Skip to content Skip to footer

Case Studies: Volt /VAR Optimization and Predictive Substation Automation for High Voltage Transmission and Distribution System

1

Background and Problem Statement

With the advent of smart grid systems and increasing focus on intelligent electrical devices (IED) deployments,  the aged grid substations are equipped with Advanced distribution automation software that can complement the  existing SCADA systems. As opposed to generic automation platforms, modern utilities are preferring more  proactive and intelligent systems that can manage both the substation as well as entire network to provide real time actionable insights and predictive signals on peak load conditions and avoid catastrophic failures.  

The dynamic Volt/VER control to stabilize the voltage and minimize the T&D loss across the Long Distanced High  Voltage Transmission and Distribution Systems would require optimal settings and timely operation of substation  assets such as Voltage regulators, reactors/ capacitor banks. Rania’s energy management platform CognitEMS™ has introduced a real-time AI powered real-time network monitoring solution to manage the Reactive Power/  technical Loss along with predictive Voltage Correction, by offering dynamic switching of reactors/ Capacitor banks across the connected substations. It also addresses the critical imperatives of load/ demand forecasting  cross the network at substation and ICT level (Interconnected transformer) to stabilize the load curve based on  real-time/ day-ahead, week-ahead operating conditions (projected peak/ off-peak scenarios and weather that  impacts the desmids).

The conventional process to manage these reactors is typically a manual operation and are  prone to human errors in delayed actions and over/ under corrections. Most of the operators monitor the changes  of Line Parameters at the substation level (Voltage, Active, Reactive Power etc.) and have limited visibility on the  impact of the action at the overall network segment.  

Hence the lack of visibility on real-time network status, and reactive manual operations have been causing  significant T&D losses, voltages instability and pre-mature wear-n-tear of the Reactors/ capacitor banks  diminishing its lifetime values.  

Ranial Systems has addressed this critical issues with its Proprietary Machine Learning & Artificial Intelligence  powered grid monitoring solution, that proactively monitors the network status, generates predictive and  prescriptive insights, system alerts, advices substation operators for the Optimal Switching operations, as well  as forecast load patterns at both substation and ICT level more accurately. The real-time web based solution  offers real-time monitoring and system management panels and historians to saves the critical losses and plan  network operations more efficiently.

 

The patented AI /ML based real-time system address the challenges of Volt-VAR optimisation by capturing real time energy and power parameters from the substation SCADA or AMRs . Within current operating environment,  the grid Operators leverage system alerts at the individual substation level to engage manual switching that are  prone to errors and delays. Latency in Decision making and its Implementation results in over/ under corrections  in Volt-VAR optimisation process . Such siloed operations inhibits in achieving desired efficiency at the network  level , and increases the Pre-mature failure of assets.  

 In order to minimize operating cost and operational inefficiency, CognitEMS™ platform offers a real-time ingestion  of energy and power readings, directional power flow, operating condition of the assets to intercept the Volt/VAR  conditions and, offer prescriptive operations suing AI based model. The runtime uses proprietary training agent  to update the model in every 30 minutes. Such incremental learning adapts to the latest operating conditions and  offer accurate suggestive operations.

The unique grid management and automation platform hosts a massive repository of substation data that are  aggregated with demands, environmental conditions, KPIs associated with the respective substations and ICTs to  forecast the load patterns(peak/ off-peak) for hourly, day-ahead, week-ahead planning more efficiently. The AI  powered multidimensional predictive models extends accurate and timely decision support capabilities to avoid  failures. 

 

3

Implementation

Ranial System has performed a detailed network study, trained the models to implement CognitEMS™ platform  for India’s largest High Voltage Transmission Operator – Grid India ( previously known as POSOCO ) and a leading  energy distribution Company CESE Limited. Grid India Manages 766 KV and 400 KV Transmission system, in  Eastern India , with its 114 Sub-Stations . The Reactive Power Loss and Voltage Fluctuations have been resulting  due to discrete manual interventions and pre-fed control operations at the substation level. Such operations infuse  latency, and result into line losses, pre-mature failure of Assets etc.

The CognitEMS™, has been trained with historical data from the SCADA feeds of the substations and back-tested. The application interfaces and historians have been configured to meet the strategic imperatives of the client’s operational flow. The system is deployed in a production cloud environment that incrementally trains the model  every 30 minutes and, uses the most updated model to intercept the real-time data streams. Threshold Parameters (in several Operational domains) and rewards are measured in relation to overall network statistics, flow direction, and, switching conditions. The system offers visual indications and alarms to facilitate switching and changing operating parameters.

The key benefits offered are as follows:

  • Realtime visibility of Capacitor banks, voltage regulators and/or Reactors’ current Switching conditions (whether in ON / OFF state ) and gain AI based expert advice on desired Voltage and reactive Power corrections
  • Realtime Alarms/ notifications to avoid any Manual Operation based Mistakes
  • Visibilities of Substations on Actual Geo Positioning, switching conditions and power flow directions with actual connectivity across  the Substations
  • Realtime & Historical switching History & Strategy (Actual & prescribed) , with its Voltages (Actual & predicted) ,Reactive Power (Actual & predictive), Active Power (Actual) and T&D loss.
  • Visibility of Improvement of Reactive Power (using model prescribed optimization with respect to existing operating conditions)Visibility of Improvement of Line Loss Actual vs prescribed trough real-time AI based runtime.
  • This platform allows Operators to view the Load Forecasting analysis ( Actual Load vs. Predicted Load analysis) along with the comparison of improvement on load patters- near real-time, hour ahead, day ahead and week ahead
  • Correction Analysis ( for a definite time range* ) of Active Power , Reactive Power & Voltage – when AI  advice is being followed . Through this, monitoring of Minimum , Maximum & Average fluctuations of  Voltages and Reactive Power trends can be traced using intelligent historians. This Option enables  operator to view over all the Substations’ correctional data along with individual Sub-Station’s Correctional  data. Such proactive Realtime analytical feature equip clients to facilitate operations in a timely fashion. 

  • Historian for Demand vs Generation patterns and accurate predictions for individual substation and ICT. 

  • Option of uploading Data for a correctional analysis and simulation to aid network planning.

 

4

Result

The deployment of the platform was completed in 3 months. The day-to-day operations could help clients to  reduce 33% of manpower utilization and achieve an average correction of 12 % Reactive Power Loss. The system  helped the operators to limit Voltage Fluctuations under 1.8 % in real-time. Such responsive automation also  lower overall Active Power Loss significantly 6.5-8% based on off-peak and peak scenarios).  

The Inbuilt predictive intelligence has allowed the client to achieve 2X efficiency on forecasting/ predicting the  load patterns, which are more efficient than standard limiter forecasting models used by conventional systems or  tools.  

The rich web-based monitoring and centralized control system allows clients to improve responsiveness and time to-value in managing network and planning operations.

Leave a comment