status, reconciling the measured process variables, estimating the unmeasured process variables, detect gross errors and identify faulty sensors.
Fig. 1 illustratively depicts the network architecture where an embodiment of the invention may be incorporated. The signals from the sensors in the power plant (171) are sent to the Distributed Control System (181). The measured data is acquired by data reconciliation system (101) from (181) at every instance to meet the real time requirement. This raw data is stored in database (161). This is achieved by implementing communication protocols for data stream (191) which may be Modbus, TCP/IP over Ethernet, remote desktop or wireless. The validated and reconciled data from Data Reconciliation (100) is stored in database (162). This post reconciliation data is then streamed to third party host (184) over data stream which may be (194) Modbus, TCP/IP over Ethernet, remote desktop or wireless. This may be a remote or local Human Machine Interface. Similarly data may be sent to process controller (182) and process optimizer (183) over data streams (192) and (193) respectively.
The invention has process configuration module (120) which uses process models to configure the power plant. In a power plant there exist multiple equipments such as turbine, heat exchanger, pump, to carry out specific process functionality. The models for equipments (111) have been specifically developed in view of the available measurements. These are interconnected in (120) by process stream models (112) which together represent the process. The measurement sensors for temperature, pressure, mass flow, status, level, current, etc are associated with the equipments and streams (120) generates the process configurations dynamically instance by instance.

The preprocessing module (130) has four subsystems (131) to (134). Initially it is checked if the power plant is in steady state by tracking the change in generated load over time in (131). If the plant is not in a transient state, further processing is done else reconciliation is not performed. Range check is done in (132) to validate the measured data and the out of range values are discarded from the reconciliation process which shall be estimated instead. The ranges are placed not based on the sensor limits, rather they are intelligently put according to the process knowledge. The equipment status is checked in (133) based on which (120) dynamically configures the process (134) arrives at the redundancy and observability of the instantaneous process.
Data reconciliation is performed at every instance by non-linear constrained optimization after dynamic process configuration based on set of constraint equations from (120). The error in the measurements maybe restricted to random error where the measured data has inconsistencies of small magnitudes, or the instrument itself may be faulty. The fault in the instrument may be due to complete failure, sensor bias, sensor drift or precision degradation. Both these can be handled by the Data Reconciliation System (100). The random errors are handled by the reconciliation module (140) and the presence of faulty sensors is handled by the Gross Error Detection (GED) Module (150).
Pig. 3 shows a flow diagram illustrating details of one method for implementing data reconciliation and gross error detection. Acquired measured data initially flows through the pre-processing module where it is validated by range check. The process configuration is dynamically updated based on the equipment status. The reconciliation module reconciles the error in measurement data by minimizing the weighted sum of squares of the adjustment made for reconciliation subject to set of constraint equations based on configuration from the process configuration.
Fig. 4 illustratively represents a process configuration with equipment model and process streams. The figure shows an input stream (421) being split into two

stream (422) and (423) at a node (411). The streams could be material streams as is the case more often or could be representative of transfer of thermodynamic properties such as energy. The node (411) could represent multiple process equipments with a single input stream and two output streams and the constraint equation pertaining to this model will be formed accordingly . (431) to (433) are the measurements on the process streams. Taking the case of a simple flow splitter where (431) to (433) represent the mass flow measurements, conservation of mass becomes one constraint which the process follows i.e. x1-(x2 + x2) =0.
Fig. 5 illustratively describes mathematical form for data reconciliation. The measurement vector given in (510) comprises of the measured process variables x1, X2, ---, X3. It may also include certain model parameters such as heat transfer coefficient. Equation 520 shows the objective function to be minimized, where x1 represent the actual measured values, xi and ai are the corresponding estimated values and the standard deviation. The objective is to minimize the sum of squares of the adjustment made weighted by the standard deviation or any suitable weight which accounts for the accuracy of the sensors. Optimization is performed by applying hard constraints which may comprise of one or more of the following: equality constraints (530) such as conservation of mass and energy or performance characteristics of equipments such as pump characteristic curves; inequality constraints 540 based on the process requirement.
Fig. 6 is flow diagram which illustrates a method by which gross errors may be detected and identified. Post reconciliation, x2 test is performed on the entire vector of estimated values. The null hypothesis here is that no gross error is present. If the test statistic is greater than the critical value, the null hypothesis is rejected i.e. gross error is detected.
If gross error is detected, identification of gross error is then taken up. GLR test statistic ?i is calculated for each of the sensors and the ones with maximum test

statistic ?max is identified as having gross error. This process is repeated on the remaining sensors till no gross error is detected. If any of the sensors are identified to have gross error, they are dripped from the objective function and reconciliation is performed again and the values for faulty sensors are now estimated.

We Claim-
1. A method for validating and reconciling measured process variables and estimating the unmeasured process variables, by means of a processor (184) configured to a storage device, for detecting, identifying and estimating for faulty sensors in a power plant at steady state, the method comprising the steps of:
- configuring the measured process by a process configuration module using steady state process models;
- receiving field data from plurality of instrument devices (111) from a central control system,
- validating the field data by a preprocessing module (130),
- configuring the process dynamically based on equipment status,
- arriving at redundancy and observability of the process,
- reconciling the measured process variables using constrained non-linear
optimization, - estimating the unmeasured process variables,
- detecting presence of faulty sensors,
- identifying faulty sensors and estimating for variables indicated.

2. The method as claimed in claim 1, wherein the preprocessing module (130) also performs steady state check and range check to validate the measured data.
3. The method as claimed in claim 1, wherein process models are constituted by constraint equations which may comprise of equality constraints and inequality constraints based on the process requirement.
4.The method as claimed in claim 1 and 3, wherein the process models that comprise of equality constraints includes conservation of mass and energy or performance characteristics of equipment such as pump characteristic curves.

5. A system for validating and reconciling measured process variables and
estimating the unmeasured process variables, by means of a processor (184)
configured to a storage device, for detecting, identifying and estimating for faulty
sensors in a power plant at steady state, wherein the measured process variables
are reconciled using constrained non-linear optimization and unmeasured process
variables are estimated.
6. The method as claimed in claim 1, as illustrated in the accompanying drawings.