"BILLING DEMAND: Either (a) or (b), whichever is higher:
(a) The MAXIMUM DEMAND occurring during the BILLING PERIOD.
(b) 65 percent of the highest MAXIMUM DEMAND established for the BILLS rendered for June, July, August or September of the preceding 11 months."
Plating Power Optimizer ยท The Learning Guide
How $325k of utility savings actually works โ explained from zero.
Nine short lessons. Each one introduces ONE concept, gives you a real-world analogy, shows you a worked example, then moves on. By the end, you can explain the whole model to anyone in five minutes.
Your electric bill isn't one number. It's three.
When the LES bill shows up, it shows one total. But underneath, that total is built from three completely different things that the utility charges you for.
Demand charge โ 70%
Energy โ 23%
Other โ 7%
Energy charge
What you pay per unit of electricity you actually use. Measured in kilowatt-hours (kWh). LES charges about 2.5 cents per kWh.
If you run a 1,000-watt heater for one hour, that's 1 kWh. Cost: 2.5ยข.
Demand charge
What you pay for the biggest amount of electricity you used at once during the month, measured in kilowatts (kW). LES charges $25.17 per kW for industrial customers. Even if you only hit that peak for 30 minutes, you pay for the whole month at that level.
If your peak was 1,200 kW for half an hour on July 15th, your demand charge that month is 1,200 ร $25.17 = $30,204 โ regardless of how little you used the other 29 days.
Other charges
A $450/month flat customer charge just for being connected, plus a penalty if your power factor is bad (more on that in Lesson 4).
Analogy โ phone bill vs gym membership
Energy charge = your phone's per-minute rate. You pay for what you use.
Demand charge = a gym membership priced by your heaviest single lift. They have to build the gym to handle your maximum effort, so they charge you based on that maximum โ not your average.
Why this matters: 70% of the bill is the demand charge. So if you want to save real money, you can't just be smarter about when you use electricity โ you have to be smarter about how much you ever pull at one time.
"Demand" doesn't mean total electricity. It means your worst 30 minutes.
The utility company drops a meter on your plant that records how much power you're pulling, every 30 minutes, all month long. They throw away the average. They only care about the maximum.
30-minute integration window
LES averages your kilowatt draw over rolling 30-minute periods. A 5-second spike doesn't count โ the meter smooths it out. But a sustained heavy load for 30 minutes? That's your peak for the month.
Worked example โ the 7am pile-on
A normal day at a small plant. Most of the time you're pulling 200 kW. But at 7:00 AM everyone fires up their equipment at the same time: 4 plating lines + 2 rectifiers + chiller + air compressor running together for 30 straight minutes.
30-minute peak (worst window)1,200 kW
Demand rate (LLP-15)ร$25.17/kW
Demand charge for the month$30,204
You owe that even if you never hit close to 1,200 kW again the rest of the month.
The fix: Stagger the startups across 20-30 minutes. The 30-minute peak drops to, say, 700 kW. Same equipment, same total energy used. Recomputing the bill:
Same day, staggered startup
New 30-minute peak700 kW
Demand rateร$25.17/kW
New demand charge=$17,619
Saved this month$30,204 โ $17,619$12,585
One bad summer day haunts you for 11 months. (The summer ratchet.)
This is the single most important concept in the whole project. If you take ONE thing away from this guide, take this.
Summer demand ratchet
A clause in the LES tariff that says: every monthly bill's "billing demand" is the larger of (a) that month's actual peak, OR (b) 65% of the worst peak you had between June and September of the preceding 11 months.
Translation: once you hit a high peak in summer, you can't escape paying for it. The utility "ratchets up" your minimum billing demand for almost a full year.
Verbatim from LES 2026 Rate Schedules PDF ยท Schedule LLP, sheet 2 of 3 ยท page 11
Analogy โ credit score that goes one way
Imagine your credit score could only go DOWN, not up, and it took 11 months to recover. One late payment in July means worse interest rates from August through next May. That's the ratchet.
The utility makes you pay for the worst version of yourself, even after you've cleaned up your act.
Worked example โ the $76,000 mistake
July 15, 2pm: A hot afternoon. Operators come back from lunch and fire up everything. The plant pulls 1,200 kW for 30 minutes. July's demand charge: 1,200 ร $25.17 = $30,204.
