GB2467159A

GB2467159A - Delayed start grid-responsive load

Info

Publication number: GB2467159A
Application number: GB0901212A
Authority: GB
Inventors: Andrew Nicholas Dames
Original assignee: Individual
Current assignee: Individual
Priority date: 2009-01-26
Filing date: 2009-01-26
Publication date: 2010-07-28
Also published as: GB0901212D0

Abstract

A responsive load controller provides a variable automatic delay to the start of a process (such as charging the battery of an electric vehicle, running a dishwasher, an electric kiln...) dependant on locally monitored grid health (eg by tracking line frequency deviation). The response function may be dependent on a real time clock, or may be remotely altered. The time delay depends upon a frequency threshold vs time delay profile. Once the process has started, the load controller makes no attempt to affect the process irrespective of subsequent grid state. This permits a wide range of processes to be simply configured as responsive loads, without the complexity of adapting the process to consume a variable amount power or have variable timing of intermittent power usage as per prior art methods. This is especially significant in processes (eg washing, battery charging) that require a defined sequence and timing for optimum results.

Description

Prior art covers line frequency driven load shedding, (lots of references, see above) timing adjust on eg fridges, water heaters...(RLtec, J Short IEEE). Other prior art covers the use of cold load pickup delay(ref above), but this is generally short term and specifically designed to minimize startup current on reconnection.

Some processes, for example the charging of batteries, are both energy intensive and complex, requiring carefully controlled sequence of power variations to complete whilst maintaining optimum battery lifetime. Consequently line frequency responsive load processes that disconnect at times of grid stress are undesirable, and approaches that attempt to moderate the power consumption by building grid response into the device are complex, and likely to entail some degradation in performance.

Other examples of energy intensive domestic appliance loads that would react badly to grid response shutdown are washing machines and dishwashers. Other items such as tumble driers can be adapted to have their heater turned on and off according to grid stress, but would not respond well to being fully turned off (ref PNNL 17079).

Our innovation is to provide automated responsive load by delaying the startup of an appliance, using an algorithm that minimizes likely startup delay, whilst giving the maximum grid response, in an amount tailored to line frequency that is of most benefit to the grid management body. As the process does not seek to alter or interrupt the behavior of the system or appliance once started, issues due to unwanted consequences of interruption do not exist, and the grid response can be straightforwardly applied to many loads previously deemed unsuited to dynamic demand management.

Note that this is different from providing cold start (or black start) assistance, which is about leaving a time delay before starting loads after loss of grid power. Eg sec 1.4.3 in ref PNNL 17079 Embodiment The basic control outline for a delayed start control module is shown in Fig 1. As can be seen, there is no modification required to the load, which can keep all of its pre existing control and timing structures if present.

The response function would typically start at +0.5Hz, and decline to +0Hz over the variable response time (for a 50Hz UK grid application, high side response). It is worth noting that on this scheme, there is no incentive to power cycle the device -this will simply reset the timer, and either not change or further delay the start time. Typical response functions (Fig 2b, 3b, 4b, 5b) are shown, with their resultant grid responses (Fig 2a, 3a,4a,5a) Optional refinements include: * Having the grid stress sensor on in standby mode, so that an immediate start decision can be made on power up, taking into account recent past grid behavior (eg running a line frequency prefilter).

* Having the response function dependant on a real time clock, by time of day and or season.

* Having the response function remotely altered (eg by radio), to optimize grid response, or dependant on commercial transactions (ie whether the transmission network operator has purchased this demand side response or not) By adapting the frequency threshold vs time delay profile, in combination with grid frequency statistics, a selection of different response profiles can be generated, in particular: linear high side (consumption change proportional to frequency divergence above nominal up to a maximum threshold)(see figs 2a, 2b), Linear low side (consumption change proportional to frequency divergence below nominal down to lower limit)(figs 3a,3b), and combined high and low side (consumption change proportional to deviation in frequency from nominal between high and low limits)(fig 4a, 4b), and finally a predominantly high side with a smaller low side response (figs 5a,5b).

High side response.

