Solvability and redundancy

Importance of redundancy

In the context of data reconciliation, redundancy means that a given measurement can be checked by a calculation based on other measurements. For example, a measured flow may also be calculable on the basis of several other flow measurements. The concept of being able to calculate a value to be compared with the direct measurement or another independent calculation is fundamental to data reconciliation. Without the extra information about error given by redundancy, data reconciliation cannot be performed. If there is no extra information, the system is just determined, and all measurements have to be accepted as is. Further, if there is even less information available, the system is underdetermined, and values have to be estimated.

It is very important to determine whether redundancy exists and, if it does not, which measurements should be added to gain redundancy at lowest cost.

Determine redundancy

The redundancy of measurements in a process can be determined by attempting to calculate the error in material balance to be considered. If all the data necessary are available, then the amount of process data is greater than the minimum necessary. Balance error is what the data reconciliation procedure works with and reduces to zero by data value adjustments.

In many processes, an overall material balance and its error can be routinely obtained, but internal balances that are important to a real understanding of the process performance are impossible to obtain due to a lack of instrumentation. Redundancy exists for the overall balance but not for some internal balances. Also redundancy may exist for some parts of the process or plant, but not others.

Variable classification

Variable classification is a method used by AVEVA Production Accounting to check the redundancy of each measured flow and solvability of each unmeasured flow. It also checks that the data reconciliation model has sufficient data that is consistent enough to perform the calculation.

The measured flows are classified into redundant (which may be adjusted or reconciled) and nonredundant (cannot be corrected and must be taken directly from measurements). Furthermore, the unmeasured flows are divided into solvable (can be calculated by means of balance calculation) and nonsolvable (cannot be determined).

For a nonsolvable flow it is theoretically impossible to calculate the value but AVEVA Production Accounting can still assist you by calculating it using internal distribution logic and weights supplied by you in the configuration. This functionality is a unique feature and very useful for distribution in utility balances. This use of distribution logic is explained further in the AVEVA Production Accounting User Guide.

For example, look at the following three examples, where flow A divides into flows B and C, and then rejoins to form flow D.

Example 1: Flow A Nonredundant and Stream B, C Nonsolvable

• Flow A is nonredundant (NR) because there is no way to deduce its value since there is no other measurement.

• Flow B and C are nonsolvable (NS) because data measurements are not provided for them, and there is no other data to deduce them from.

• Flow D is solvable (S), since it can be deduced from the measurement on Flow A.

Example 2: Flows B, C Nonsolvable and Flows A, D Redundant

• Flow A and D is measured. Because it has no information for flows B and C, they are nonsolvable (NS).

• Flows A and D are now redundant (R) because each value can be deduced from the other value.

Example 3: Flows B Nonredundant and Flows C Solvable

• Here, the values for all flows can be deduced, since there is measured data for flows A, D and B, and a value can be deduced for the unmeasured flow C.

• Flows A and D are redundant, because they are measured. This makes flow B nonredundant (NR) (since no other flow data is available to deduce its value) and flow C solvable(S).

Set up a data reconciliation model

When setting up a model, we recommend that all boundary streams should be redundant. These are the links to the outside world. Making them redundant gives AVEVA Production Accounting a chance to crosscheck the measurements. In a yield model, transactions from receipt points to tanks and from tanks to shipment points are almost always redundant, because they are measured and can also be deduced from changes in inventory.

In an ideal world all streams should be at least solvable – you then have a value for each flow in the facility. Although nonsolvable areas inside a facility are not desirable, in practice they may not be a problem, so long as the major flows are available. But nonsolvable boundary flows represent a major problem.

If there is insufficient redundancy (ideal redundancy is 100%, but realistically it will probably be about 33 %), you may need to extend the scope of the model. You may also consider adding artificial (or calculated) measurements derived from existing measurements that can be set up through the formula feature of Smart Objects. Examples of calculated measurements are:

• If there is a pressure control tag in an unmeasured flow, we can make a calculated measurement from the valve position value of the pressure control.

• We can make a calculated measurement from an engineering equation and/or laboratory analysis data such as

  • Hydrocarbon loss from the desalter, using only water analysis and flow

  • Coke build-up in the coker drum, using a simple calibrated model based on furnace severity and feed conditions

  • Coke burnt in the FCCU regenerator

When you are constructing a data reconciliation model, remember that in an ideal situation all unmeasured streams are solvable and most measurements are redundant. But this is not always the case, and in extreme situations it may be necessary to simplify the model, relocate instruments or install new ones to gather sufficient data for an adequate reconciliation. You can use AVEVA Production Accounting as a tool to determine the best location for new or changed instruments.

When building a data reconciliation model you need to consider the solvability and redundancy of your plant. The main principle in building a good model is to strive for as many redundant flows as possible and as few nonsolvable flows as possible. Normally, more redundant flows improve the overall quality of reconciled data.

If you study the above example carefully, you can see that there are no nonsolvable flows. Flows that are not metered are marked Solvable (S), indicating that AVEVA Production Accounting can calculate a measurement from other readings. Since all the boundary flows are metered, they are all redundant. There are three balance envelopes which consist of only redundant flows. Flow A can be crosschecked by Balance Envelope Ae and Ce. Flow B can be crosschecked by Balance Envelope Ae and Be.

Flow C can be crosschecked by Balance Envelope Be and Ce. The redundancy extent for all redundant flows is 2.

In precise terms, the redundancy extent largely influences the quality of reconciled data. If a certain flow has higher redundancy extent that other flows, the reconciled value of a certain flow is statistically more accurate than others. This is the key concept of redundancy.