Plant Studied
- Capacity
- 67 MW
- Location
- Maharashtra, India
- Sensor Level
- Combiner Box
- Pre-Engagement Daily Loss
- Avg Loss 7.3%
Deploys Quadrical Ai’s Digital Twin Based Solution for Performance Improvement
SolarArise invests and manages large scale, grid connected solar projects - both commissioned and greenfield - with a long-term hold philosophy. Backed by Kotak Infrastructure, Global Energy Efficiency, Renewable Fund (GEEREF) and Thomas Lloyd, SolarArise aims to develop, own, and operate a large Solar portfolio in India.
Results
3.1%Post-Engagement Daily Loss
4.2%Reduction – Average Daily Loss | Downward Trend Continuing
Plant Configuration
- # of inverters
- 20
- # of string-combiner boxes
- 400
- Configurations
- Single-Axis and Seasonal Tracker
Challenge
The team at SolarArise knew there had to be a way to increase generation at their plants – perhaps even beyond budgeted expectations set forth by PV Syst models. Too dependent on pyranometer accuracy, those models were losing precision with Solar plants becoming ever larger. Though Quadrical Ai can operate with data from any CMMS, SolarArise already had a Quadrical Ai Monitoring platform. The next step was to use the data to utilize O&M teams more effectively in order to increase yield generation.
The Plant
The Yield Improvement plan for SolarArise was to use data to benchmark its 18-Month Pre-Engagement performance. It was followed up with a 3-Month Pre-Engagement Audit and a 1-Week Remediation Report. This was iterated multiple times as a monthly Audit - Remediation cycle. Culminating in RealTime information being updated daily in an Operation Dashboard made Performance Improvement become both continuous (daily) and seamless. Solar Industry specific algorithms used by Quadrical Ai are part of a robust, secure, engineering solution. With plug & play integration capability, and an intuitive interface, they continue to provide speed-to-value and enhanced security over time.
Our Digital Twin Benchmark
Our solution for SolarArise used Digital Twin AI, built on Peer Comparison, Environmental Context and Time. Being a weather and module corrected benchmark, our Digital Twin based Machine Learning continues to be reliable at tracking needle-in-a-haystack issues even down to the module level.Using SCADA, tracker, weather, calibration and manufacturer data, plus data from their own RealTime Monitoring Platform.
SolarArise received predictive insights for O&M. Integrating all these types of data allowed us to build stacked Digital Twins for all hardware relationships in the plant. This allowed for Automatic RealTime Plant Monitoring with tickets issued to solve O&M issues prioritized by revenue impact. Though not required by SolarArise, Yield Forecasting is also available on the Quadrical Ai Platform. Contracted (Budgeted) energy is less useful as its overly dependent on:
- Simplistic PVSyst calculations of Irradiance
- Pyranometer accuracy
This makes it more unreliable as plants become larger.
Digital Twin – 3-Dimensional Benchmark:
- Context - Satellite Weather Data, Weather Station (Pyranometer), and Plant Yield
- Peer and Proximity Data for better gauge of granular conditions
- Meter specific using Time Series performance of individual meters
Pre-Engagement Performance Benchmark (18 Mos)
Contracted (Budgeted) Benchmark Variance = -43% to +100%
Digital Twin Benchmark Variance = +2.29 to +15%
A Stable Benchmark (3x lower sigma) = More Precise Loss Tracking & Remediation
We are deeply focused on energy uplift and believe our plants can produce more than the budget targets.