The delayed start protocol described can stack up many devices waiting to start, triggered by a high side line frequency excursion. The more devices, and the higher each of their loads on trigger, the higher the high side response obtained.

This is best provided by devices that can wait for a while before starting without undue user inconvenience. The other metrics are how high the power at startup is -response is normally required within 10 seconds, so for devices with a slow ramp up use the power at 10 seconds after start. Duration of power usage is an issue if it for less than the delay time to nominal frequency and less than 30 to 15 minutes -if this is the case, reduce the calculated response accordingly. Very long nominal delays (many hours or days) on long processes also start to reduce the available response, as the possible need for more than one demand for response during the period cannot be met.

Approximate amount of response provided is equal to the nominal delay (hours)times the power times average number of starts per hour. Alternatively, this can be expressed as the average power (24/7) of the device or process times the ratio of the nominal delay time to the process time (assuming process uses power at a constant rate -if not use the time that would give the same energy as the real process based on the power 10 seconds after start.

High side response example -Electric vehicle battery charger: Maximum consumption 3kw. Full charge energy 10kWh, Typical charge energy around 7kwh. Average load (24hrs, 365 days) 100W. Total charge time 7hrs. Consumption for first 30 mins 3kw. On charge start request, starting line frequency threshold is 50.5Hz (nominal line maximum), declining linearly to Hz (nominal minimum) over 2 hours, at which point if still untriggered the load will start independent of line frequency. This gives a typical high side response of 2hrs*3kw* 100W/7kw, = 86W.

Note that the high side response can be more or less than the average (24/7) consumption of the device, and is proportional to the starting power times the nominal delay.

Low side response: As devices are not turned off on this protocol, low side response is only obtained by not starting devices that would otherwise have started. Thus to get low side response that reacts in lOseconds or less, only those devices that would have started in that 10 second period if line frequency was at nominal can come into play. Similarly for low side response reacting in 30 seconds only those devices that would have started in that 30 second period contribute.

Thus for fast reacting low side response, nominal delays greater than 30 seconds do not increase the average available response in that period.

Thus fast low side response is best provided by high power process with many starts, and a short nominal delay.

The amount of response available is the startup power times the number of starts per unit time times the maximum delay acceptable for the response.

For example, a domestic kettle, doing 2.7kw for 2 minutes, 10 times per day would have an average daily consumption of 0.9kwh, or 37.5W. With a start delay varying from zero at nominal line frequency to 30 seconds at minimum line frequency, this would give a low side response of 2.7kw* 30sec/2.4hrs = 9.5W across the 24 hours. In reality this would be weighted to the times that kettles are most used -which is probably quite useful. The average amount of response available to react within 10 seconds will be 1/3 of this.

An equivalent high side response could be added with no additional hardware, see figures 4a, 4b. If a nominal delay of 30 seconds was used, this would give a high side response equal to the low side response. However this would then give an average delay to starting the kettle of around 33 seconds, vs the average delay of only 3 seconds (typical, dependant on grid statistics) for the low side only implementation.

Similarly, a low side response could be added to the battery charger in the high side example. This would be done by adding a 30 second long tail to the 2 hour high side delay (see fig 5a, Sb). However, due to the relatively infrequency of charge cycle starts, (compared, for instance, to a kettle), the amount of low side response obtained is small; around 0.7W for 30 second reaction, or 0.24W for 10 second reaction.

Delayed Start Responsive Load A Dames 22 January 2009 Likely PC class HO 2J 3/14 References: US4317049, MIT, FAPER PNNL 17079 Gridwise project, http://gridwise.pnLgov/docs/gfa proect final report pnnll7O79.pdf J Short et al, IEEE, http:jjieeexplore.ieee. org/starnp/starnpsp?arnurnber=4282051&isnumber=42820O2 US20070129851.."shedding" US20070222294A1..."shedding" EP1739806A4...shedding US20050180075A1...shedding US20070198133A1... adjusts power consumption U57242 114... "shedding" U56603218.."complex cooperative system..

US6314378...Shedding EP0893001A4..shedding US4385241...shedding GB 2361118...continuous operation, timing adjust.