Solar parks and batteries: dispatch ignores ageing and prices

Large storage parks drift apart in temperature and charge across many strings

What happens. A storage park is built from containers with many battery strings. The strings do not age evenly: some are warmer, some are fuller, some weaker. As the park grows, the differences grow, and the operator sees only the total and a few alarms.

What it costs. Capacity that is lost because the weakest or hottest string sets the limit, and cooling that works against the spread instead of removing it.

Why the usual fix fails. Treating the park as one big battery hides the spread. Balancing by a fixed rule in the battery management does not know the temperature of the neighbour string. More sensors give more numbers and no decision.

How a fixed-price remote diagnosis checks it. From string data (voltage, current, temperature, cooling state) I estimate the state of each string and check how much capacity the spread costs. The fixed-price remote diagnosis answers whether computed string states and a dispatch that considers them can give back usable capacity.

Background: Computing unmeasured states for control

Battery dispatch ignores the ageing it causes

What happens. The storage is dispatched to the price: charge when cheap, discharge when dear. How much each cycle costs in lifetime does not enter the decision. A deep cycle at high power and a gentle one earn nearly the same on the day and age the cells very differently.

What it costs. A battery that loses capacity faster than planned, a warranty that is used up early and revenue that falls in the later years.

Why the usual fix fails. A fixed limit on cycles per day is blunt and wastes good price spreads. Treating wear as a cost per kilowatt hour without a model is a guess. Waiting for the manufacturer’s warranty to decide hands the control to a contract.

How a fixed-price remote diagnosis checks it. From the operating history, price data and the specification of the cells I check how the ageing depends on the way of operation and what a dispatch that prices the ageing would do differently. The fixed-price remote diagnosis answers whether ageing-aware dispatch is worth the effort for your storage.

Background: Economic predictive control

Solar plant, storage and loads share one grid connection limit

What happens. A site has solar, a battery and loads behind one grid connection. When the sun is strong and the load is low, the feed-in hits the limit, and the plant is throttled. Each device follows its own rule, and nobody decides jointly who gets the connection when.

What it costs. Curtailed solar energy that was free, a battery that is full when it would be needed and connection capacity that stands unused at other hours.

Why the usual fix fails. Throttling the solar plant at the limit is simple and wastes energy. A battery charged by a fixed rule is full at noon or empty at the wrong hour. A bigger connection takes years and costs money.

How a fixed-price remote diagnosis checks it. From the profiles of generation, load and connection I check how much a joint schedule of battery and flexible loads can reduce curtailment under the limit. The fixed-price remote diagnosis answers what such an optimisation brings for your site and what forecasts it needs.

Background: Economic predictive control

Solar forecast errors turn into imbalance cost

What happens. The operator sells the forecast output. The sun does something else: a cloud front is later, haze is stronger. The difference between promise and delivery has to be bought or sold at the imbalance price, which is poor at exactly the bad hours.

What it costs. Imbalance costs that fluctuate from day to day and a forecast service that nobody can compare with the plant’s own data.

Why the usual fix fails. Buying a better forecast service helps on average and not at the park. A battery as a buffer is expensive for forecast errors alone. Selling less than the forecast shifts the loss into lost sales.

How a fixed-price remote diagnosis checks it. From plant data and past forecasts I check how large the error is at your park, in which weather it arises and whether a short-term correction from your own measurements improves it. The fixed-price remote diagnosis answers how much imbalance cost a better short-term forecast can save.

Solar, wind and storage in one park are operated separately

What happens. In a hybrid park, each technology has its own operator view and its own rule. The battery is charged when the price is low, without asking whether the wind will produce the same hour. The solar plant is throttled while the battery is empty.

What it costs. Lost value, because the three do not complement each other, and grid connection capacity that is used badly.

Why the usual fix fails. Optimising each part is easy and leaves the interplay open. A joint rule such as “always charge from the surplus” is not good in every season. A single forecast for the whole park needs data from all three.

How a fixed-price remote diagnosis checks it. From generation data, prices and the grid limit I check how much a joint dispatch gains against separate operation. The fixed-price remote diagnosis answers whether a common plan for solar, wind and storage can run in your operation and what it needs from the existing control.

