District heating: supply temperature by curve, sources by rule

The supply temperature follows the outdoor curve, mostly too hot

What happens. The supply temperature of the network is set from the outdoor temperature by a fixed curve. To be safe against complaints at the end of the network, the curve is placed high and rarely revisited.

What it costs. Heat losses in the pipes, higher pump power and generation at a temperature the customers did not need. Heat pumps and condensing plants lose efficiency at high supply temperatures.

Why the usual fix fails. Lowering the curve by hand works until a cold morning brings complaints from the far end, and then the curve goes back up. The curve knows the outdoor temperature but not the wind, the sun on the buildings or the load tomorrow.

How a fixed-price remote diagnosis checks it. From supply and return temperatures, flows, heat meter data and weather I check how much margin the curve carries. The fixed-price remote diagnosis tells you whether a predictive supply temperature based on a load forecast and a network model could lower the average temperature, and what measurements it needs.

Background: Economic predictive control

The return temperature is too high

What happens. The water comes back from the customers warmer than the network was designed for. Some substations pass on hot water that they did not cool down.

What it costs. Heat pumps and combined heat and power plants lose efficiency, the network carries more water for the same heat, and pumps work harder.

Why the usual fix fails. General appeals to customers and fixed return-temperature clauses in contracts rarely change much. Finding the substations that cause it means visiting many of them.

How a fixed-price remote diagnosis checks it. From meter data (flow, supply and return temperatures per customer) I look for the customers and the hours that contribute most. The fixed-price remote diagnosis tells you whether a customer-side diagnosis can single out the worst substations and what it would take to fix them.

There are few measuring points in the network

What happens. Temperatures and pressures are measured at the plant, at a few key points and at the customer meters. In between, the state of the pipes and the flows is not known.

What it costs. Bottlenecks and low-pressure spots are discovered through customer complaints. Planning for new customers rests on estimates.

Why the usual fix fails. Adding measuring points is costly in an old network, and the data from the points that exist is not combined into one picture.

How a fixed-price remote diagnosis checks it. From the meter and measurement data you have, with the pipe plan, I check how well the state of the network can be computed. The fixed-price remote diagnosis tells you whether a state estimate is possible with your measurements, and where an extra sensor would give the most.

Background: Computing unmeasured states for control

The differential pressure is kept high

What happens. The network pumps hold a fixed differential pressure, set high enough that even the worst customer is supplied on the coldest day.

What it costs. Pump power, all year round, for a pressure that is mostly more than needed.

Why the usual fix fails. Lowering it by hand risks the far customer. Without knowing which customer is the critical one at a given hour, nobody dares to change the setpoint.

How a fixed-price remote diagnosis checks it. From pump, pressure and flow data and the differential pressures at the critical points I check how often the setpoint is above need. The fixed-price remote diagnosis tells you whether a pressure that follows the critical point could cut pump power without risk.

Background: Economic predictive control

The storage tank is run by rule of thumb

What happens. The heat storage is charged and discharged by simple rules, for example at fixed times or at fixed levels. It is not known whether the choice is the best one for the day ahead.

What it costs. Value is lost: heat is produced at expensive times and the storage is full when cheap or free heat is available.

Why the usual fix fails. A bigger tank adds investment without fixing the rule. Operators adjust the rule by experience, which works on average but not on unusual days.

How a fixed-price remote diagnosis checks it. From tank temperatures, levels, production and demand data, plus price signals, I check how much better a planned use of the storage would have been. The fixed-price remote diagnosis tells you whether a storage dispatch looking ahead is worth the effort for your system.

Background: Economic predictive control

Waste heat, combined heat and power, boilers and heat pumps are not coordinated

What happens. Several heat sources feed one network: industrial waste heat, a combined heat and power plant, boilers and heat pumps. Each is started by its own rule or by the operator on duty.

What it costs. Higher heat cost than necessary, because the cheapest mix is not the one that runs. Electricity prices, fuel and carbon prices all change the best mix during the day.

