1. Smart Home Energy CPS Scheduling
Professor Shiyan Hu
Department of Electrical and Computer Engineering
Michigan Technological University 1 1

2. Smart Home: Academic Perspective
5% energy efficiency improvement in residential home energy systems leads to carbon emission reduction equivalent to removing 53 million cars in U.S. 2

3. Smart Home
To Minimize Expense, Balance Energy Usage and Maximize Renewable Energy Usage 3

4. Why we schedule? The Single User Smart Home 4 Power flow Internet
Control flow 4

5. Varying Energy Consumption
Typical summer energy load profile in State of Ontario, Canada. One can see the peak load around 7:00pm which usually involves a lot of human activities.
Peak
Average
PAR
Source: Ontario Energy Board 5

6. Dynamic Electricity Pricing
Set high prices at peak energy hours to discourage energy usage there for energy load balancing
Hourly Price from Ameren Illinois 6

7. Renewable Energy 7

8. Energy Scheduling for a Single Smart Home
Given the electricity pricing, to decide
when to launch a home appliance
at what power level
for how long
utilize renewable energy
subject to scheduling constraints
Targets
Reduce user bill
Reduce PAR (peak to average ratio) of grid energy usage
Maximize renewable energy usage
The smart home scheduler computes scheduling solutions for future, so it needs the future pricing. How? 8

9. Two Pricing Models: Guideline and Realtime Pricing
Guideline price: utility publishes it one day ahead to guide customers to schedule their appliances, through providing the predicted pricing in the next 24 hours.
Real time price: utility uses it to bill customers, e.g., it obtains the total energy consumption in the past hour, computes the total bill as a quadratic function of the total energy, and then distributes the bill to each customer proportionally. 9

10. Electric Vehicles (EV)
Powered by one or more Electric Motors 10

11. Multiple Mode Charging of EV
2014 Honda Accord PHEV
120-volt: less than 3 hours
240-volt: one hour
2013 Toyota Prius PHEV
240-volt: 1.5 hours
2014 Chevrolet Volt PHEV
120-volt: 10 – 16 hours
240-volt: 4 hours
Using mobile connector
29 miles of range per hour charge
The fastest way to charge at home
58 miles of range per hour charge 11

12. 12 Landry machine Dish washer EV AC Start End …… 13:00 18:00 09:00
08:00
17:00
N/A 12

13. 13 Multiple Power Level (VFD) Impact Power Powerr 5 cents/kwh2 1 2 3
Time
Time
(b)
(a)
cost = 10 kwh * 5 cents/kwh = 50 cents
cost = 5 kwh * 5 cents/kwh + 5 kwh * 3 cents/kwh = 40 cents 13

14. Uncertainty of Appliance Execution Time and Energy Consumption
In advanced laundry machine, time to do the laundry depends on the load. How to model it? 14

15. Problem Formulation
Given n home appliances, to schedule them for monetary expense minimization considering multiple power level considering variations
Solutions for continuous VFD/power level
Solutions for discrete VFD/power level
Solutions for continuous VFD
Solutions for discrete VFD 1 2 3 4 15

16. The Procedure of the Our Proposed Scheme
Offline Schedule
A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 16

17. The Outline A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Optimal Greedy based Deterministic Scheduling
Optimal DP based Deterministic Scheduling
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 17

18. Linear Programming for Deterministic Scheduling with Continuous Power Level
minimize:
subject to: 18

19. Max Load Constraint
To avoid tripping out, in every time window we have load constraint 19

20. Appliance Load Constraint
Sum up in each time window appliance power consumption is equal to its input total power 20

21. Appliance Speed Limit and Execution Period Constraint
The power is upper bounded
Appliance cannot be executed before its starting time or after its deadline 21

22. Power Resource
Power resource can be various 22

23. Solar Energy Distribution Constraint
Solar Energy can be directly used by home appliances or stored in the battery 23

24. Battery Energy Storage Constraint and Charging Cost
Solar Energy Storage
Battery Charging Cost 24

25. The Proposed Scheme Outline
A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Optimal Greedy based Deterministic Scheduling
Optimal DP based Deterministic Scheduling
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 25

26. Greedy based Deterministic Scheduling for Task i
Power t1 t2 t3 t4
Time
Price
Time
Cannot handle noninterruptible home appliances 26

27. The Proposed Scheme Outline
A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Optimal Greedy based Deterministic Scheduling
Optimal DP based Deterministic Scheduling
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 27

28. Dynamic Programming
Given a home appliance, one processes time interval one by one for all possibilities until the last time interval and choose the best solution
Choose the solution with total energy equal to E and minimal monetary cost 28

29. Characterizing
For a solution in time interval i, energy consumption e and cost c uniquely characterize its state
Time interval i
Time interval i+1
(ei, ci)
(ei+1, ci+1) 29

