The Role of Modeling and Accounting in Estimation of Unbilled Revenue Presented by: J. Stuart McMenamin

Transcription

1 Electric / Gas / Water The Role of Modeling and Accounting in Estimation of Unbilled Revenue Presented by: J. Stuart McMenamin

2 Please Remember In order to help this session run smoothly, your phones are muted. If you would like to make the presentation portion of the screen larger, press the expand button on the toolbar. Press it again to return to regular window. If you need technical assistance during the meeting, dial *0 and you will be connected to a Premiere Conferencing technician. If you need to give other feedback to the presenter during the meeting, such as, slow down or need to get the presenters attention for some other reason, use the pull down menu in the seating chart and we will address it right away. If you have general questions regarding the presentation, please type your question in the Q&A box in the bottom, right corner. We will try to answer as many questions as we can at the end of the session.

3 Estimating Unbilled Energy and Revenue Utilities deliver energy continuously throughout the month. They know how much has been shipped (net generation), but they do not know who the buyer is until they read meters. They do not know the revenue until they read meters and calculate bills. To close the books at the end of a calendar month, utilities must estimate the revenue that goes with unbilled energy. For a given month, unbilled energy may be between 30% and 70% of the total, depending on the timing of read cycles and the timing of extreme weather. This presentation covers three topics: -- Modeling issues in estimating booked and unbilled energy -- How accounting uses the model results -- Comparison of two accounting methods (prior unbilled method vs. direct estimation of unbilled energy)

4 Estimating Unbilled Energy and Revenue The biggest source of uncertainty is from meters that are read early in the month. On a cycle chart, these early reads create the unbilled corner use on these days will be measured and billed in the following month.

5 Geometry of the Problem Billed Sales(m) =energy delivered in the revenue month parallelagram (A+B) Calendar Month Sales (m) = energy delivered in the calendar rectangle (B+C) Unbilled(m) = energy delivered in the unbilled triangle (C) Unbilled(m-1) = prior month unbilled triangle (A) A B C A C B

6 Constructing Cycle Month Variables Daily Average Temperature Cycle Data Days = 28.5 CDD = CDD/Day = 7.7 Calendar Month Days = 31, CDD = 370, CDD/Day = 11.9

7 Constructing Unbilled Corner Variables Daily Average Temperature

8 Monthly Model Estimation and Simulation Estimated Model: The model parameters are estimated by regressing billed sales on billing month inputs (billing days, billing cycle weather,...). Sales = a + b CycCDD + e m Cyc Pr edm = â + bˆ CycCDD m m m ê m = Sales m CycPr ed m Calendar Month Simulated Model: Simulate using the estimated parameters with calendar month inputs. CalSim m = â + bˆ CalCDD m Unbilled Corner Simulated Model: Simulate using the estimated parameters with unbilled corner inputs. Unb Simm = â + bˆ UnbCDD m

9 Issues with Monthly Models Modeling Issues for Monthly Models Plan to use the model to simulate usage in three ways Predicted values with cycle month estimation data Simulated values with calendar month weather and days Simulated values with unbilled weather and days Model must be structured so that the predicted values are scaled properly whether the number of days is 34 (cycle based), 31 (calendar month), or 15 (unbilled corner). Seasonal patterns (monthly binaries) need to be shifted. One solution is to build models on a per day basis. The dependent variable in the model is then: SalesPerDay(m) = Sales(m)/BillingDays(m) Weather variables should also be in per day form (HDD(m)/BillingDays(m)) Predicted values are comuted as: Predicted Sales = SalesPerDay Days Days can be 34, 31, or 15.

10 Treatment for Monthly Binary Variables Seasonality and Holidays create issues for Monthly Binary variables. Example: January calendar day events impact estimated binary coefficients for billing data in January and February. Solution: For January calendar month simulations, use half of the January binary coefficient and half of the February binary coefficient. This requires using two calendars, structured as follows. Use the calendar month binaries as over-ride variables in the simulations.

11 Setting up Calendar Month Calcs in MetrixND

12 Alternative Using Load Research Data Compute calendar month energy and unbilled energy using predicted or measured values for class profiles. Daily energy models can be used here to backcast daily usage. Profile Scaled to Billed Energy (m-1) Profile Scaled to Billed Energy (m)

13 How the Numbers Get Used Calendar Month Sales All methods start with cycle models for each class Sales, CycPred, e Then simulate sales for the calendar month (CalSim) Based on Calendar weather and calendar days Then use the results to estimate Calendar Month Sales for each class Delta Method Delta = CalSim - CycPred CalSales = Sales + Delta = CalSim + e Ratio Method (Booked-to-billed Ratio) Ratio = CalSim/CycPred CalSales = Sales Ratio = (CycPred + e) Ratio = CalSim + e Ratio Direct Modeling Method CalSales = CalSim Methods differ in treatment of the cycle month residual. Method 1 adds it, method 2 scales it, method 3 ignores it. Calendar estimates can be calibrated to system load A B C B