August through next May: You're very careful. Actual peak each month is only 400 kW. But the ratchet floor kicks in:
Last summer's worst peak1,200 kW
Ratchet floor1,200 ร 65%780 kW
This month's actual peak400 kW
Billing demand (larger of the two)max(400, 780)780 kW
Extra kW you're billed for780 โ 400380 kW
Penalty per winter month380 ร $25.17$9,565
Across 8 winter months$9,565 ร 8$76,520
That's $76,520 you pay for electricity you never actually used โ all because of one 30-minute window in July.
Now flip it. If you prevent that 1,200 kW peak in the first place โ staggered startups, smarter scheduling, lower simultaneous load โ the ratchet floor drops in lock-step. Stop one summer event, save almost a year's worth of money.
This is why the scheduler exists. Its job isn't really to find cheap hours. It's to prevent your operators from accidentally setting a summer peak that haunts you for the next 11 bills.
Power factor = how efficient your equipment is at using the electricity it pulls.
This is the most physics-y lesson. Don't worry about the equations โ focus on the analogy.
Analogy โ pushing a shopping cart
Imagine pushing a shopping cart through a parking lot. If you push it straight forward, all your effort goes into moving the cart. Power factor = 1.0 (perfect).
But if you push at an angle, only part of your push moves the cart forward โ the rest is wasted, pushing the cart sideways into nothing. Power factor < 1.0 (wasted effort).
The utility has to deliver enough electricity for BOTH the forward push AND the sideways waste. They charge you a penalty if too much of your push is sideways.
Real power (kW)
The "forward push." Actually does work โ heats metal, spins motors, runs pumps. This is what your equipment is FOR.
Reactive power (kVAR)
The "sideways push." Doesn't do useful work, but flows back and forth between your equipment and the grid. Heavy motors, transformers, and old SCR rectifiers cause a lot of this.
Power factor (PF)
A number from 0 to 1 that tells you how much of the electricity is "forward push" vs "sideways waste." Higher = better. LES penalizes anything below 0.93 at $2.60 per excess kVAR.
Old SCR rectifiers typically run at PF = 0.75 (lots of wasted push). Modern IGBT rectifiers run at 0.95+ (almost all forward).
Verbatim from LES 2026 Rate Schedules PDF ยท Schedule LLP, sheet 3 of 3 ยท page 12
"EXCESS kVARS: Maximum kVARS for the BILLING PERIOD minus the product of the MAXIMUM DEMAND for the BILLING PERIOD multiplied by .39523 (representing a power factor of 93 percent). The CUSTOMER will be responsible for limiting kVAR requirements at or below this calculated level at all times and will be charged for EXCESS kVARS."
Worked example โ a 1,200 kW load at 0.75 PF
Peak demand this month1,200 kW
Actual reactive power1,200 ร tan(arccos(0.75))1,058 kVAR
Allowed reactive (the LES limit)1,200 ร 0.39523474 kVAR
Excess (what you get billed for)1,058 โ 474584 kVAR
Monthly penalty584 ร $2.60$1,518
Annual penalty$1,518 ร 12$18,224
The fix: Add a $18,000 capacitor bank (it absorbs the sideways push). Penalty drops to zero. The capacitor pays for itself in 12 months.
Note: Only LLP (large light & power) customers get charged for power factor. GSD (smaller commercial) customers don't see this penalty.
Rectifiers are the big DC power supplies that drive plating tanks.
Electroplating needs direct current (DC) electricity โ like a battery โ to deposit metal evenly. But the grid delivers alternating current (AC). A rectifier is the box that converts AC โ DC. Every plating line has one. They're often the single biggest electrical load in the shop.
SCR rectifier (old)
Efficiency~80%
Power factor~0.75
Output qualityChoppy, 5-10% ripple
Failure modeWhole line down
Plating cycle timeStandard
IGBT rectifier (modern)
Efficiency~92%
Power factor~0.95+
Output qualitySmooth, <1% ripple
Failure modeModular, hot-swap
Plating cycle timeUp to 40% shorter
Analogy โ carburetor vs fuel injection
SCR rectifier = old carbureted car engine. Wastes fuel, runs rough, struggles at low RPM.
IGBT rectifier = modern fuel-injected engine. More efficient, runs smooth, no warm-up needed.
