SYSTEMS AND METHODS FOR MODELING ENERGY CONSUMPTION AND CREATING DEMAND RESPONSE STRATEGIES USING LEARNING-BASED APPROACHES

• US 20150248118A1
• Filed: 02/26/2015
• Published: 09/03/2015
• Est. Priority Date: 02/26/2014
• Status: Active Grant

• Current Assignee: Board of Trustees of The University of Alabama (University of Alabama System)
• Original Assignee: Board of Trustees of The University of Alabama (University of Alabama System)
• Inventors: Li, Shuhui, Sun, Ming, Zhang, Dong
• Granted Patent: US 9,817,375 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: F24F 11/30 for purposes related to the operation of the system, e.g. for safety or monitoring
• CPC Class Codes: F24F 11/46 Improving electric energy efficiency or saving
• CPC Class Codes: F24F 11/62 characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
• CPC Class Codes: F24F 11/66 Sleep mode
• CPC Class Codes: G05B 13/04 involving the use of models or simulators
• CPC Class Codes: G06N 3/006 based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent

Abstract

According to various implementations, a demand response (DR) strategy system is described that can effectively model the HVAC energy consumption of a house using a learning based approach that is based on actual energy usage data collected over a period of days. This modeled energy consumption may be used with day-ahead energy pricing and the weather forecast for the location of the house to develop a DR strategy that is more effective than prior DR strategies. In addition, a computational experiment system is described that generates DR strategies based on various energy consumption models and simulated energy usage data for the house and compares the cost effectiveness and energy usage of the generated DR strategies.

Claims

1. An energy consumption management system comprising:

one or more local receivers disposed adjacent a respective one of one or more energy consuming units in a building, each local receiver comprising a processor configured for receiving usage instructions for the adjacent energy consuming unit and causing the usage instructions to be executed for the energy consuming unit; and

a central computing system comprising;

a memory configured for storing actual usage data associated with at least one energy consuming unit in the building for one or more time windows on each of days i through j, wherein day i is the first day for which data is stored and day j is the most recent day for which data is stored; and

a processor configured for;

receiving the actual usage data from the memory;

executing at least one computer-based learning system to model energy consumption for day j+1 based on at least the actual energy usage data for the energy consumption unit;

generating a demand response strategy for the energy consuming unit for day j+1 based on the modeled energy consumption and next-day energy pricing for each time window for day j+1; and

communicating the demand response strategy of the energy consuming unit to the local receiver associated with the energy consuming unit, the demand response strategy comprising the usage instructions.

2. The system of claim 1, wherein the computer-based learning system is a neural network system.

3. The system of claim 1, wherein the computer-based learning system is a regression based system.

4. The system of claim 1, wherein causing the usage instructions to be executed for the energy consuming unit comprises communicating the usage instructions to the energy consuming unit.

5. The system of claim 1, wherein causing the usage instructions to be executed for the energy consuming unit comprises controlling the energy consuming unit according to the usage instructions.

6. The system of claim 1, wherein the demand response strategy comprises scheduled usage for the energy consuming unit during each time window of day j+1.

7. The system of claim 1, wherein the energy consuming unit is a HVAC system and the demand response strategy comprises thermostat settings for the HVAC system for each time window of day j+1.

8. A system for creating a demand response strategy for a building, the system comprising a computing device comprising:

a memory configured for storing actual usage data associated with an energy consuming unit in the building for one or more time windows on each of days i through j, wherein day i is the first day for which data is stored and j is the most recent day for which data is stored; and

a processor configured for;

receiving the actual usage data from the memory;

executing at least one computer-based learning system to model energy consumption for day j+1 based on at least the actual usage data for the energy consumption unit, and

generating a demand response strategy for the energy consuming unit for day j+1 based on the modeled energy consumption and next-day energy pricing for each time window for day j+1.

9. The system of claim 8, wherein:

the energy consuming unit is a regulatable energy consuming unit; and

the processor is further configured for;

receiving usage data associated with operating a deferrable load energy consuming unit for each run cycle,

receiving next-day energy pricing for each time window for day j+1, and

generating a demand response strategy for the deferrable load energy consuming unit, the demand response strategy for the deferrable load energy consuming unit comprising at least one time window in day j+1 during which operation of the deferrable load energy consuming unit is allowed based at least on the usage data associated with the deferrable load energy consuming unit and the next-day energy pricing for day j+1.

10. The system of claim 9, wherein the processor is further configured for electrically communicating the demand response strategy for day j+1 associated with the regulatable energy consuming unit to a first control receiver and the demand response strategy for day j+1 associated with the deferrable load energy consuming unit to a second control receiver, the first control receiver being associated with and in electrical communication with the regulatable energy consuming unit and the second control receiver being associated with and in electrical communication with the deferrable load energy consuming unit, the first and second control receivers configured for controlling operation of the regulatable energy consuming unit and the deferrable load energy consuming unit, respectively, based on the communicated demand response strategies.

11. The system of claim 9, wherein generating the demand response strategy for the deferrable load energy consuming unit comprises executing a binary integer programming strategy.

12. The system of claim 8, wherein:

the memory is configured for storing a current charge for a battery on day j, the battery configured for receiving and storing energy generated by a photovoltaic system; and

the processor is further configured for;

receiving the current charge for the battery from the memory, and

generating a control and management strategy for the battery for day j+1 based at least on the current charge and next-day energy generation and consumption difference between at least one renewable energy source and at least one energy consuming unit, wherein the control strategy for the battery comprises identifying one or more time windows during which the battery is charged when PV generation is higher than the demand or energy from the battery is discharged for consumption by one or more energy consuming units when PV generation is below a predetermined level.

13. The system of claim 8, wherein:

the energy consumption unit is an HVAC system,

the actual usage data comprises an outdoor temperature for each time window on days I through j, an indoor temperature for each time window on days i through j, and a thermostat setting for each time window on days i through j, and

the processor is further configured for using one or more of the computer based learning systems to model energy consumption of the HVAC system for day j+1 based on the weather forecast for day j+1 and the actual usage data for days i through j.

14. The system of claim 13, wherein generating the demand response strategy comprises generating thermostat settings for one or more time windows on day j+1 such that energy costs for day j+1 are minimized.

15. The system of claim 14, wherein the computer-based learning system comprises a neural network system.

16. The system of claim 15, wherein the neural network system comprises a multilayer perceptron, and the multilayer perceptron comprises:

first, second, and third input nodes, the first input node configured for receiving from the memory the outdoor temperature for each time window during each day i through j, the second input node configured for receiving from the memory the indoor temperature for each time window for each day i through j, and the third input node configured for receiving from the memory the thermostat setting for each time window for each day i through j;

a computation layer comprising a plurality of computation nodes, the processor configured for propagating signals from the input nodes through the computation layer in a forward direction on a layer-by-layer basis; and

an output layer comprising at least one output node, the output node indicating energy consumption for the HVAC system for each time window for day j+1.

17. The system of claim 14, wherein the computer-based learning system comprises a regression based system.

18. The system of claim 17, wherein the regression based system solves the following non-linear equation to learn energy consumption of the HVAC system:

Q_k=q(T_k+1^I,T_k^I,T_k^O,β

)

wherein Q_k is the k^th HVAC energy consumption observation stored in the memory, q(•

) is a third order polynomial function of a predictor variable vector (T_k+1^I,T_k^I,T_k^O) which includes thermostat settings, indoor temperatures, and outdoor temperatures for observations k=1 . . . n, respectively, n is the number of observations saved in the memory, and β

is a parameter vector that includes β

₀, β

₁, . . . β

_p.