30. Pruning
For one time interval, (e1, c1) will dominate solution (e2, c2), if e1>= e2 and c1<= c2
Time interval i
(15, 20)
(15, 25)
(11, 22) 30

31. Dynamic Programming based Appliance Optimization
Power level: {1, 2, 3}
Dynamic Programming returns
optimal solution
(6, 9)
(5, 8)
(4, 7)
(5, 7)
(4, 6)
(3, 5)
(4, 5)
(3, 4)
(2, 3)
(3, 3)
(2, 2)
(1, 1)
(3,6)
(3,3)
Price
(2,4)
(2,2)
(1,2)
(1,1)
Time
(0,0) t1
(0,0) t2 31

32. DP based Deterministic Scheduling For Multiple Home Appliances
Determine Scheduling Appliances Order
An appliance
Schedule Current Home Appliance by DP
Not all the appliance(s)
processed
Update Upper Bound of Each Time Interval
All appliances are processed
Schedule 32

33. The Proposed Scheme Outline
A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Optimal Greedy based Deterministic Scheduling
Optimal DP based Deterministic Scheduling
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 33

34. Variation impacts the Scheme
Worst case design
It can be improved
Cost can
be reduced
Best Price
Window t1 t2 t3 t4 34

35. Best Case Design t1 t2 t3 t4 35

36. Variation Aware Design
An adaptation variable β is introduced to utilize the load variation. t1 t2 t3 t4 36

37. Uncertainty Aware Algorithm
Trip rate = trip out event / total event 37

38. The Design Flow Uncertainty Aware Algorithm38

39. Algorithmic Flow Input: Task set with tasks which can be scheduled
Core 1
up date task load based on β
Generate appliances schedule by solving the LP
Derive current trip rate using Monte Carlo simulation
Current trip rate ≤ Target
Update β No
Yes
Core 2
Core 3
Core 4
β from 0 to 0.25
β from 0.25 to 0.5
β from 0.5 to 0.75
β from 0.75 to 1
Yes
up date task load based on β
Generate appliances schedule by solving the LP
Derive current trip rate using Monte Carlo simulation
Current trip rate ≤ Target
Update β No
Output: Schedule 39

40. Algorithm Improvement
Monte Carlo Simulation takes 5000 samples
Latin Hypercube Sampling takes 200 samples
Latin Hypercube Sampling is a statistical method for generating a distribution of plausible collections of parameter values from a multidimensional distribution
Current S 40

41. The Proposed Scheme Outline
A deterministic scheduling with continuous power level
A deterministic scheduling with discrete power level
Optimal Greedy based Deterministic Scheduling
Optimal DP based Deterministic Scheduling
Stochastic Programming for Appliance Variations
Online Schedule for Renewable Energy Variations 41

42. Online Tuning Actual renewable energy < Expected
Utilize energy from the power grid
Actual renewable demand > Expected
Save the renewable energy as much as
possible
Actual renewable demand = Expected
Follow the offline schedule 42

43. Experimental Setup
The proposed scheme was implemented in C++ and tested on a Pentium Dual Core machine with 2.3 GHz T4500 CPU and 3GB main memory.
500 different task sets are used in the simulation. The number of appliances in each set ranges from 5 to 30, which is the typical number of household appliances [1].
Two sets of the KD P series PV modules from Inc [2] are taken to construct a solar station for a residential unit which are cost $502.
The battery cost is set to $75 [3] with 845 kW throughput is taken as energy storage.
The lifetime of the PV system is assumed to be 20 years [4].
Electricity pricing data released by Ameren Illinois Power Corporation [5]
[1] M. Pedrasa, T. Spooner, and I.MacGill, “Coordinated scheduling of residential distributed energy resources to optimize smart home energy services,” IEEE Transactions on Smart Grid, vol. 1, no. 2, pp. 134–144,2010.
[2] Data Sheet of KD P series PV modules, available at
[3] T. Givler and P. Lilienthal, “Using HOMER software, NRELs micropower optimization module, to explore the role of gen-sets in small solar power systems case study: Sri lanka,” Technical Report NREL/TP , 2005.
[4] Lifespan and Reliability of Solar Panel,available at
[5] Real-Time Price, available at 43

44. Experimental Setup on Weekday Using DP44

45. Energy Consumption Distribution on Weekday
Fig1. Energy consumption distribution comparison of Test Case I. (a) Traditional scheduling
(b) Dynamic Programming based scheduling. 45

46. Monetary Cost Distribution on Weekday
Fig2. Monetary cost comparison of Test Case I. (a) Traditional scheduling (b) Dynamic Programming
based scheduling. 46

47. Experimental Setup on Weekend Using DP47

48. Energy Consumption Distribution on Weekend
Fig3. Energy consumption distribution comparison of Test Case II. (a) Traditional scheduling
(b) Dynamic Programming based scheduling. 48