Both move you the same distance. One just uses less fuel and runs cleaner.
Worked example โ upgrading one 1,000 kW DC plating line
The DC output stays the same (the plating tanks need the same DC power either way). What changes is how much AC the rectifier has to pull from the grid.
SCR draws from grid (AC)1,000 รท 0.801,250 kW
IGBT draws from grid (AC)1,000 รท 0.921,087 kW
Reduction in AC draw1,250 โ 1,087163 kW
Energy saved per year163 kW ร 6,300 hrs ร $0.0245$25,166
Demand saved per year163 kW ร $25.17 ร 12$49,242
PF penalty avoided (IGBT fixes PF too)$18,975
Total annual savings per line$25,166 + $49,242 + $18,975$93,400
Capex per line$50,000
Payback$50,000 รท ($93,400 รท 12)6.4 months
The scheduler is a robot operator with one job: never let the peak hit summer-ratchet territory.
Think of the scheduler as a layer that sits between operators and the plating equipment. Operators say "we need to run jobs A, B, C, D today." The scheduler figures out when each job starts, in what order, with how much overlap.
Its objective function is simple to describe:
- Cap the daily peak. Don't let total instantaneous demand exceed a target (say, 700 kW). Stagger startups, sequence jobs across lines.
- Especially in summer. A summer peak ratchets the next 11 bills, so summer minutes are worth ~8ร more than winter minutes.
- Shift continuous loads to off-peak. Run the chiller, DI/RO water plant, and air compressors at night when there's headroom โ using thermal storage and water tanks as buffers.
- Respect line capacity and deadlines. Don't break the workflow. Jobs still have to finish on time.
Analogy โ a smart dishwasher timer
Some dishwashers can be programmed to start at 3am instead of 7pm so they don't compete with your electric oven during peak hours. The plating scheduler is that, but for an industrial site with 4-10 simultaneous "dishwashers" running.
What the scheduler does NOT do: It doesn't chase wholesale electricity prices. For retail customers on LES, energy is essentially flat ($0.0245 winter, $0.0255 summer). Saving 1 cent/kWh by running at midnight isn't worth tripping a summer ratchet. The scheduler's whole job is preventing peaks, not surfing prices.
Where AI actually shows up in this project (and where it doesn't).
The honest answer trips a lot of people up, so let's get it straight.
What is and isn't AI
The $325k savings number is NOT from AI. It comes from straightforward arithmetic against the verified 2026 LES tariff. Multiplication and addition. No machine learning involved in the savings math itself.
So where is AI? Three places:
1. Machine-learning model that forecasts SPP wholesale electricity prices (V1/V2). An XGBoost model trained on 16 months of real SPP prices. Predicts the next 48 hours of wholesale rates within ~17% accuracy. This model is parked in the database, not used in V3. Why? Because retail customers don't see wholesale prices. The forecast becomes useful again in V4 for facilities that enroll in SPP demand-response programs (where they get paid for curtailing during grid emergencies).
2. Scheduling optimization (LP). The "Linear Program" scheduler is an optimization algorithm (linear programming). Some people call this AI. Strictly it's operations research โ same math used in airline crew scheduling, factory production lines, route planning. It's not learning from data; it's solving a constraint problem.
3. The build process itself. Claude (this assistant) read the LES tariff PDF, synthesized industry research on demand ratchets and PF penalties, wrote the cost model, ran tests, caught bugs, generated this learning page. That's where "AI helped build this fast" is honest.
Honest framing
"AI helped build this analysis in days instead of months. The savings number itself is arithmetic on the verified tariff. There's a machine-learning component for wholesale price forecasting, but it's parked because retail customers don't qualify for those rates. If a facility ever enrolls in demand response, that forecast becomes the revenue model."
Why "$325k" is the model number, not necessarily any specific plant's number.
The math is verified against the LES tariff. The plant inputs are sized for a generic mid-size finisher. To get a specific number for a specific facility, those inputs need to be swapped for real data. Here's the honest split.