Why the usual fix fails. A merit-order list on paper is out of date at the next price change. Handling it manually works for one operator who knows the system and not for the next shift.

How a fixed-price remote diagnosis checks it. From the production, price and demand data of a year I compare what ran with what an optimised plan would have chosen. The fixed-price remote diagnosis tells you whether producer dispatch is worth building, and what forecast it would need.

Background: Economic predictive control

New low-temperature networks behave differently

What happens. A new network is designed for low temperatures, with new kinds of customer stations and heat sources. The way supply and demand interact is different from the old high-temperature networks.

What it costs. Comfort complaints and an operation that stays cautious for years, because no one has experience with the new behaviour.

Why the usual fix fails. Operating rules carried over from old networks fit badly. Learning by trial costs customers’ comfort.

How a fixed-price remote diagnosis checks it. From design data, early operating data and the plan of the network I check how its dynamics differ. The fixed-price remote diagnosis tells you whether a network model and a control adapted to it should be set up from the start.

Background: Predictive control of process plants

Faulty substations raise return temperatures

What happens. A substation with a stuck valve, a fouled heat exchanger or a wrong setting passes on hot water. Nobody notices, because the customer is warm and does not complain.

What it costs. Lower efficiency of the whole network, caused by a handful of substations among many.

Why the usual fix fails. Inspecting every station on a cycle is slow. A fixed limit on the return temperature of each customer flags many in good order and misses those in the middle.

How a fixed-price remote diagnosis checks it. From the hourly or daily meter data of each substation I look for patterns that point to faults. The fixed-price remote diagnosis tells you whether fault detection from meter data you already collect can pick out the stations worth a visit.

The heat demand forecast is made by hand

What happens. The operator estimates tomorrow’s demand from the weather forecast and experience. It works on typical days and poorly on odd ones, such as holidays, sudden cold or a change in customer behaviour.

What it costs. Wrong scheduling: too much produced and wasted, or too little and an expensive boiler started at short notice.

Why the usual fix fails. A forecast by the weather service gives the temperature, but not the behaviour of your customers. Spreadsheets with fixed factors age as soon as the customer mix changes.

How a fixed-price remote diagnosis checks it. From a few years of demand, weather and calendar data I check how well a short-term forecast can be built and how it compares with the current estimate. The fixed-price remote diagnosis tells you whether a forecast would improve scheduling, and what data to keep.

Water travel times in the network are ignored

What happens. Heat put in at the plant arrives at the customer after hours, in a large network even after much longer. Control and planning treat the network as if it responded at once.

What it costs. Under-supply in the cold hours, when a change at the plant arrives too late, and over-supply when it arrives after the need has passed.

Why the usual fix fails. Faster feedback does not help when the water cannot travel faster. A correction for the delay by rule of thumb is wrong when the flow changes.

How a fixed-price remote diagnosis checks it. From temperature and flow measurements along the network I check how long the water takes along its paths and how that changes with load. The fixed-price remote diagnosis tells you whether a model with transport delays would improve control and planning.

Background: Computing unmeasured states for control

Industrial waste heat varies

What happens. A factory could feed heat into the network, but its heat comes in when the production runs, not when the network needs it. The amount and temperature change by the shift.

What it costs. Heat that is available but unused, and a supply the network operator cannot count on.

Why the usual fix fails. A fixed contract amount ignores the swings. A large storage tank to absorb them is expensive.

How a fixed-price remote diagnosis checks it. From the factory’s heat profile and the network’s demand profile I check how well they overlap and how much storage or control would bridge the gap. The fixed-price remote diagnosis tells you whether an integration control can raise the share of waste heat that is used.

Background: Economic predictive control

Solar thermal fields with storage overheat or lose heat

What happens. A field of solar collectors feeds a storage tank and the network. The sun does not follow demand: the field delivers most when the demand is least.

What it costs. Overheating and venting of the collectors on sunny days, losses from the storage in poor weather and a field that does not reach its yield.

Why the usual fix fails. Rules based on the tank temperature alone react when the problem is already there. Bigger storage helps but costs.

