ELECTRICAL LOAD SHEDDING REDUCES EXPENSES OF TEXAS WATERFLOOD

Forrest Collier, Eric N. Hardgrave
Mobil Exploration & Producing U.S. Inc.
Midland, Tex

Raymond J. Stanley, Keith Hatfield
R.I. Stanley & Associates Inc.
Dallas

An electrical load-shedding project at the Salt Creek field unit (SCFU) in Kent County, Tex., has reduced electrical expenses by $510,000 in the first year of implementation.

The load-shedding project was possible because of a dual demand electrical rate schedule. With this rate schedule, a large portion of the electrical demand charge is based on the SCFU load during the wholesale utility's monthly peak load. The goal is to interrupt SCFU electrical load at the utility's peak load thereby reducing the demand charge.

To effectively load shed, SCFU personnel must predict when the wholesale utility's monthly peak load will occur.

The utility's system load is highly dependent on the temperature in the utility's main load area. By closely monitoring the utility's load and weather data in the load area, the time of the utility's peak load can be consistently forecasted.

Currently, artificial lift installations and high-pressure injection pumps are interrupted for load-shedding purposes. Because the interruption of artificial lift installations results in deferred production, priority for interruption must be carefully assigned to wells so that lost revenue is minimized.

Limiting total interruption time and frequency during each billing period is also very important in minimizing lost revenue and cycling of lift equipment.

DUAL DEMAND RATE

Most large commercial electrical rates contain three cost categories:

Customer or facilities charge
Energy charge
Demand charge.

The customer or facilities charge is a monthly flat rate designed to recover the expenses associated with the customer's individual account services and metering facilities.

The energy charge is the cost for total consumption of electrical power used during the billing period. The energy charge is measured in kilowatt-hours (kw-hr) and is denoted by the area under the curve in Fig. la.

The demand charge is the charge for the utility company to provide capacity to meet the customer's needs at any time and is measured in kilowatts (kw). This charge is based on the highest average load required by the customer during any consecutive 15 min interval within the billing period. Some electrical rates may stipulate a different time interval and method for demand averaging. The electrical demand is denoted by the highest point on the curve in Fig. la.

The dual demand rate is a special power rate negotiated with the distribution cooperative that allows a direct pass-through of wholesale power costs from the wholesale power supplier to the customer.

The wholesale power costs are passed to the customer in the same rate form that is used to sell wholesale power to the distribution cooperative. The dual demand rate form has the same cost categories as the normal commercial rate except that the demand charge is divided into two components:

The noncoincident peak (NCP) demand charge
The coincident peak (CP) demand charge.

The NCP demand basis is the same as the demand basis for the normal commercial rate but the cost, under the dual demand rate for SCFU, has been discounted from $9.00 to $1.56/kw. The CP demand charge is an additional charge based on the wholesale utility's charges to the distribution cooperative.

The CP demand charge for each billing period is based on the customer's highest average 15 min load within the hour during which the wholesale utility's highest average hourly load for the billing period occurs.

For SCFU, the CP demand charge varies with the time of year, and the costs are $7.16/kw in summer billing periods (May-October) and $6.03/kw in winter billing periods (November-April). Fig. 1b illustrates the difference between the NCP and CP demand charges.

The CP demand charge can be reduced by forecasting the wholesale utility's average hourly peak load for the billing period and then shedding SCFU electrical load when the peak occurs.

A dual demand rate benefits the wholesale power supplier because it flattens electrical generation requirements by giving customers incentives to interrupt load during peak periods. Effective demand management allows the wholesale supplier to defer or eliminate costly power generation investments.

FORECASTING THE PEAK

To take full advantage of the dual demand rate, SCFU personnel must be able to accurately forecast the time of the wholesale utility's average hourly peak load for each billing period.

A key activity in forecasting the wholesale utility's peak load is the monitoring of current weather data in the utility's main load area, in and around Stephenville, Tex.

The system load of the wholesale utility serving SCFU is driven by residential load. These residential loads are very predictable because they are primarily dependent on the weather (especially temperature) in the Stephenville area. Fig. 2 illustrates the percent of the wholesale utility's total system load that is dependent on the weather for a typical summer day.

Historical load data for the wholesale utility is also very important in peak load forecasting. Fig. 3 illustrates typical load shapes for past summer, winter, and transitional billing periods. Load magnitudes are highest in the summer and winter months when temperatures are most severe and the use of air conditioning or space heating is increased.

