EP3353872A1

EP3353872A1 - System for balancing an electricity network with a grid

Info

Publication number: EP3353872A1
Application number: EP16767041.3A
Authority: EP
Inventors: Egbert Wouter Joghum Robers; David BEIJER; Wouter Smit; Jorrit LUCAS
Original assignee: Peeeks BV
Current assignee: Peeeks BV
Priority date: 2015-09-22
Filing date: 2016-08-30
Publication date: 2018-08-01
Also published as: NL2015490B1; WO2017052362A1

Abstract

The present invention relates to a system for balancing an electricity network with a grid, at least one power source and multiple loads, the system comprising at least one load control unit, for determining a regular energy consumption of one of the power loads, determining a current state of said load, storing and/or communicating the determined values, switching the load on and off the network based on the determined state and a switching signal, at least one grid behavior model unit, for modelling grid behavior based on grid parameters and/or pricing parameters, storing a grid behavior model in a database, at least one central controller, having access to the database, for calculating a switching strategy for matching the power requested by the loads and the power available from the power source, based on the grid behavior model from the database, the current state of the at least one load and sending a switching signal to the load.

Description

System for balancing an electricity network with a grid

The present invention relates to a system for balancing an electricity network with a grid.

An electrical grid in itself cannot store energy. At any moment, the amount of electricity generated and put into the grid has to be equal to the amount of electricity drawn by energy consumers. If there is a shortage (the whole grid needs more electricity than what is generated) this can lead to a dropping of voltage in the grid, also called a brown- out. Brown-outs can damage equipment connected to the grid. Also, the whole grid or parts thereof can experience black-outs: when this occurs the energy transport comes to a full stop and the grid has to be started up completely. If there is a surplus of electricity (i.e. more electricity is generated than what is consumed) this can lead to over voltage, which can also damage equipment connected to or part of the grid. Also these over voltages can result in tripping of safeguards, creating a black out.

Surpluses or shortages also have other effects on the grid besides the voltage drops or spikes. The spinning mass present in generators connected to the grid experiences a larger resistance, as the load on the grid is larger than was intended. This leads the generators to slow down, which in turns leads to a lower frequency in the grid, as the rotational speed of these generators is tied directly to the grid frequency. This poses an additional reason to keep the grid balanced: some equipment uses this grid frequency as a time keeping mechanism. Keeping stability of the grid is not easy, and the complexity of this task is increasing. Not only is the amount of generators and consumers running in the millions, but many of the consumers are not under any form of control by the parties responsible for keeping the grid stable: any individual can at will switch on or off their light, water heater, computer or other appliance. On top of this, the amount of electricity consuming assets is ever increasing, and due to the prices of photovoltaic panels dropping, the amount of electricity generated out of control of utilities or grid operators is also increasing. Due to the increase in photovoltaic and wind generators driven by government incentives to promote clean energy generation the relative amount of controllable generation compared to controlled generation is increasing: wind and solar is strongly depending on weather circumstances.

Most countries have a national organization charged with the task to keep the national grid balanced, the national grid operator, sometimes also referred to as TSO which stands for Transmission System Operator. These grid operators use a multitude of ways to do this such as multilateral agreements with large energy consumers such as steel mills, rewards for large consumers that implement frequency-control or parties that are willing to run of their emergency power generators upon the request of the grid operator.

Another way to regulate this is by using an open imbalance market where real-time electricity prices keep the grid stable: in such a system energy consumers can choose to buy their electricity from a spot market where a price for electricity is determined multiple times per day, variations exist ranging from a price every minute to pricing per hour. This way energy consuming parties can be incentivized to schedule their consumption based on the actual state of the grid: a shortage of electricity will lead to high prices which will hopefully lead parties to postpone their electricity consuming processes, whereas surpluses will lead to low or even negative prices, leading to people receiving money for consuming electricity. If parties have processes that can be started quickly this should lead them to put these surpluses to good use.

These incentives lead to a number of companies buying their electricity straight from these spot markets. In order to get the best possible price for their electricity in order to make their processes as cheap as possible, the owners of electricity steering assets want to predict what will happen on the grid, since many assets require some time to start up or stop. Besides this prediction of the situation of the grid and prices of electricity, it is desirable to have a mechanism in place that automatically checks the process state and boundary parameters and decides to switch on or off certain assets. Having this automated will prevent human error, and allow the system to optimize faster, leading to a better match of the processes to the electricity prices, which in turn leads to the lowest costs of energy.