How a fixed-price remote diagnosis checks it. From collector, tank and network data and the weather forecast, I check how often the plant overheated or ran at a poor point. The fixed-price remote diagnosis tells you whether predictive operation of field and storage can use the sun better.

Background: Economic predictive control

Electric boilers are timed badly

What happens. Electric boilers turn power into heat. They are started when heat is needed or when a rule says the power is cheap, not on a plan that follows the price.

What it costs. Power cost that is higher than needed, and opportunities missed when power is very cheap or oversupplied.

Why the usual fix fails. Fixed times of day and price thresholds do not fit a market where the price curve changes every day.

How a fixed-price remote diagnosis checks it. From the price history, boiler logs and heat demand I compare what ran with a price-led plan. The fixed-price remote diagnosis tells you whether price-led dispatch of electric boilers, within the limits of the network and storage, is worth building.

Background: Economic predictive control

Water losses go unnoticed

What happens. A leak in the pipes loses hot water, which is replaced by treated make-up water. Small leaks may run for months, as the make-up quantity is read rarely.

What it costs. Treated water, the heat that goes with it and damage to the ground and the pipe. A small leak can grow into a pipe failure with a large repair.

Why the usual fix fails. Reading the make-up meter by hand once a month finds only the large ones. Walking the line with detectors is slow.

How a fixed-price remote diagnosis checks it. From the make-up water, pressure and flow data of the network I check whether the normal losses can be told apart from leaks. The fixed-price remote diagnosis tells you whether leak detection from the data you already have is realistic.

Background: Computing unmeasured states for control

Deciding which streets to connect carries investment risk

What happens. The operator must decide where to extend the network. The choice rests on connection requests, estimates of demand and rough costs.

What it costs. Investment that does not pay back if customers do not connect or if the heat source cannot cover the added demand.

Why the usual fix fails. A spreadsheet with averages ignores how the extension changes the flows, pressures and temperatures in the existing network.

How a fixed-price remote diagnosis checks it. From the network plan, meter data and the candidate streets I check which extensions the network can carry. The fixed-price remote diagnosis tells you whether a network simulation of the options is possible with your data and what it would show.

Municipal heat plans need scenarios

What happens. Municipalities must draw up a heat plan, and the utility is asked to describe how the network could develop: which sources, which temperatures, which areas.

What it costs. Planning effort for a question with many variants, and results that depend on assumptions that are hard to check.

Why the usual fix fails. Scenarios put together by hand in a spreadsheet are few, and the link between source, network and demand is simplified.

How a fixed-price remote diagnosis checks it. From the data on demand, sources and network that you can share, I check how a simulation and optimisation of scenarios could be set up. The fixed-price remote diagnosis tells you what a scenario tool would need and how much work it is to build.

Background: Optimal paths, schedules and setpoint plans

District cooling networks see load peaks

What happens. A district cooling network supplies chilled water to buildings. On hot afternoons the load peaks, and all chillers run together.

What it costs. Chiller power at the least efficient time and the high cost of capacity that is needed only a few hours a year.

Why the usual fix fails. More chillers cover the peak but cost investment. Storage shifts the load but needs a rule on when to charge it.

How a fixed-price remote diagnosis checks it. From load, chiller and storage data and the weather forecast I check how large the peaks are and how often they could have been shifted. The fixed-price remote diagnosis tells you whether predictive operation of the cooling network can flatten them.

Background: Economic predictive control

Network control by fixed curves

What happens. The control of the network consists of fixed curves and setpoints, implemented years ago in the control system. Nobody changes them because nobody remembers why they were chosen.

What it costs. Losses and pump power that the original design accepted, but which are avoidable with today’s knowledge.

Why the usual fix fails. A replacement of the control system alone brings the same curves in new hardware. A rewrite without a concept repeats the problem.

How a fixed-price remote diagnosis checks it. From the control documentation, logs and plans I judge what the control does today and what a new concept would have to change. The fixed-price remote diagnosis gives you a basis to decide between keeping, tuning and redesigning.