Historical load shapes are used to forecast both the time of day that the peak load will occur and the magnitude of the peak load.

To facilitate the forecasting of the wholesale utility's peak loads, a load-forecasting computer model was developed for the SCFU load-shedding program. The computer model is a data base containing the last 4 years of hourly temperature and load data from the Stephenville weather station and the wholesale power utility.

When furnished with temperature and time of year information, the computer model searches the data base and develops a subset data base which exhibits all the characteristics defined by the user. The model then calculates statistics based on the subset data base and provides the user with a set of columnar and graphical data for that day's forecasted load shape.

Using the statistical data, the model will forecast the time and magnitude of the daily peak load when provided with actual load data for the day being monitored.

Fig. 4 shows a typical output sheet from the computer model. The first three columns on the right-hand side numerically represent the typical load shape for the temperature and time of year input criteria. As actual hourly loads from the wholesale utility become available during the day being monitored, the loads are input into Column 4.

The forecasted peak load for that day is then calculated in Column 5 with an upper and lower boundary in Columns 6 and 7.

Fig. 4 is the average load shape representing the temperature and time of year input parameters.

FORECASTING PROBLEMS

Forecasting the wholesale utility's peak load can be very tricky.

One problem is that during a load-shedding operation, the utility's peak shifts because of the interruption of too much load. This is best illustrated by the following example.

Assume that the wholesale utility's daily peak load is forecast between 7 and 8 p.m. and the load-shedding team interrupts 15,000 kw of SCFU load starting at 7 p.m. Also, assume that the wholesale utility's average load from 6 to 7 p.m. was 840,000 kw while the load from 7 to 8 p.m. was 830,000 kw.

If the 15,000 kw had not been interrupted, the peak load for this hypothetical day would have been 845,000 kw between 7 and 8 p.m. However, the peak load is shifted to 6 7 p.m. because 15,000 kw of load is removed from. the system from 7 to 8 p.m.

This problem is most likely to occur in the months with higher temperatures because the utility's average hourly loads around the normal peak period are close in magnitude (Fig. 3a). Chances of shifting the utility peak can be minimized by interrupting the desired load through the entire potential peak period or interrupting smaller magnitudes of load during the hour in which the peak is forecasted.

Another problem of forecasting is changing weather conditions during a potential peak load day. An example is a cold front during winter. If the cold front passes through the load center during and after the normal peaking period, the utility load may continue to increase throughout the day rather than form the usual spike and subsequent decrease.

Similarly, in summer months, a thunderstorm passing through the load center during the peak period may suddenly decrease the temperature thereby reducing the utility load by a significant amount.

As more industries use the dual demand rate on the same electrical system, forecasting increases in difficulty because wholesale utility load interruption increases during peak days.

With only one operation forecasting and shedding load, wholesale utility loads are dependent primarily on the temperature in the utility load center. With several operations forecasting and shedding load, utility loads become dependent on temperature and on the magnitude of load interruption that occurs on the system.

The additional variable makes accurate forecasting more difficult.

LOAD-SHEDDING PROCESS

Prior to each billing period, the load-shedding team discusses unique characteristics of that billing period. From historical weather and load data, the team determines the portion of the billing period and the time of day that the wholesale utility's hourly peak load is likely to occur.

The team also predicts the approximate magnitude of the peak load and the temperature in Stephenville at which the peak load is likely to occur. Each of these parameters will differ for each billing period in a given year.

Fig. 3 illustrate how different months require different strategies. For example, Fig. 3a indicates that the wholesale utility's peak load in the summer months will occur sometime between 4 and 8 p.m. Conversely Fig. 3b indicates that in winter months the peak will be in the morning between 6 and 9 a.m.

After a monthly strategy is formulated, a load-shedding technician is assigned to monitor daily weather forecasts for Stephenville. If a peaking-type day is projected, pertinent field personnel will be notified of a potential load-shedding interruption.

As the likely time of the system peak approaches, the technician will rigorously monitor Stephenville temperatures and wholesale utility loads. Pertinent data will be input into the load-forecasting computer model to further refine the forecast of the peak load.