Background: Economic predictive control

Underperforming strings and inverter faults go unnoticed

What happens. A string that produces less than its neighbours, an inverter that derates in the heat or a fuse that has tripped: all of that shows as a slightly lower total, which the weather also explains. The operator looks at the portal and sees an average day.

What it costs. Lost yield for weeks until someone finds it, and repair visits that are planned late and combined badly.

Why the usual fix fails. Alarm limits of the inverter only report outright failures. A visual walk through the park takes days and happens twice a year. Comparing the output to a simple reference value flags too many good days.

How a fixed-price remote diagnosis checks it. From the string and inverter data and the weather data I check how well a model of the expected output separates a real fault from weather. The fixed-price remote diagnosis answers whether fault detection from existing data finds underperformance early at your park.

Panels are cleaned by calendar, not by soiling

What happens. Dust, pollen and bird droppings reduce the yield, and the soiling depends on the season, the site and the rain. The cleaning is booked at fixed times. Sometimes the panels are clean already, and sometimes they have lost output for weeks.

What it costs. Cleaning cost without a gain, or yield that is lost between two dates. Neither is visible, because nobody measures the soiling directly.

Why the usual fix fails. More frequent cleaning costs more and is not needed everywhere. A reference cell measures one place. Waiting for the rain is a plan until it fails to rain.

How a fixed-price remote diagnosis checks it. From yield data, weather and rain events I estimate how the soiling builds up and when cleaning pays for itself. The fixed-price remote diagnosis answers whether a computed soiling level can steer the cleaning dates at your park.

Several markets at once: day-ahead, intraday and reserve

What happens. A storage or a flexible plant can sell on the day-ahead market, the intraday market and for reserve. The bids are due at different times, with different information, and what is sold in one market is missing in another. In practice, one market is chosen by habit.

What it costs. Lost revenue because the plant is in the wrong market at the hour, and a trading desk that works by gut feeling and spreadsheet.

Why the usual fix fails. Following the highest price of the day looks right and overlooks the limits of capacity and ageing. A fixed split between markets is easy and does not adapt when the prices move. A forecast alone does not decide the bid.

How a fixed-price remote diagnosis checks it. From the history of prices, the technical limits of the plant and your bidding rules I check how much a joint decision across the markets could gain and how robust it is against forecast errors. The fixed-price remote diagnosis answers whether a multi-market optimiser is worth building.

Background: Economic predictive control

Loads, solar and storage on an industrial site are not coordinated

What happens. A factory has production loads, a solar plant and a battery, each with its own control. The battery charges when the sun shines, even if a machine will start in five minutes and take the power from the grid anyway. Nobody plans the use of these three together.

What it costs. A power bill that is higher than needed, solar energy that is fed in at a low price and bought back at a high price later, and a battery that does not deliver what the investment plan promised.

Why the usual fix fails. Rules of thumb in the battery control follow the last hour, not the plan of the shift. An energy management system that only shows the data does not decide. A big battery covers the lack of coordination with money.

How a fixed-price remote diagnosis checks it. From load profiles, the production plan, solar data and tariffs I check how much a coordinated plan of the three gains against today’s operation. The fixed-price remote diagnosis answers whether a site optimisation is worth building and what data it needs from production.

Background: Economic predictive control

Power peaks still hit despite the storage

What happens. A battery is installed to cut the peak of the grid power, because the grid charge depends on the highest quarter-hour of the month. Still, a peak gets through. The battery was empty or its limit was too low when a few machines started together.

What it costs. Peak charges that run for the whole month or the year because of a single quarter-hour, and a storage that does not pay back as expected.

Why the usual fix fails. A fixed threshold for the battery is a trade-off between empty and over-cautious. Following the present load alone reacts too late. A bigger battery buys safety with money.

How a fixed-price remote diagnosis checks it. From load profiles, the shift plan and the battery data I check which peaks were avoidable and how a prediction of the load could set the threshold. The fixed-price remote diagnosis answers how much a predictive peak control brings against your present setting.