Background: Predictive control of process plants

Smart meter data is collected but not used

What happens. The meters of the customers send readings at short intervals, and the data is stored for billing. Few people look at it for operating the network.

What it costs. Missed insights: hidden faults, load patterns, customers that behave unusually, and information the network control could use.

Why the usual fix fails. The billing system is not designed for analysis, and a data warehouse project takes long without a clear question.

How a fixed-price remote diagnosis checks it. From a sample of meter data and your system description I check what questions the data can answer and what needs to be prepared. The fixed-price remote diagnosis tells you which analyses are worth setting up first and what data layer they need.

There are few network engineers

What happens. Operating and improving a heat network takes someone who knows the hydraulics, the control and the data. Many utilities have one such person, or none.

What it costs. Improvements wait. Known problems stay on the list for years because no one has the time.

Why the usual fix fails. Hiring is slow, as the profile is rare. External studies are one-off and the knowledge leaves with the report.

How a fixed-price remote diagnosis checks it. From a short description of the network and exports of your data I find the three or four points with the most to gain. The fixed-price remote diagnosis gives you a ranked list of what to do, so a small team can use its time where it counts.

Network models are not calibrated on live data

What happens. A model of the network built from first principles needs to match the real network. The parameters, such as pipe roughness, heat losses and the real behaviour of the substations, are estimated once and drift.

What it costs. Planning of production and storage rests on a model that is trusted less over time, and the model is rebuilt for each question.

Why the usual fix fails. A model tuned by hand against a few measurements fits those days and not the rest. A purely statistical model has no physical meaning for the cases it has not seen.

How a fixed-price remote diagnosis checks it. From the network plan and a few months of live data I check how well a physical model can be calibrated and kept current. The fixed-price remote diagnosis tells you whether a calibrated network model could support production and storage planning, and what it takes to maintain.

Background: Predictive control of process plants

Lower supply temperatures for heat pumps risk under-supply at the far end

What happens. To suit heat pumps, networks lower their supply temperature, and the connection rules change for new connections. But many substations are not known well, and some need the high temperature.

What it costs. A risk of under-supply at the end of the network and complaints, and the temptation to keep the safe, high temperature.

Why the usual fix fails. A fixed step down of the curve is a gamble: either it works or the far customers freeze. Waiting for complaints is the slow way to find the weak points.

How a fixed-price remote diagnosis checks it. From return temperatures, meter data and the network plan I check which parts of the network have little margin. The fixed-price remote diagnosis tells you whether the supply temperature can be controlled from the computed state at the network end, with return temperatures as a computed measurement.

Background: Computing unmeasured states for control

Large heat pumps are new plants with few operating data and permit limits

What happens. Large heat pumps take heat from rivers, wastewater or industrial waste heat. They are among the first plants of their size, and the source temperature swings with the season and the weather.

What it costs. Part-load operation, unclear interplay with combined heat and power plants and storage, and permit limits such as how much a river may be cooled.

Why the usual fix fails. The supplier’s operating instructions describe the standard point. Without operating experience the first years are run cautiously, at a cost.

How a fixed-price remote diagnosis checks it. From the heat pump design data, source temperatures and the surrounding plant I check how dispatch could respect the permit limits and the other sources. The fixed-price remote diagnosis tells you whether an operating strategy for heat pump, storage and combined heat and power can be worked out before the data exists.

Background: Economic predictive control

Funded transformation plans now have to be built and operated

What happens. Many heat network operators have had plans funded for their transformation. Now the plans must be built, with new sources and lower temperatures, and operated.

What it costs. A plan on paper does not say how the network behaves in operation. Surprises in operation cost the savings the plan counted on.

Why the usual fix fails. The consultant’s plan stops at the investment decision. Operation is learned on the running network.

How a fixed-price remote diagnosis checks it. From the plan, the network data and the planned sources I check how the funded network would be run. The fixed-price remote diagnosis tells you whether a network model and operating optimisation should be set up with the build, and what they need.

Background: Predictive control of process plants