If the technician believes that a system peak load will occur, proper operations personnel will be notified so that the SCFU electrical load can be promptly interrupted during the hour of the peak. After the peak load has occurred, the technician will expedite the start-up of the affected load.

The majority of the affected load can be turned on and off by the load-shedding technician from the central production office. A typical load-shedding interruption will last 2-4 hr. A typical billing period will have zero to four interruptions.

All interrupted SCFU electrical loads are thoroughly documented for future analysis. The technician also monitors and documents deferred production, artificial lift installation failures resulting from load shedding, and other important data.

SCFU ELECTRICAL LOAD

Electrical loads at SCFU consist of the following equipment:

Artificial lift units (electric submersible pumps and beam pumps)
High-pressure injection pumps
Vapor recovery units
Product/transfer pumps
Miscellaneous small loads such as office needs.

The electric submersible pump (ESP) load accounts for 73% of the total demand at SCFU. ESP motors at SCFU range in size from 50 to 560 hp.

The high pressure injection pump load accounts for 20% of the demand. Currently, six injection pumps, all driven by 1,500 hp motors, are active. The majority of other, less significant loads, such as beam-pump installations, vapor recovery units, and product/transfer pumps, range in size from 1 to 100 hp.

Because of the large contribution to the total electrical demand at SCFU, ESP and injection pump loads were initially targeted for load shedding. Beam pump installations were added to the load-shedding strategy after 10 billing periods.

ARTIFICIAL LIFT INSTALLATIONS

Interruption of artificial lift installations for load-shedding purposes has two major concerns:

Lost revenue resulting from deferred production.
Potential for higher maintenance costs from excessive cycling of equipment.

The key to lessen both of these problems is to minimize interruption time and frequency.

When the evaluation of artificial lift load shedding began, the most critical unknown was how many interruptions would be necessary to assure that the CP demand was reduced significantly. As interruption time and, therefore, artificial lift downtime increases, lost revenue from deferred production also increases.

Because of the increased cycling of lift equipment, frequent interruption will impact surface and subsurface maintenance costs. For each producing well, a single, unique interruption time per month is associated with the break-even point at which potential electrical cost savings equals lost revenue plus increased maintenance costs.

The key to successful load shedding of producing wells is to keep the total interruption time and frequency well below the break-even time.

Fig. 5 illustrates the potential profit and risk of load shedding 10,000 kw of artificial lift load at various interruption times. The financial risk of load shedding is the lost revenue associated with deferred production and is represented by the dotted line below the break-even line.

In calculating the financial risk of load shedding, it is assumed that the oil price is $20/barrel and that maintenance expenses are not affected by load shedding. Also, production is assumed to be deferred at the same rate that it is produced.

In other words, a well producing 5 bo/hr will lose 15 bbl if it is interrupted for 3 hr.

The financial risk is the profit lost if load shedding does not reduce electrical expenses despite the interruption of load.

The two lines above the break-even line in Fig. 5 represent the estimated net profit from successful load shedding. The net profit is the difference between the electrical savings and the financial risk of load shedding.

Net profit for summer billing periods is higher than that of winter billing periods because the CP demand charge is $7.16 and $6,03/kw for summer and winter months, respectively.

Although the above assumptions are somewhat simplified, Fig. 5 illustrates an important point. After 4 hr of interruption, the net profit from successful load shedding is $49,300-60,500/month while the financial risk of interrupting production is only $11,100/month. However, at 16 hr of interruption time, net profit is only $18,50030,000/month while the financial risk is $44,600/month.

Therefore, it is very important to minimize interruption time so that the ratio of potential net profit to financial risk is as high as possible. Prior to beginning the SCFU load-shedding program, an accurate estimation of interruption time and frequency was necessary to properly evaluate project economics.

To estimate the expected monthly interruption time and frequency, a mock load-shedding exercise was conceived. The mock exercise involved six past billing periods from 1991 and was conducted over a 2 day period ' Although the six billing periods had already passed, the exercise was designed so that the mock load-shedding technician would have no prior knowledge of the hour in which the monthly utility peak loads occurred.

The mock load-shedding technician was provided with all information necessary to forecast peak load occurrences. This information included hourly load data from the wholesale utility, weather forecasts, and actual temperature data from Stephenville.

