Tata Consultancy Services Limited

University System of Maryland

Arunchandar Vasan

Anand Sivasubramaniam

Venkatesh Sarangan

Srinarayana Nagarathinam

Rajesh Jayaprakash

Venkata Ramakrishna Padullaparthi

Sivabalan Thirunavukkarasu

Pandeeswari Indumathi Sankaranarayanan

Shiva R Iyer

Perumal Rajaram

Avinash Achar

Ajay Narayanan

Arvind Ramanujam

Praveen Orvakanti

Rajesh Subbiah

Venkata Ramakrishna P

Ramasubramanian Suriyanarayanan

Prasant Kumar Misra

Ankush Ojha

Alok Ranjan

Iyswarya Narayanan

Depak Sudarsanam

Karpagam Murugappan

Ajjampur Udaya Shankar

Krishna Kumar Sunil Komdam

Praveen Manoharan

Tejasvi Malladi

Venkataramakrishna P

Maathangi Sankar

Shravan Srinivasan

Abinaya Manimaran

Gollakota Phani Bhargava Kaushik

Sneha Mary Thumma

Harihara Subramaniam Muralidharan

Kundan Kandhway

Krishna Kumar Sunil Komadam

Vankatesh Sarangan

Reinforcement Learning agent interacting with a real-world building to determine optimal policy may not be viable due to comfort constraints. Embodiments of the present disclosure provide multi-deep agent RL for dynamically controlling electrical equipment in buildings, wherein a simulation model is generated using design specification of (i) controllable electrical equipment (or subsystem) and (ii) building. Each RL agent is trained using simulation model and deployed in the subsystem. Reward function for each subsystem includes some portion of reward from other subsystem(s). Based on reward function of each RL agent, each RL agent learns an optimal control parameter during execution of RL agent in subsystem. Further, a global optimal control parameter list is generated using the optimal control parameter. The control parameters in the global optimal control parameters list are fine-tuned to improve subsystem's performance. Information on fine-tuning parameters of the subsystem and reward function are used for training RL agents.

Multi-agent deep reinforcement learning for dynamically controlling electrical equipment in buildings

Considering the dependency of a flight hold time on multitude of dynamically varying factors, determining an optimal hold time balancing between passenger utility and airline utility is challenging. State of art approaches are limited to use of only deterministic approaches with limited ML assistance that require huge labelled training data. Embodiments disclosed herein provide a method and system for computing and recommending optimal hold time for every flight of an airline so as to minimize passenger misconnects in airline operations through Reinforcement Learning (RL). The method disclosed utilizes RL, which is trained to make decision at a flight level considering local factors while still adhering to the global objective based on global factors. Further method introduces business constants in the rewards to the RL agents bringing in airline specific flexibility in reward function.

Method and system for minimizing passenger misconnects in airline operations through learning

This disclosure relates generally to method and system for maximizing space utilization in a building. Due to current pandemic scenario many organizations eventually need to plan for the return of employees to office space ensuring biosafety. The challenge of maximizing the office space utilization ensuring occupants biosafety and comfort thereby minimizing HVAC energy consumption is necessary. The method utilizes two heuristic approaches for determining maximum allowable occupants placement in the open plan space using an optimal occupant placement technique. This minimizes the HVAC energy if the actual count is lesser than the possible maximum occupants can be placed which further optimizes energy using a joint actuator control technique. Additionally, the proposed two heuristic approaches improve space utilization for the infection rate ensuring bio safety. Full utilization of open plan space is possible when the community infection rate and exposure duration are relatively low resulting low risk probability for uninfected occupants.

Method and system for maximizing space utilization in a building

This disclosure relates generally to method and system for detecting on-street parking violations. The method include capturing, by using an media capturing device embodied in an electronic device mounted in a vehicle, a video stream of a scene during a trip of the vehicle. The video stream is processed at the electronic device to identify target objects such as no-parking signage and vehicles parked in the vicinity thereof. A meta-information associated with the target objects is stored in form of a short-term historian in a repository associated with the electronic device. The absolute locations of the target objects are determined and the historian is updated with the values of the absolute locations. A set of unique target objects is determined from amongst the target objects and a meta-information associated with the unique objects is sent to a cloud server for determining parking violations.