Background: Economic predictive control

Panels and crops compete in agri-solar plants

What happens. Agri-solar plants put panels over fields. The panels shade the crop, and the crop needs light at certain times of the season. The panel angle that gives the most power is not the angle that gives the best harvest. Today the tracking follows only the sun.

What it costs. A harvest that is lower than hoped, or solar yield given away without need, and an uncertain return on both sides.

Why the usual fix fails. Fixed panel positions are a compromise for the whole year. Tracking for power only treats the crop as a side issue. Asking the farmer to decide by hand per day is not practical.

How a fixed-price remote diagnosis checks it. From weather, solar data and crop growth information I check how a tracking plan could balance the two targets over the season. The fixed-price remote diagnosis answers whether tracking that is optimised for both is worthwhile for your crops and your plant.

Trackers follow the sun angle only, and wind loads are ignored

What happens. A tracker turns the panels to the sun by the calendar and the clock. On a cloudy day with diffuse light, the best angle is a different one, and in strong wind the panels have to go to a safe position. The wind position is chosen by a threshold on the wind speed.

What it costs. Yield lost on cloudy days, hours of stow position when the wind is only gusty and the mechanical wear of frequent movements.

Why the usual fix fails. A better clock does not know the weather. A lower wind threshold protects the plant and costs yield. A sensor on each tracker is expensive and still leaves the question of the best angle open.

How a fixed-price remote diagnosis checks it. From tracker, weather and yield data I check how much a model of the light and the wind loads could add to the yield and how the stow decisions could be based on forecasts. The fixed-price remote diagnosis answers whether model-based tracking pays at your plant.

Monitoring portals have gaps and allow no analysis

What happens. Inverters, meters and weather stations each report to their own portal. The data have gaps, different time steps and different names. A question such as “why was Tuesday weak?” needs hours of work with exports, and often the answer is not found.

What it costs. Faults that stay hidden, performance reports built by hand each month and engineers who spend their time on data instead of the plant.

Why the usual fix fails. A new portal adds one more place to look. Spreadsheets break at the next change of format. A data platform without a first useful analysis stays empty.

How a fixed-price remote diagnosis checks it. From the exports of your portals I check the gaps and consistency, and build a data layer that makes the first analyses possible: expected against actual output, fault search. The fixed-price remote diagnosis answers what is usable in your data today and what is missing before a model can use it.

Signs of thermal runaway in a battery are found too late

What happens. A battery fire starts with small signs: a cell that is warmer than its neighbours, a voltage that falls a little faster, a gas sensor that reacts. The monitoring has fixed alarm limits, which are crossed when the process has started. Operators are afraid of both missing a fire and false alarms that shut the plant.

What it costs. A safety risk for people and the site, insurers who ask for more and downtime after every alarm.

Why the usual fix fails. Tighter fixed limits give more false alarms. Additional sensors give more data and no earlier decision. A fire protection system reacts after the event, not before.

How a fixed-price remote diagnosis checks it. From temperature, voltage and gas data and recorded events I check which patterns come before an alarm and how a model of the heat in the cells could report a deviation earlier. The fixed-price remote diagnosis answers whether early detection by model can be added to your monitoring and what proof of reliability it would need.

Background: Computing unmeasured states for control

Contracts require grid-supportive operation of the storage

What happens. The grid operator or a contract partner requires that the storage stays within given limits: a power that may be fed in, ramp rates, a window for the state of charge. The trading side wants to use the battery as hard as the prices allow. The two requirements collide in the dispatch.

What it costs. Penalties or loss of the contract if a limit is violated, and lost revenue if the operator stays far from the limit.

Why the usual fix fails. A conservative margin on all limits is safe and loses revenue. A rule check after the plan is made throws away plans that were not feasible. Two teams who each have one half of the task do not find the compromise.

How a fixed-price remote diagnosis checks it. From the contract terms, the plant data and the trading rules I check how the limits can be written into the plan as constraints and how much revenue remains. The fixed-price remote diagnosis answers whether an optimisation that respects the contract by construction is realistic for your storage.

Background: Economic predictive control

Island grids with diesel, solar and storage run on fuel by rule

What happens. A small grid without a connection to the main grid is supplied by diesel generators, solar and a battery. The diesel runs because the rule says “start when the battery falls below a level”, often at a poor load point, and for a short time only. Solar is throttled when the battery is full.

