Cognitive platform and method for energy management for enterprises

• US 10,481,629 B2
• Filed: 07/24/2017
• Issued: 11/19/2019
• Est. Priority Date: 04/21/2014
• Status: Active Grant

• Current Assignee: Amber Flux Private Limited
• Original Assignee: Amber Flux Private Limited
• Inventors: Goparaju, Muralidhar, Goparaju, Vinulata
• Primary Examiner(s): Chen, Alan
• Application Number:
  US15/658,220
• Publication Number:
  US 20170322579A1
• Time in Patent Office: 848 Days
• Field of Search: None
• CPC Class Codes: G05B 15/02 electric
• CPC Class Codes: G05F 1/66 Regulating electric power
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence

Abstract

This invention is a platform and method for using cognition for managing enterprise energy needs. Cognitive platform allows for dynamic management of energy consumption, demand and baseline calculations by use of a cognitive platform and cognitive device. Employees as well as internal and external stakeholders can set performance indicators and monitor the parameters against the energy performance indicators. Based on the initial knowledge, the system identifies improvements in order to reach the energy key performance indicators. Depending on the feedback, the system learns and improves the accuracy of the predictions and suits them to a given industry or given enterprise scenario. Enterprise-wide energy and/or environmental management covers policies, planning, key performance indicators, goals, targets, works flows, user management, asset mapping, input-output energy flows, conservation options, performance management, analytics. The system and method allow for monitoring and verification by internal and/or external stakeholders.

Claims

1. A method for processing information for resource management, said method comprising the steps of:

(a) receiving data related to at least one resource from a plurality of data sources;

(b) performing at least one learning operation, via a cognitive decision maker, to improve resource optimization using a cognitive or machine learning process that assesses the effect of parameter variation in reaching a predetermined goal and maintains feedback information, wherein said at least one learning operation uses a priori and derived knoweledge using plurality of processes, wherein the said plurality of processes extracts one or more features from the data based on said a priori and derived knowledge and maps one or more features using a classification system employing at least one ordering selected from a group of rank, score and weights;

(c) providing at least one command or processed information to any one of;

an equipment, an external application or a storage element, based on parameters to best reach said predetermined goal and maintain feedback information.

2. The method of claim 1, wherein data from said plurality of data sources is related to any one of a furnace, air conditioner, heater, motor, fan, blower, pump, compressor, boiler, meter, heat exchanger, flare stack, battery, computer, computer software, router, antenna, switch, transmission device, wire-line network, or a wireless network.

3. The method of claim 1, wherein said at least one learning operation is related to at least one of energy consumption, maintenance, load, a resource demand, resource supply, resource usage, power off, power on, idle state, sleep state, CPU utilization, memory utilization, spectrum allocation, bandwidth allocation or sensed information.

4. The method of claim 1, wherein said at least one learning operation uses multi-criteria decision making to approach a target or predetermined goal.

5. The method of claim 1, wherein said at least one learning operation uses spatial reasoning based on qualitative abstractions of spatial aspects of data.

6. The method of claim 1, wherein said at least one learning operation performs data enrichment for knowledge derivation.

7. The method of claim 1, wherein said at least one learning operation is based on at least feedback data which is selected from the group of observation, decision and result.

8. The method of claim 1, wherein said at least one learning operation uses:

(a) a network of nodes and edges where each node or edge is comprised of at least one member representing at least one subset of training data;

(b) a training dataset that has information related to features;

(c) a module that groups data into more than sets wherein each data point is member of at least one set and generates at least one transformation dataset wherein each feature representation is associated with at least one set and applies at least one machine learning model to the transformation dataset and generates at least one predictive model.

9. The method of claim 1, wherein said at least one learning operation is based on differentiation of predicted outcomes from actual outcomes.

10. The method of claim 1, wherein said at least one learning operation uses:

(a) searching of data to extract at least one search result comprising plurality of events wherein each event includes one or more field values;

(b) derivation of probability for each field value of a plurality of field values; and

(c) probabilities to determine the likelihood of occurrence of an event.

11. The method of claim 1, wherein the received data comprises positive class data and unlabeled data and said at least one learning operation includes:

(a) generating a training dataset of a negative class from positive class data;

(b) generating an unlabeled data subset and a dataset that does not belong to any class of training data;

(c) classifying positive dataset based on the said training dataset, unlabeled data subset and the said dataset that does not belong to any class of training data;

(d) ranking said classified positive data.

