Advisory Systems and Prediction Results

This module provides infrastructure for look-ahead predictions that support real-time advisory systems.

Overview

Modern process operations benefit from predictive capabilities:

• Look-Ahead Simulation: Predict process behavior minutes to hours ahead
• Uncertainty Quantification: Confidence intervals on predictions
• Constraint Monitoring: Early warning of limit violations
• Operator Guidance: Actionable recommendations

Classes

PredictionResult

Structured output from predictive simulations for advisory systems.

Key Features

• Time Horizon: Predictions from minutes to days ahead
• Uncertainty Bounds: 95% confidence intervals on values
• Violation Detection: Identify which limits may be exceeded
• Explainability: Human-readable prediction drivers
• Recommendations: Suggested operator actions

Usage Example

// Run look-ahead simulation (conceptual - requires integration)
PredictionResult prediction = new PredictionResult(
    Duration.ofHours(2), 
    "Base Case"
);

// Add predicted values with uncertainty
prediction.addPredictedValue(
    "separator.pressure",
    new PredictedValue(52.5, 2.1, "bara")  // mean, stddev, unit
);

prediction.addPredictedValue(
    "compressor.power",
    new PredictedValue(4500.0, 150.0, "kW")
);

// Check for constraint violations
if (prediction.hasViolations()) {
    System.out.println(prediction.getViolationSummary());
}

// Get operator guidance
String advice = prediction.getAdvisoryRecommendation();
System.out.println(advice);

PredictedValue

A predicted value with uncertainty quantification.

Constructors

// With standard deviation (auto-calculates 95% CI)
PredictedValue value = new PredictedValue(
    50.0,   // mean
    2.5,    // standard deviation
    "bara"  // unit
);

// With explicit confidence bounds
PredictedValue value = new PredictedValue(
    50.0,   // mean
    45.0,   // lower 95% bound
    55.0,   // upper 95% bound
    "bara", // unit
    0.95    // confidence
);

// Deterministic (no uncertainty)
PredictedValue value = PredictedValue.deterministic(50.0, "bara");

Accessors

double mean = value.getMean();
double stdDev = value.getStandardDeviation();
double lower = value.getLower95();
double upper = value.getUpper95();
String unit = value.getUnit();
double confidence = value.getConfidence();

ConstraintViolation

Predicted constraint violation with timing and severity.

Severity Levels

Level	Description	Response
LOW	Minor deviation expected	Monitor
MEDIUM	Significant deviation	Plan response
HIGH	Equipment limits may be exceeded	Act soon
CRITICAL	Safety limits may be exceeded	Immediate action

Usage Example

ConstraintViolation violation = new ConstraintViolation(
    "HP Separator Pressure Limit",  // constraint name
    "separator.pressure",            // variable
    85.0,                            // predicted value
    80.0,                            // limit value
    "bara",                          // unit
    Duration.ofMinutes(45),          // time to violation
    Severity.HIGH                    // severity
);

violation.setSuggestedAction("Reduce inlet flow or open bypass");

String description = violation.getDescription();
// "HP Separator Pressure Limit: separator.pressure expected to 
//  reach 85.00 bara (limit: 80.00) in 45 min"

Prediction Status

Status	Description
SUCCESS	Prediction completed normally
WARNING	Completed but violations detected
FAILED	Prediction did not converge
DATA_QUALITY_ISSUE	Input data problems detected

Integration Patterns

With Real-Time Data

// Pseudo-code for real-time integration
while (running) {
    // Update process with live data
    updateFromOPC(process, liveData);
    
    // Run prediction
    PredictionResult prediction = runLookAhead(process, Duration.ofHours(1));
    
    // Send to advisory system
    if (prediction.hasViolations()) {
        sendAlert(prediction.getAdvisoryRecommendation());
    }
    
    // Wait for next cycle
    Thread.sleep(60000);
}

With MPC Integration

// Provide prediction to MPC controller
PredictionResult prediction = runLookAhead(process, mpcHorizon);

// Extract predicted values for MPC
Map<String, PredictedValue> values = prediction.getAllPredictedValues();
for (var entry : values.entrySet()) {
    mpc.setDisturbanceForecast(
        entry.getKey(), 
        entry.getValue().getMean()
    );
}

Assumptions and Documentation

PredictionResult prediction = new PredictionResult(Duration.ofHours(2));

// Document assumptions used
prediction.addAssumption("Inlet conditions remain constant");
prediction.addAssumption("No equipment trips");
prediction.addAssumption("Cooling water available at design temperature");

// Set explanation for operators
prediction.setExplanation(
    "Pressure rising due to increasing gas-oil ratio in feed"
);

// Set confidence level
prediction.setOverallConfidence(0.85);

Best Practices

1. Appropriate Horizons: Match prediction horizon to decision timescale
2. Uncertainty Communication: Always show confidence bounds
3. Actionable Recommendations: Provide specific suggested actions
4. Assumption Transparency: Document all prediction assumptions
5. Status Monitoring: Track prediction quality over time

Related Documentation

• Physics Constraint Validation - Validate predictions against physics
• Safety Scenarios - What-if analysis
• MPC Integration - Real-time control integration