14 How the Numbers Get Used Revenue & Balance Sheet Balance Sheet: At the end of the month, the estimated revenue from unbilled energy appears on the balance sheet as an asset. Assets are: Cash Bills that have been paid Receivables Bills that have not been paid Unbilled accrual Estimated revenue from unbilled energy Income Statement: At the end of the month, income includes estimated revenues accruing from energy deliveries made during the month. Booked revenue is calculated as: B C = A - A B + Booked(m) = (Billed(m) Unbilled(m-1)) + Unbilled(m) (A+B-A)+C = Billed(m) + (Unbilled(m) Unbilled(m-1)) (A+B)+(C-A) = Billed(m) + Change in Unbilled The above is true no matter how unbilled energy is estimated. Unbilled revenue is called the unbilled accrual The Unbilled(m-1) effectively reverses the prior months unbilled estimate So...we need to estimate current month unbilled energy (C) to estimate revenue. C

15 Unbilled Energy (1) Prior Unbilled Method This method infers the current month unbilled value from other values that have already been computed. Models are not used to simulate energy use in the unbilled corner. C = C A A B - B + Unbilled(m) = CalSales(m) Billed(m) + Unbilled(m-1) B+C is the estimate of aggregate calendar month energy. A+B, the billed energy value, is the sum of customer bills on a billing cycle basis for the month. A, the prior month unbilled, is the estimate of unbilled energy that was made at the beginning of month m and that was used to close the books in month m-1. Another Perspective A+B is measured. Prior A implies B. B+C is estimated. Implied B implies C. A B C A+B=100 B+C=90 A=30 B=70 C=20

16 Unbilled Energy (2) Direct Method Direct estimation of the unbilled corner is based on use of models to simulate sales for the unbilled corner UnbilledSim = Model simulated value based on Unbilled HDD, Unbilled CDD, and Unbilled Days 1. Delta Method (Delta = Already Billed) Delta = CalSim - UnbilledSim Unbilled = CalSales - Delta 2. Ratio Method (Unbilled Fraction) Ratio = UnbilledSim/CalSim CalSales = CalSales Ratio 3. Direct Method Unbilled = UnbilledSim B C C Note methods 1 and 2 apply a delta or ratio to CalSales. CalSales may include a scaled residual (e) from the cycle month. In method 1, this gets carried forward. In method 2 it gets scaled by the unbilled fraction. In method 3, it is ignored.

17 Comparing the two methods Prior Unbilled Method A Close Prior Month Estimate A Direct Unbilled Method A B C B C Close Current Month A+B is now measured B+C is estimated Prior A implies B and C Close Current Month A+B is now measured B+C is estimated C is estimated C implies B and revised A With the Prior Unbilled approach, A is estimated before A+B is known. Any error estimating A passes on to C. Errors accumulate until a manual adjustment is made. With the Direct Unbilled approach, B+C is estimated after A+B is known. And C is estimated directly. The estimate for C implies a value for B, which implies a revised value for A. Revenue is adjusted accordingly and the unbilled accrual is correctly represented as an asset.

18 Example of the Comparison Prior Unbilled Method = Income Statement Rev = (20 30) = 90 Close Prior Month Estimate A = 30 Close Current Month A+B = 100 B+C = 90 (estimate) Balance Sheet Unbilled Accrual = 20 Note: Rev = Cal Sales = 90 Direct Unbilled Method Income Statement Rev = = 110 Close Current Month A+B = 100 B+C = 90 (estimate) C = 40 (estimate) Implies B = 50 Implies A = 50 Balance Sheet Unbilled Accrual = 40 Note: Rev =110 Cal Sales = 90

19 Checking the Unbilled Volumes using Zone Output Daily System Energy (GWh) Unbilled energy at the class level should be roughly consistent with an unbilled calculation using total daily system energy. This is a simple check to see if aggregate results are reasonable.

20 Conclusion This is all very confusing but... Forecasting has a strong role here. -- Models can be used to estimate BTB deltas or ratios -- Models can be used to estimate unbilled fractions or unbilled energy -- Models can be used to weather adjust booked sales -- Monthly models or more detailed daily profiling models can be used The Direct Unbilled method looks like the right thing to do. -- It is self correcting. -- It avoids accumulation of errors that will show in the balance sheet. -- It avoids the occasional need to make large adjustments -- If done right, it will be consistent with common sense checks

21 Questions? Press *1 to ask a question on the phone or type in the box at the bottom, right corner. Next Brown Bag Seminar: Short-term System Operations and Retail Forecasting Tuesday, Sept 12 Upcoming Itron Workshops and Meetings Fundamentals of Short-Term and Hourly Forecasting - September 14-15, San Diego, CA Itron Users Conference - October 15-17, Palm Desert, CA Forecasting November 6-8, San Diego, CA NEW Applied Load Research Workshop - November 13-14, San Diego, CA Contact us at: or forecasting@itron.com