System and method for detecting on-street parking violations

Vehicle-to-Grid (V2G) technologies are being adopted to reduce peak demand and to take over as energy sources during grid instability. It is necessary to estimate attributes of electric vehicle trips and residual battery charge in order to correctly predict spatio-temporal availability of energy from EVs to form a micro grid. However, it may not be feasible to get the required attributes for all vehicles in a city-scale traffic scenario. Embodiments of the present disclosure and system address the problem of accurately estimating the local energy reserve that is available from parked EVs during a given time of the day. In addition, the system also determines which neighborhoods have the potential to form micro grids using the parked EVs during a given time period. This will help grid operator(s) to plan and design smart grids which can create EV-powered micro grids in neighborhoods during periods of peak demand or during disruptions.

Electric vehicle based micro grid identification for on-demand power supply

Traditionally, benchmarking of asset performance involves comparing actual performance with ideal values that correspond to test conditions which may not be realized in practice leading to inappropriate ranking of the assets. Systems and methods of the present disclosure use condition-aware reference curves for estimating the maximum possible operating efficiencies (under specific operating conditions) instead of the theoretical maximum efficiencies. The reference curves are received from the manufacturer or obtained from on-site test results. Benchmarking is then performed based on two dimensions, viz., an inter-asset metric and an intra-asset metric that are analogous to the first law and second law of thermodynamics respectively. The two-dimensional benchmarking then helps in identifying inefficient assets that may be analyzed further for finding the root cause. Tracking the performance of assets over time greatly helps in operations and maintenance, and thus reducing downtime of systems and accordingly the operating costs.

Methods and systems for benchmarking asset performance

Electrical utilities offer incentives to customers to reduce consumption during periods of demand-supply mismatch. A building's participation in demand response (DR) depends both on its ability (due to building constraints), and its willingness (a function of incentive) to reduce electricity. Customers prefer a large incentive whereas a utility would want to minimize the revenue outflow to achieve a target reduction. Systems and methods of the present disclosure identify optimal incentive from the utility's perspective reflecting this trade-off. A model is built to estimate the demand response potential (DRP) of a building for a given incentive offered by the utility. The models for individual buildings are used to characterize the behavior of an ensemble of buildings. The utility may then decide optimum incentives that should be offered to achieve a target DR, using the associated DRP.

Systems and methods for optimizing incentives for demand response

The present disclosure provides system and method for determining optimal decision parameters for a demand response (DR) event involving a District Cooling Plant (DCP). Most of conventional DR event techniques address control of building-level energy consumption loads alone while in presence of District Cooling (DC) has not received much attention when a plurality of buildings are served by a District Cooling Plant (DCP). The disclosed system and method determine set points of optimal decision parameters of the plurality of buildings and the DCP, by conditioning and un-conditioning on the DCP parameters such that a thermal discomfort of occupants residing in the plurality of buildings is minimum and achieves a maximum target energy demand reduction during the DR event. The disclosed system and method work for hundreds of buildings and able to determine the optimal decision parameters for each building and the DCP efficiently.

Determining optimal decision parameters for a demand response event involving a district cooling plant

Various fault types occurring at multiple possible locations in the solar panel network are simulated using the network simulation model. The dataset covering multiple fault scenarios and multiple no-fault scenarios is determined for training the CNN model. The fault scenarios include one fault type alone at particular location or multiple locations, as well as multiple fault types at multiple locations. The fault types include a short circuit fault, an open circuit fault, a shading fault, a soiling fault, a hot-spot fault, an arc fault, a degradation fault, and a clipping fault, the short circuit fault comprises a line-line fault, and a line-ground fault The convolutional neural network (CNN) model is trained with fault datasets and no-fault datasets covering various fault sensors and no-fault scenarios to generate the FDDL model. The fault datasets and no-fault datasets are determined based on the network simulation model of the solar panel network.