Verified โ same for every customer on LES
| Element | Source |
|---|---|
| $25.17/kW combined demand rate (LLP-15) | LES 2026 Rate Schedules PDF, sheet 2 of 3 (page 11) |
| 65% summer ratchet, 30-min integration | LES tariff Schedule LLP, sheet 2, BILLING DEMAND definition |
| $2.60/kVAR penalty above 0.93 PF | LES tariff Schedule LLP, sheet 3 of 3 (page 12) |
| 0.39523 multiplier (= tan(arccos(0.93))) | Same sheet, EXCESS kVARS definition |
| $450 customer charge + seasonal energy rates | Same PDF rate table |
| LES SEP grant amounts (EMS, VFD, compressors) | LES Sustainable Energy Program 2026 PDF |
| SCR vs IGBT efficiency physics | Standard industrial electrical engineering |
Assumed โ would need swapping for a specific plant's number
| Element | What's in the model | How to verify for your facility |
|---|---|---|
| Rate class | LLP-15 Secondary | Check any monthly bill. Could be LLP-16 Primary ($23.15/kW) or LLP-39 35 kV ($19.65/kW) |
| Daily plating energy | 5 MWh/day | 12-month average from utility bills |
| Continuous load | 650 kW (chiller + DI/RO + air) | Walk-around inventory + nameplate ratings |
| Number of plating lines | 2 | Actual count on the floor |
| Rectifier inventory | 1,000 kW DC total, all SCR, 80% efficient | Maintenance records, nameplate data |
| Current power factor | 0.75 (typical for SCR-heavy load) | Request 12 months of kVAR data from the utility |
| Current operating habits | Everything fires up at 7am | Compare interval data to actual operating logs |
Three concrete pieces of homework take the model from generic to plant-specific:
- Pull 12 months of interval data from the utility (30-minute kW, kVAR, kVA readings). Free.
- Walk the floor and record rectifier nameplate kW, manufacturer, and install year per line.
- Read one utility bill to confirm the active rate class.
With those three inputs, the model produces a facility-specific number instead of a generic one. Until then, the honest framing is a range โ the modeled $325k might land anywhere from $150k (smaller plant, fewer rectifier candidates) to $500k+ (larger plant or worse current state).
Other LES rate options worth knowing about
LLP-TOUD (Time-of-Use Demand, Rates 30/31/32): split demand into on-peak ($13/kW) and off-peak ($4.50/kW) charges. Useful for plants that can genuinely shift load to off-peak hours (12am-12pm and 9pm-12am in summer). Same 65% ratchet still applies.
LLP-OPD (Off-Peak Daily, Rates 27/28/29): designed for plants whose biggest demand happens during off-peak. Adds a penalty if off-peak demand exceeds the summer base. Rarely useful for day-shift finishers.
Curtailment Rider: separate program that pays customers $2.40/kW of demand they curtail during summer 4pm-8pm events when LES calls for it. Modest revenue, requires SCADA capability. V4 territory.
The whole math model in one diagram.
Memorize this and you can rebuild the whole model on a napkin.
Monthly bill
monthly_bill = customer_charge
+ billing_demand ร ($25.17/kW)
+ monthly_kwh ร $0.0245/kWh (winter)
+ excess_kvar ร $2.60
Billing demand (the ratchet)
billing_demand = max(
this_month_peak_kw,
0.65 ร max(last_year's_summer_peaks),
minimum_floor
)
Excess kVAR (the PF penalty)
excess_kvar = max(0,
peak_kw ร tan(arccos(power_factor))
โ peak_kw ร 0.39523
)
Annual savings (the hero number)
annual_savings = ฮฃ (current_state_monthly_bill โ tier3_monthly_bill)
across 12 months
Self-check โ can you answer these from memory?
If your peak last July was 1,200 kW, what's the minimum demand you'll be billed for next February?
Show answer
780 kW (= 65% ร 1,200). Even if you only USE 400 kW in February, you pay for 780 kW. That's the ratchet. Number to memorize: 65%.
If our energy is essentially flat at retail, why doesn't the scheduler save money by chasing cheap hours?
Show answer
Because the cheap-hour savings are tiny โ about $0.001/kWh between summer and winter. For our scale, that's ~$1,800/year total. The scheduler's real job is preventing demand peaks, not finding cheap hours. The savings live in the demand charge, not the energy charge.
Someone asks, "is this AI?" What do you say?
Show answer
"AI helped build the analysis fast and synthesize the research. The savings number itself is arithmetic on the verified LES tariff โ no machine learning required. There's a price-forecasting ML model parked for a future demand-response use case, but that's separate." Don't overclaim AI on the savings math.