For each day of the six billing periods, the mock load-shedding technician reviewed the weather forecasts and temperature/load data, input parameters into the load-forecasting computer model, and made artificial decisions on whether to interrupt operations. Because the mock load-shedding technician had no prior knowledge of peak load occurrences, the artificial decision to interrupt operations was biased only by the weather/load data and the output of the forecasting model.

The main conclusions from the mock load-shedding exercise were:

Zero to 10 hr/month of interruption time could be expected from a load-shedding program.
Zero to four interruptions/month could be expected from load shedding.
The monthly CP demand charge could be significantly reduced in 50-75% of the billing periods.

From the simulation, it became apparent that there would be months in which the CP demand would not be reduced despite interruptions of SCFU load. Conversely, there also would be months in which the CP demand would be reduced with only 2 hr of interruption time.

With only 0-10 hr/month of interruption time, lost revenue (at $20/bbl) from interruptions is well below potential electrical savings. Also, an interruption frequency of only zero to four times per month requires minimal cycling of lift installations.

Therefore, it is assumed that surface and subsurface maintenance costs are not impacted significantly.

INJECTION PUMP LOAD

Interruption of injection pump load has the following problems and concerns:

Storage of water during load-shedding interruptions
Nuisance shutdowns of engine-driven injection pumps
Deferred production resulting from disruption of target injection rates
Additional pump/motor maintenance resulting from excessive cycling.

The problem of storing produced water during a load-shedding operation is partially offset by interrupting high water producing lift installations at the same time that injection pumps are interrupted. This reduces the volume of produced water entering the injection facility and partially negates the need for water Storage.

During a typical load-shedding interruption, producing wells are normally interrupted anyway because significant electrical savings are possible with the load shedding of lift installations. Even without a reduction of produced water entering the injection facility adequate storage capacity is available to handle water during a load-shedding interruption.

Another concern associated with water storage is the ability to inject stored water after an interruption. The electric-driven injection pumps are centrifugal pumps; therefore, some stored water is injected because of increased injection rates resulting from lower field injection pressure after an interruption. When necessary, existing standby pump capacity is sufficient to remove the remainder of the stored water.

Using standby pump capacity to inject stored water could potentially increase the NCP demand charge; however, at an NCP demand charge of only $1.56/kw, the additional expense is minimal.

In addition to the electric-driven injection pumps, the SCFU injection system consists of four engine-driven, centrifugal injection pumps. When the electric-driven injection pumps are interrupted for load-shedding purposes, the SCFU field injection pressure decreases significantly. When the field pressure decreases, injection rates of the engine-driven pumps increase and high flow shut-downs occur. This problem is managed by manually imposing artificial head on the engine-driven pumps during load-shedding operations so that flow rates will stay within the operating limits.

Deferred production from disruption of target injection rates during load-shedding operations is another concern because of high formation communication between offsetting well bores.

With an interruption time of 10 hr/month, target injection rates are disrupted a maximum of only 1.4%. This deferred injection is partially offset by the injection of stored water after an interruption. For these reasons, it is assumed that overall production deferral from injection interruptions is minimal. It is also assumed that maintenance because of increased cycling of injection pumps will not increase noticeably with zero to four interruptions/month.

PROGRAM RESULTS

The SCFU load-shedding program began Oct. 21, 1991, and has been in progress for 12 billing periods, as of this writing.

Fig. 6 shows the effect of the load-shedding program on the SCFU electrical bill.

The gap between the average demand line and the CP demand line is the total reduction in CP demand after load-shedding started in October 1991.

Table 1 summarizes electrical billing data and results of the SCFU load-shedding program from Oct. 21, 1991, to Oct. 20, 1992.

For the first year, the CP demand charge was successfully reduced in nine billing periods.

Total electrical cost reduction for the first year was approximately $510,000 while estimated revenue reduction and operating costs were $150,000.

Therefore, net profit from load shedding was $360,000.

The effective reduction of the CP demand was estimated by subtracting the actual CP demand from the average demand for each billing period. For the first year of load shedding, average interruption time ranged from 0 to 10 hr/billing period while interruption frequency ranged from zero to four times per billing period.

In the two unsuccessful winter billing periods (Nov. 21 to Dec. 20 and Jan. 21 to Feb. 20), no interruptions occurred because the strategy was to wait for low temperatures that never occurred. And, in the unsuccessful summer billing period (July 21 to Aug. 20), no interruption occurred because the strategy was to wait for high temperatures that never occurred.