12. The method of claim 1, wherein said at least one learning operation includes:

(a) application of a first level of complexity of a machine learning model to an input value;

(b) determination of whether the first level of complexity is able to classify the input value; and

(c) if the first level of complexity is determined to be unable to classify the input value, application of a second level of complexity of the machine learning model to the input value, wherein the second level of complexity is to classify the input value.

13. The method of claim 1, wherein data is received using:

(a) a software based intelligent agent comprising a software model having at least one condition associated with one rule;

(b) comparison of data to at least one condition to determine whether the at least one condition threshold is met and providing a response if the at least one condition threshold is met.

14. At least one non-transitory computer readable storage medium encoded with processing instructions performed by at least one processor, said at least one computer readable storage medium and at least one processor comprising an information processing system for resource management, said system comprises:

(a) a module for receiving data related to at least one resource from plurality of data sources;

(b) a module for performing at least one learning operation, via a cognitive decision maker, to improve resource optimization using cognition or machine learning wherein the said at least one learning operation uses a priori and derived knowledge using plurality of processes, wherein said plurality of processes extracts one or more features from the data based on said a priori and derived knowledge and maps one or more features using a classification system employing at least one ordering selected from a group of rank, score and weights;

(c) a module for providing processed information or at least one command to any one of an equipment, external application or storage of any type.

15. The computer readable storage medium of claim 14, wherein data from said plurality of data sources is related to any one of a furnace, air conditioner, heater, motor, fan, blower, pump, compressor, boiler, meter, heat exchanger, flare stack, battery, computer, computer software, router, antenna, switch, transmission device, wire-line network, or a wireless network.

16. The computer readable storage medium of claim 14, wherein said at least one learning operation is related to any one or more of energy consumption, maintenance, load, resource demand, resource supply, resource usage, power off, power on, idle state, sleep state, CPU utilization, memory utilization, spectrum allocation or bandwidth allocation or sensed information.

18. The computer readable storage medium of claim 14, wherein said at least one learning operation uses spatial reasoning based on qualitative abstractions of spatial aspects of knowledge.

19. The computer readable storage medium of claim 14, wherein said at least one learning operation performs data enrichment for knowledge derivation.

20. The computer readable storage medium of claim 14, wherein said at least one learning operation is based on at least one feedback whose type is selected from the group of observation, decision and result.

21. The computer readable storage medium of claim 14, wherein said at least one learning operation uses:

(a) a network of nodes and edges where each node or edge is comprised of at least one member representing at least one subset of training data;

(b) a training dataset that has information related to features;

(c) a module that groups data into more than one sets wherein each data point is member of at least one set and generates at least one transformation dataset wherein each feature representation is associated with at least one set and applies at least one machine learning model to the transformation dataset and generates at least one predictive model.

22. The computer readable storage medium of claim 14, wherein said at least one learning operation is based on differentiation of predicted outcomes from actual outcomes.

23. The computer readable storage medium of claim 14, wherein said at least one learning operation uses:

(a) searching of data to extract at least one search result comprising plurality of events wherein each event includes one or more field values;

(b) a probability of occurrence derivation for each field value of the plurality of field values and using probabilities to determine a probability of occurrence of the event.

24. The computer readable storage medium of claim 14, wherein the received data comprises positive class data and unlabeled data and said at least one learning operation includes:

(a) generating a training dataset of a negative class from positive class data;

(b) generating an unlabeled data subset and a dataset that does not belong to any class of training data;

(c) classifying positive dataset based on the said training dataset, unlabeled data subset and the said dataset that does not belong to any class of training data;

(d) ranking said classified positive data.

25. The computer readable storage medium of claim 14, wherein said at least one learning operation includes:

(a) application of a first level of complexity of a machine learning model to an input value;

(b) determination of whether the first level of complexity is able to classify the input value; and

(c) if the first level of complexity is determined to be unable to classify the input value, application of a second level of complexity of the machine learning model to the input value, wherein the second level of complexity to classify the input value.

26. The computer readable storage medium of claim 14, wherein data is received using:

(a) a software based intelligent agent employing a software model comprised of at least one condition associated with one rule;

(b) correlation of data with at least one condition to determine whether the at least one condition threshold is met and responding when at least one condition threshold has been met.

17. The computer readable storage medium of 14, wherein said at least one learning operation uses multi-criteria decision making to approach a target or goal.