Why does upgrading an SCR rectifier to IGBT save money in three different ways?
Show answer
(1) Higher efficiency means less AC draw from the grid โ energy savings. (2) Less AC draw means lower kW peak โ demand savings. (3) Higher power factor means no PF penalty โ kVAR savings. Three savings layers from one piece of equipment. For a 1,000 kW DC line: ~$93k/year combined, 6.4-month payback.
If the model says $325k but a specific plant doesn't match every assumption, how do you frame the number?
Show answer
"The math is verified against the LES tariff. The plant inputs are generic. To go from generic to plant-specific, you need 12 months of utility interval data, a rectifier inventory walk, and a quick call with LES to confirm the rate class. Worst case it lands around $150k; best case higher. Either way, real money." Always give a range, never a single point estimate when stakes are real.
What's the cheapest, fastest single project that pays for itself?
Show answer
A 700 kVAR detuned automatic capacitor bank. ~$18,000 capex. Kills the power factor penalty. ~$18,000/year in penalties avoided. 12-month payback, zero process changes, zero operator retraining. The "easy yes" project to start with.
Glossary โ every term you'll see
- kW (kilowatt)
- Unit of power. How fast electricity is being used right now. Like miles-per-hour for cars.
- kWh (kilowatt-hour)
- Unit of energy. Total amount used over time. Like total miles driven.
- kVAR (kilovolt-amperes reactive)
- The "sideways push" โ reactive power that doesn't do useful work but still flows on the grid.
- kVA (kilovolt-amperes)
- Total apparent power. Hypotenuse of the triangle of real power (kW) + reactive power (kVAR).
- Power Factor (PF)
- kW รท kVA. A number from 0 to 1. Higher = more efficient. 0.93 is LES's threshold.
- Demand charge
- $ per kW of the highest 30-minute peak in the month. The big chunk of your bill.
- Energy charge
- $ per kWh you actually use. Flat-ish, doesn't vary much by time of day.
- Summer ratchet
- The clause that locks your minimum billing demand at 65% of the previous summer's worst peak, for the next 11 months.
- LLP
- Large Light & Power โ LES's industrial rate class (Rates 15, 16, 39). Where most mid-size finishers fall.
- GSD
- General Service-Demand โ smaller commercial rate class (Rates 11, 12). Below 400 kW peak.
- SCR
- Silicon-Controlled Rectifier โ older thyristor-based AC-to-DC converter. 80% efficient. Bad PF.
- IGBT
- Insulated Gate Bipolar Transistor โ modern switch-mode AC-to-DC converter. 92% efficient. Good PF.
- SEP
- LES Sustainable Energy Program โ annual grant pool ($2.2M for 2026) that offsets capex for EMS, VFDs, efficient compressors.
- EMS
- Energy Management System โ software that monitors and controls major plant loads. Eligible for SEP grant.
- VFD
- Variable Frequency Drive โ motor controller that varies speed to match load. Saves energy + qualifies for $50/HP grant.
- SPP
- Southwest Power Pool โ the wholesale electricity market that LES buys from. Retail customers don't transact here directly.
- LMP
- Locational Marginal Price โ the wholesale price of electricity at a specific grid location. What SPP customers pay. Not what we pay.
- Ratchet floor
- The minimum demand you'll be billed for in any month, set by 65% ร previous summer's peak.
- Current state
- The model's baseline โ what the plant looks like if nobody does anything different. Compared against each "tier" of savings.
- Tier 1 / 2 / 3
- Three stacked layers of improvements. Tier 1 = software only. Tier 2 = + aux load shifting. Tier 3 = + hardware retrofits.
- LLP-TOUD
- LES rate variant with separate on-peak and off-peak demand charges (Rates 30/31/32). Useful for plants that can genuinely shift load away from the 12pm-9pm summer on-peak window.
- LLP-OPD
- LES rate variant that penalizes excess off-peak demand (Rates 27/28/29). Designed for night-heavy operations. Rarely useful for day-shift plating shops.
- Curtailment Rider
- Optional LES program that pays $2.40/kW for demand curtailed during summer 4-8pm events when LES calls them. Requires SCADA capability. V4 territory.