DEFERRED PRODUCTION

Deferred production because of load-shedding interruptions was analyzed and quantified by two different methods.

Method 1 compares potential production on load interruption days with potential production on normal, noninterruption days.

Daily potential production is defined as the production possible if all wells at SCFU were producing for the 24-hr period. This number is generated by adding downtime production to actual measured daily production.

Potential production on each load-shedding interruption day is compared to the production potential on the closest noninterruption day (usually the prior day). Potential production on each interruption day was also compared to the average of potential production for the 7 preceding noninterruption days.

Method 2 is based on the assumption that the rate of deferred production for an interrupted well is equal to the normal producing rate. In other words, a well producing 5 bo/hr will defer 15 bbl if it is interrupted for 3 hr. Normal producing rates of each well are determined from monthly well tests.

With Method 2, total deferred production because of load-shedding interruption is the sum of production deferred from each well that is interrupted. It is also assumed, with Method 2, that interruption of water injection will have minimal impact on total production.

Method 1 is considered the more accurate of the two because it involves the measurement of actual daily production rates and accounts for the effects of water injection interruptions.

The calculation of deferred revenue in Table 1 is based on deferred production estimates from Method 1.

Table 2 displays the estimated production change resulting from load shedding for the 18 days in which a load-shedding interruption occurred. Also listed in the table are the average interruption times and the number of interrupted ESPS, beam pump installations, and injection pumps.

In Table 2, a negative production change indicates that the load-shedding interruption resulted in deferred production. A positive production change indicates that production increased because of the interruption.

Production increases are indicated on 5 of the 18 interruption days. These increases are primarily the result of normal production fluctuations. Otherwise, the increased production on the 5 days is unexplained.

It is very unlikely that a load-shedding interruption would result in increased production rates; therefore, if a positive production change is indicated, it is assumed (for project economics and otherwise) that the production change is zero.

The listed production change for each interruption day in Table 2 is estimated from Methods 1 and 2. Production change estimates from Method 1 include a comparison of interruption days with the preceding noninterruption day and with the average of the preceding 7 noninterruption days.

Average production change for the two Method 1 comparisons are within approximately 2%. The estimate of total deferred production from Method 1 is approximately twice as high as that of Method 2.

The magnitude and rate of deferred production per well are highly dependent on the number of interrupted wells and the interruption time of each well. Prior to a load-shedding interruption, producing wells are prioritized for interruption according to the potential net profit from load shedding.

Because lower priority wells generally have higher oil cuts, more prolific production is impacted as more wells are interrupted.

Table 2 indicates that the magnitude of deferred production usually increases noticeably when more than 11 producing wells are interrupted.

The rate of production deferred per well will also increase with interruption time because of increased well bore loading. As well bore loading time increases, the time it takes for a well to recover to normal production will also increase.

The occurrence of well bore loading during load-shedding interruptions is supported by the fact that deferred production estimates from Method 1 are twice as high as Method 2.

Three load-shedding operations involved the interruption of injection pumps only. Positive production changes are shown on two of these days. Although inconclusive, the data support the assumption that injection interruptions have a minor effect on production.

The magnitude of deferred production during load-shedding interruptions is very difficult to accurately quantify. Normal production fluctuations mask the effect of load-shedding interruptions on total production. As more production data become available in the upcoming years, a more reliable statistical method of analysis will be possible.

MAINTENANCE COST

To date, no ESP or injection pump failures can be directly attributed to the load-shedding program. Increased cycling of rotating equipment is detrimental to equipment life; however, analysis of failure data indicates no noticeable increases after the load-shedding program began.

OPERATING COSTS

The initial cost required to start the SCFU load-shedding program was about $40,000. This included:

Development of the load-forecasting model
Computer hardware and other software necessary for load shedding
Development of historical temperature and load information.
Minimum operating costs required for the SCFU load-shedding program include:
Time required from the load-shedding technician and operating personnel
Acquisition of weather forecasts and data
Periodic updating of the load-forecasting model and other historical data.

The load-shedding technician and other operating personnel spend 1-15 hr/week on monitoring actual loads and weather data, forecasting the wholesale utility's peaks, and expediting load-shedding operations when necessary.

Weather data are accessed daily via a computer weather network that costs $400-500/month. Annual updates of forecasting software and supporting data are projected to cost $5,000/year.