What it costs. Diesel consumption higher than necessary, many starts that wear the engines and a fuel logistic that is expensive in remote places.

Why the usual fix fails. A second generator as reserve covers the uncertainty with fuel. A larger battery costs capital. Rules by the operator’s experience are hard to hand over to the next shift.

How a fixed-price remote diagnosis checks it. From load, solar and generator data I check how much a predictive dispatch, which looks ahead at load and sun, saves in starts and fuel. The fixed-price remote diagnosis answers whether such a dispatch is feasible on the existing control of the island grid.

Background: Economic predictive control

Forecasts must sit inside the optimiser, not beside it

What happens. Companies that sell flexibility from storage and production need forecasts of price, load and generation, and an optimiser that decides. In many set-ups the forecast comes from one team and the optimiser from another. The optimiser trusts the forecast as if it were certain, and the plan breaks at the first deviation.

What it costs. Plans that must be corrected during the day, flexibility offers that are not kept, and unused potential.

Why the usual fix fails. A better forecast reduces the problem but never removes it. Rules to correct the plan at every deviation produce a nervous plant. A separate model of the forecast error does not enter the decision.

How a fixed-price remote diagnosis checks it. From the forecasts, the plans and the actual values of a few weeks I check how large the errors are and how the optimiser can take their size into account. The fixed-price remote diagnosis answers whether optimisation under uncertainty gives a more robust plan for your flexibility.

Background: Economic predictive control

Control and plant management for large solar and storage plants is hard to join

What happens. An integrator builds large solar and storage plants from parts of many suppliers. Each part has its own control and its own interface. The plant controller that has to follow the grid operator’s requirement must bring them together, and the control engineers for this are few.

What it costs. Commissioning that drags on, a plant that does not meet the grid requirements at first test and costs for contractual penalties.

Why the usual fix fails. Copying the last project’s architecture brings its quirks. A supplier’s central controller locks the integrator to one maker. Hiring takes long and the project will not wait.

How a fixed-price remote diagnosis checks it. From the single-line diagram, the interface lists and the grid requirements I check how the plant control could be structured and which loops decide whether the requirement is met. The fixed-price remote diagnosis answers what a model-based control concept with supervision and documentation would involve.

Flexible grid connections cap the site at times the site cannot plan around

What happens. Grid operators begin to offer flexible connection agreements: the site gets its connection sooner, and in return the operator may limit the power at times, with notice. New heat pumps, electric boilers, charging parks, batteries and data centres are the typical users. The site has to keep its whole operation below a limit that changes.

What it costs. Without a controller, the site either stays far below the limit and loses output, or crosses it and pays or is cut off. Production plans become uncertain.

Why the usual fix fails. Dimensioning the site for the worst case wastes the connection. A manual reaction to the notice of the operator comes too late. A simple load shedding switches off the most expensive process first, not the one that can wait.

How a fixed-price remote diagnosis checks it. From the load profiles, the process constraints and the expected limit profile I check which loads can wait, for how long, and what a site controller that keeps the sum under the limit would have to do. The fixed-price remote diagnosis answers how much of the planned output can be kept under a time-varying limit.

Background: Economic predictive control

Midday solar earns nothing unless it is used on site

What happens. For new solar plants the feed-in payment stops in hours with negative prices, and a cap on the feed-in applies until a smart meter is installed. The farm or the company with solar on the roof sees that the midday output, which was a gain, now earns nothing or costs.

What it costs. Solar energy that is thrown away or sold at a loss, and loads such as cooling, charging and pumps that run at the evening price.

Why the usual fix fails. Switching off the plant in negative hours saves the penalty and gives up the energy. Moving loads by hand depends on a person who remembers. A battery is expensive when the loads could do the job.

How a fixed-price remote diagnosis checks it. From load profiles, solar data and price data I check which flexible loads could move to the sun and how much of the own generation could be used on site. The fixed-price remote diagnosis answers what a schedule of flexible loads and storage against your own solar yield would bring.

Background: Economic predictive control