Methods and systems for fault detection, diagnosis and localization in solar panel network

Method and system for predicting temporal-spatial distribution of load demand on an electric grid due to a plurality of Electric Vehicles (EVs) is described. The method includes creating an EV load demand (EVLD) model for a Region of Interest (ROI) serviced by the electric grid, wherein the EVLD model integrates an EV model and a transport simulator simulating EV traffic conditions for the ROI. Further, the method includes computing the load demand in time and space in terms of State of Charge (SOC) of a battery for each EV among the plurality of EVs in the ROI, based on the EVLD model. Furthermore, the method includes aggregating the computed the load demand, in terms of the SOC, of each EV in time domain and space domain to create a temporal-spatial impact of the load demand by the plurality of EVs on the electric grid for the ROI.

Method and system providing temporal-spatial prediction of load demand

Conventional systems for monitoring pipe networks are generally not scalable, impractical in the field with uncontrolled environments or rely of static features of pipes that are vary depending on the pipes under consideration. The ideal sensor-ed monitoring systems are not economically viable. Systems and methods of the present disclosure provide an improved data-driven model to rank pipes in the order of burst probabilities, by including dynamic feature values of pipes such as pressure and flow that depends on network structure and operations. The present disclosure enables estimating approximate values for the dynamic features since they are hard to estimate accurately in the absence of a calibrated hydraulic model. The present disclosure also validates the estimated approximate dynamic feature values for the purpose of estimating bursts likelihood vis-a-vis accurate values of the dynamic metrics.

Ranking pipes for maintenance in pipe networks using approximate hydraulic metrics

A technique for energy sample forecasting of heating, venting and air conditioning-refrigeration (HVAC-R) systems is disclosed. In an example, a first expected energy sample of a HVAC-R system at a first time period is forecasted by modelling actual energy samples of the HVAC-R system at previous time periods using a statistically-based seasonal-autoregressive integrated moving average (SARIMA) model. Further, an anomaly is detected at the first time period when deviation between the first expected energy sample and an actual energy sample at the first time period is greater than a dynamic context sensitive threshold. Also, an expected energy sample at next time period is forecasted by modelling a second expected energy sample of the HVAC-R system at the first time period using the statistically-based SARIMA model upon detecting anomaly. The second expected energy sample is forecasted by modelling the actual energy samples at the previous time periods using a physical model.

System and method for energy sample forecasting of HVAC-R systems

Devices and methods comprise determining an actual weather independent energy consumption value and actual weather dependent energy consumption value of a building. The devices and methods further comprise, obtaining energy simulation information from a data source. Further, an ideal weather independent energy consumption value and one or more ideal loads are generated based on the energy simulation information, building information, and total energy consumption of the building. The devices and methods further comprise determining a benchmarking score for the building using the actual weather independent energy consumption value, the actual weather dependent energy consumption value, the ideal weather independent energy consumption value, and the one or more ideal loads for managing energy consumption associated with the building.

Devices and methods for energy benchmarking of buildings

A system and method for optimizing energy consumption in a plurality of air handling units (AHUs) in a zone is provided. The system comprising a zone thermal unit that is configured to obtain a first set of input parameters including an internal heat gains, a surface convective loads, an intra-zone mixing, a supply air temperature, a second set of input parameters including internal moisture gains, a supply humidity ratio, and a third set of input parameters including an air contaminant concentration and an ambient contaminant concentration of the AHUs and generates a first set of output parameters including a zone temperature, a humidity ratio and an air concentration. The system further includes an optimizer that is configured to generate a second set of output parameters including an optimum combination of AHU flow rates for the AHUs based on at least one of the first set of output parameters.

Method and system to optimize energy consumption in a zone with a multi-air handling unit (AHU) setup

Described herein, are methods and systems for locating a leak in a water distribution network. According to an implementation, a leak situation in the water distribution network is detected based on a flow difference value between an actual flow value and a predicted flow value of an inlet flow meter of the water distribution network at at least one time interval. Leak signature values of demand nodes in the water distribution network at the at least one time interval are determined. A leak signature value of a respective demand node at a respective time interval is determined based on centrality metrics, the predicted flow value at the respective time interval, and static physical properties related to the water distribution network. At least one possible leak node is identified based on the flow difference value and the leak signature values of the demand nodes at the at least one time interval.

Growing community of inventors

Chennai, India

Arunchandar Vasan