Shading Loss – 301 MWh (3 Months)
Actual Energy Produced= 31, 267 MWh
(Contracted) Expectations= 32, 072 MWh
With Right Conditions, Plant May Produce Even More= 33, 730 MWh
Examples of Needle-in-Haystack / Temporal Issues
| Category of Loss | # of tickets | Loss % | Loss (MWh) | Comments |
|---|---|---|---|---|
| Slacker SCBs | 5 | 0.10% | 26.8 | Check for structural issues, shading, and wire heating during various times of day |
| Others (SCBs uncorrelated with Digital Twin Benchmark) | 3 | 0.03% | 9.8 | Out-of-Sync SCB’s for additional observation |
| SCBs with Long Term Issues | 21 | 0.30% | 113 | Investigate from structural perspective |
| SCBs with Shading Loss | 87 | 0.90% | 302 | Table-to-Table Morning/Evening Shading |
Weekly Remediation Reports
| Category of Loss | # of tickets | Loss % | Loss (MWh) |
|---|---|---|---|
| Failures - Y-Connectors, Cable (Others) | 179 | 1.12% | 26.7 |
| Inverter (DC to AC) | 20 | 0.84% | 20.0 |
| Slackers - Late Starters / Early Finishers | 15 | 0.44% | 10.6 |
| Uncorrelated to Benchmark | 13 | 0.38% | 9.1 |
Long Term Structural Losses = 21 SCB’s 113 MWh (0.3%)
Identified 7 SCB’s which contributed to 44 MWh (0.3%) for Further Investigation and Manufacturer Escalation
Soiling Analysis
Continuously investing in new feature additions, Quadrical Ai developed Digital Twins to detect Soiling Losses. During our Case Study window, we found additional Soiling Loss of 372 MWh and identified 3 Blocks which required more frequent cleaning.
RealTime Revenue Prioritized Operation Dashboard
Post-Quadrical Ai Check-In
| 18-Mos Baseline | Reductions | ||
|---|---|---|---|
| Generation Loss | Shading:Fixations around Fixing | 0.3% | 0.1% |
| Structural Losses: Degraded Panels | 0.2% | 0.2% | |
| Temporal Losses: MC4, Y-Connector, Insulation, Ground Fault Reductions and Panel Errors | 3.0% | 1.6% | |
| Soiling | 1.7% | 1.4% | |
| Other (Site Outages) | 0.6% | 0.1% | |
| Invertor Loss | Clipping | ||
| Temporal Losses (Fan failures, Overheating etc) | 1.0% | 0.7% | |
| Other | 0.2% | 0.0% | |
| AC Loss | Curtailment | ||
| Temporal losses | |||
| Temporal losses | 0.2% | 0.0% | |
| Total | 7.3% | 4.2% |
Total Loss%
Quadrical Ai’s Digital Twin approach was instrumental in setting realistic expectations and then pinpointing areas of underperformance that focused our teams’ efforts.
Digital Twin Based AI for Complete Solution
Unlike traditional models which rely on an “ideal component” formula, we use a Machine Learning model of efficiency and behavior. Based on data collected from 100’s of input parameters and actual inverter and plant yields, these merged individual and average performances create sophisticated special purpose models. This makes them fine tuned to capture sensor data, degradation to create forecasts and foresee future anomalies. Adapting to tolerances, the environment and history, the Quadrical Ai Digital Twins tune continuously, resulting in steady optimization of the aging plant.
- Robust Metrics and Benchmarking
- Energy Forecasting with Storage, Trading, and Pricing Optimizations
- Monitoring with Accurate Needle-in-Haystack Identification
- Predictive Maintenance for Outage Prevention
- O&M Guidance with Revenue Prioritized and Actionable Tickets
- Plant Audits
We know that a Digital Twin AI solution can tell natural variances in the system apart from problematic performance. With Quadrical Ai’s ability to adapt with time and learning, they’ll be optimizing your plant assets even at Year 25.
Benefits
- Management
- RealTime Portfolio Performance Monitoring & Revenue Impact.
- O&M Team
- Corrective, Preventive, & Predictive O&M Directions & Ticketing Workflow.
- Internal Data Science Team
- RealTime Structured Schema Based Data of Actual as well as Reliable System Benchmark – Digital Twin data and all of it available on a RealTime dashboard.
Future Performance
Loss percentages will continue to be further improved as timely actions are taken on O&M guidance. With on-site O&M feedback, the Quadrical Ai system is also trained to become more intelligent with time. Globally, almost every solar company faces similar challenges. As assets age, prices paid per MWh are decreasing, while scale and complexity are increasing. This makes holistic long-term planning essential for asset owners and managers.
Our partnership with SolarArise has been fruitful as they think in KWh, not just KW. Looking at technology as a revenue generation tool, not a cost center, makes it a value addition and the foundation of organic growth.
Deployed on 3064 MW Solar + 667 MW / 2433 MWh Storage + 200 MW Wind