49. Monetary Cost Distribution on Weekend
Fig4. Monetary cost comparison of Test Case II. (a) Traditional scheduling (b) Dynamic Programming
based scheduling. 49

50. Experimental Results Using LP
Energy Cost (cents)
Runtime (s)
Cost
time
household appliances
household appliances 50

51. Traditional vs. LP vs. Discrete Greedy
Cost
Household appliances 51

52. Only DP Can Handle Non Interruptible Task set
Cost
Household appliances 52

53. Comparison of Worst Case, Best Case Design and Stochastic Design
Energy Cost (cents)
Trip Rate (%)
Cost
Rate
10 seconds
Household appliances
Household appliances 53

54. Online vs. Offline
Cost (cents)
Household appliances 54

55. Example of a Task Set 55

56. The Implementation Using FPGA56

57. Schematic of FPGA Implementation57

58. Summary
This project proposes a stochastic energy consumption scheduling algorithm based on the time-varying pricing information released by utility companies ahead of time.
Continuous power level and discrete power level are handled.
Simulation results show that the proposed energy consumption scheduling scheme achieves up to 53% monetary expenses reduction when compared to a nature greedy algorithm.
The results also demonstrate that when compared to a worst case design, the proposed design that considers the stochastic energy consumption patterns achieves up to 24% monetary expenses reduction without violating the target trip rate.
The proposed scheduling algorithm can always generate a monetary expense efficient operation schedule within 10 seconds. 58

59. Multiple Users in a Community59

60. Multiple Users
Pricing at 10:00am is cheap, so how about scheduling everything at that time?
Energy Accumlation
10:00am 60

61. Game Theory Based Scheduling61

62. Game Theory Based Scheduling
For every player in a game, there is a set of strategies and a payoff function which is the profit of the player.
Each player chooses from the set of strategies in order to maximize its payoff.
When no player can increase its payoff without decreasing other users’ payoff, Nash Equilibrium is reached. 62

63. Game Formulation in Community Level
Players: All users in the community
Strategy: Choose power levels and launch time to maximize payoff while satisfying constraints 63

64. Community Size Small community: Less than 100 users
Medium community: 100 ~5,000 users
Large community: More than 5,000 users 64

65. Small Community: Fully Distributed Architecture
In the fully distributed architecture, each customer uses own smart home scheduler to communicate with other users for information exchange and computes smart home scheduling solution. 65

66. Algorithmic Illustration For Small Community
Communication/Synchronization ……
Equilibrium/Schedule …
Embedded Processor
User2
Usern
User1
Iteration 1
Iteration 2 66

67. Algorithmic Flow For Small Community
Each user schedules their own appliances separately to maximize payoff using dynamic programming
Appliances
Determine scheduling appliances order
All users share information with each other
Schedule current appliance by dynamic programming
Each user reschedules their own appliances separately by dynamic programming No No
All appliances scheduled
Equilibrium
Yes
Yes
Schedule
Schedule 67

68. Single User Smart Home Scheduling
Baseline Energy Usage
Scheduling Range
Energy Usage
Scheduling Range
Energy Usage
Scheduling Range
Energy Usage 68

69. An Example for Two Customers
Total Energy Load
Change this
Customer 1
Customer 2 69

70. Problem With The Fully Distributed Architecture
Communication/synchronization problem
Assume that there are 100 iterations needed for the game theory based algorithm.
Communication/synchronization needs to be performed at the end of every iteration.
It is not realistic for big community to deploy the fully distributed architecture due to the complexity of synchronization among a large number of users.
Each user performs the game theory based algorithm at their own side and communicates with all other users after every single iteration. 70

71. Medium Community: Fully Centralized Architecture
Users only communicate with computer cluster twice, at the beginning and end.
Communication/synchronization is not needed any more among users.
Communication/synchronization within computers or CPU cores is much easier and faster.
Each user sends the scheduling tasks to a computer cluster which compute the scheduling solutions of all users. 71

72. Algorithmic Illustration For Medium Community
Parallel Computing …
Each core schedules assigned tasks of users in parallel
All cores share information with each other to synchronize
Each core reschedules the assigned tasks given the information of other users
Schedule
Equilibrium
Run iteratively until convergence
Interface
User1
User2
Usern 72

73. Algorithmic Flow For Medium Community
Solve the continuous fashion problem combinatorially
Discretize the continuous solution
Flag all computers to be available
Assign task fractionally to the available computer with lowest ratio of 𝒄/𝒇
Sort all computers increasingly by ratio of 𝒄/𝒇
Runtime of computer is reaching TC
Flag the computer to be unavailable
Yes No
Each computer runs tasks of users in parallel
All computers share information with each other to synchronize
Each computer reruns the tasks of users given the information of other users
Schedule
Equilibrium
Run iteratively
Users send tasks to computers
Schedule tasks of users to computers
Game theory based algorithm
Computers send back the results to users
# iterations = kϒ ……
User1
User2
User3
Usern

74. Problem With The Fully Centralized Architecture
Cannot handle large community
Communication delay
Limited computation power and high maintenance cost
Security concerns 74

75. Large Community: Hierarchical Architecture
There are 10 million users in a big community. It can be partitioned into 2k smaller groups, in which the number of users is 5k.
The communication overhead within each group is acceptable.
There is no flooding packets problem. 75

76. Algorithmic Flow For Intra-Community Optimization
Parallel Computing …
Each core schedules assigned tasks of users in parallel
All cores share information with each other to synchronize
Each core reschedules the assigned tasks given the information of other users
Schedule
Equilibrium
Run iteratively until convergence
Interface
User1
User2
Userx1
Continue to Inter-community optimization 76

77. Algorithmic Flow For Inter-Community Optimization
Energy consumption summation of Intra-community optimization
Pick k time intervals with the largest total energy consumption
Reduce the k energy consumption by δ
Pick k time intervals with the smallest total energy consumption
Increase the k energy consumption by δ
Continue to Intra-community optimization/Schedule 77

78. Algorithmic Illustration For Large Community
Parallel Computing …
Each core schedules assigned tasks of users in parallel
All cores share information with each other to synchronize
Each core reschedules the assigned tasks given the information of other users
Schedule
Equilibrium
Run iteratively until convergence
Interface
User1
User2
Userx1
Continue to Inter-community optimization
Energy consumption summation of Intra-community optimization
Pick k time intervals with the largest total energy consumption
Reduce the k energy consumption by δ
Continue to Intra-community optimization/Schedule
Pick k time intervals with the smallest total energy consumption
Increase the k energy consumption by δ 78

79. Community-level Solar Energy
All the customers in the building belong to the community and share the energy generated by the PV panel as a public resource. 79

80. Microgrid
A microgrid is a localized grouping of electricity generation, energy storage, and loads that normally operates connected to a traditional centralized grid. Microgrid generation resources can include fuel cells, wind, solar, or other energy sources. 80

81. Benefit of Renewable Energy Integration?
Mitigate the pressure of the peak generation.
Avoid pollution due to excessive generation.
Reduce the power flow of transmission and distribution system, ensure the security.

82. Challenges
Utility designs the predictive pricing. However, it only considers the electricity availability from utility.
Smart home scheduling (e.g., dynamic programming) only schedules the energy according to the predictive pricing. When renewable energy is available, how do smart home customers use it in smart home scheduling?
Suppose that the total available renewable energy is 100kWh, while the total energy demand is 200 kWh. How do we distribute the renewable energy resources?
Discount everybody’s bill by 50%?
What if a person only uses energy significantly at 8pm?
After merging, cannot physically distinguish utility energy and renewable energy.
We can only virtually distribute them to reward customers, but how much energy/reward should be assigned to a customer? 82

83. New Idea
Design an effective and efficient smart home scheduling method to virtually distribute energy and the associated reward to each customer.
If an aggregator decides the energy distribution, it intends to minimize the community-wide monetary cost, while expecting the monetary cost of each customer is also reduced. 83

84. Model of Community-level Renewable Energy
Power Line
Data Line 84

85. Energy Consumption & Monetary Cost
Power Line
Data Line 85

86. Problem Formulation 86

87. Constraints 87

88. All customers solve the multi-customer smart home scheduling problem
Algorithmic Flow No
Yes
Converge?
All customers solve the multi-customer smart home scheduling problem
End 88

89. 89

90. Experimental Setup 90

91. Wind Farm Data
Wind Power from 01/01/2014 to 01/05/2014, every 15 minutes, Forecast Error 9%
Scaled for 20% penetration 91

92. Background Energy 92

93. No Smart Home Scheduling93

94. Centralized Solution
$ over all the customers 94

95. Decentralized Solution
$ over all the customers 95

96. Energy Consumption During Cross Entropy Iteration96

97. Sell Energy Back to Grid
Home level renewable energy generation unit is encouraged such that customers could sell residual renewable energy back to the utilities.
Standard: Clarify how to connect renewable generation unit into the power grid.
Net Metering: Measure the power flow in both directions.
Selling Price: Partial the retail price or generation cost (varies in different locations).
Already applied in 27 states of U.S. 97

98. Net Metering
Net Meter
The power flow injected into the distribution network is measured by net meter as the selling back amount. 98

99. Making Smart Home Scheduling More Complex
PV panel
Power Line
Data Line 99

100. Summary Single user smart home scheduling
Variation aware optimization in smart home scheduling
Multiple user smart home scheduling
Considering renewable energy
Net metering