Currently predictions of the grid status are already being made by different parties. TSO's, utilities, commodity-traders, large energy consumers such as steel mills already have a keen interest in predicting the grid behavior. This is done by a combination of signal processing, monitoring certain assets, weather models and behavioral models of consumers (e.g. on a hot summer day more air-conditioning will be used). All these methods have their drawbacks, as many of the information required to get accurate predictions is either not publicly available (state of certain assets in the grid), or has a low resolution (both in time and space, such as weather predictions),

It is known that in some cases net frequency is used to predict the imbalance on the system. However, a frequency is not a good quantity for a prediction, because it only shows an actual net status, but gives no indication of future behavior.

It is therefore an object of the present invention to provide a solution that provides better results in demand response, or at least to provide a useful alternative to the state of the art.

The invention thereto proposes a system for balancing an electricity network with a grid, at least one power source and multiple loads, the system comprising at least one load control unit, for determining a regular energy consumption of one of the power loads, determining a current state of said load, storing and/or communicating the determined values, switching the load on and off the network based on the determined state and a switching signal, at least one grid behavior model unit, for modelling grid behavior based on grid parameters and/or pricing parameters, storing a grid behavior model in a database, at least one central controller, having access to the database, for calculating a switching strategy for matching the power requested by the loads and the power available from the power source, based on the grid behavior model from the database, the current state of the at least one load and for sending a switching signal to the load.

A switching strategy is to be understood here as a set of conditions and measures to be taken based on the conditions. In particular it determines which loads are to be switched on or off under what circumstances. A switching strategy may be optimized based on current pricing, expected future pricing and the current state of all assets in the network. Optimization may in this case refer to a best match of supply and demand and/or a lowest price for a consumer. The load control units are connected to an electricity consuming asset in such a way that the unit can switch the asset on or off, but also check the current state of the asset and whether it is safe to switch. Besides that the unit can also measure the energy consumption of the asset at regular intervals in order to measure the effect of switching. This whole set of information and feedback of regulating actions is passed on back to the central control unit. The load control unit may further determine a regular energy consumption of one of the power loads and a current state of said load at regular intervals in order to measure or predict the effect of switching.

Preferably, the grid behavior model unit is coupled to at least one data gathering unit, for gathering current grid parameters such as frequency and phase shift, but also current pricing data, public usage data and environmental parameters such as weather data. This is combined in models with behavioral models of electricity consumption in the grid (such as time of year, day of week and time of day leading to certain usage patterns) to reach a prediction of energy consumption and resulting energy prices. This data will be forwarded to the central orchestration unit such that these predictions can be taken into account when determining the optimal switching strategy.

The phase shift measurement module in the system according to the invention is to improve on the predictions and speed compared to similar solutions known as the state of the art. Frequency measurement is being used, but this is a reaction on things that have happened already, so it does not provide any predictionary value. The phase shift measurement module measures the phase shift of the 50Hz signal in the grid. Extremely small changes in this signal can provide an indication of changes in the grid equilibrium and can be detected faster than the published prices provided by the TSOs, so can provide a significant advantage in a system that tries to capitalize on these imbalance markets.

This invention also covers an embodiment of this grid behavior model that is implemented through neural networks that combine the input data to a prediction. Also covered is an embodiment in which the neural network is improved by reinforcement learning where the realized prices are fed back into the system and compared to the prediction, which is then used to optimize the neural network model to improve future predictions.

The invention will now be elucidated with reference to the following figures. Herein:

Figure 1 shows a variation where the load control units are integrated within the existing control systems of the electricity consuming assets or software-only; Figure 2 shows an embodiment wherein a meter has to measure the electrical installation;

Figure 3 shows an embodiment wherein a power using plant is coupled to two balance responsible parties;

Figure 4 shows a graph of the phase shift over a period of a year;

Figure 5 shows an embodiment of the system where a control system (3) is used to control several assets;

Figure 6 shows an embodiment of the method where information about one or more assets is communicated to a control system

Figure 7 shows an embodiment of the system where a control and/or metering device is coupled to the control system and a trading system or grid balancing system; and

Figure 8 shows an embodiment of a device that measures and determines the phase shift and actions from the measurements.

Figure 1 shows a variation where the load control units are integrated within the existing control systems of the electricity consuming assets or software-only. The system comprises a processing unit (3), connected to a controller (10) that controls an electric installation (5). The processing unit is also connected to other systems, such as energy trading systems (1) and/or grid balancing systems (2). The processing unit is connected to a database (31) of properties and/or states of the electrical installations and controllers. Also it is connected to a set of optimization rules, a database of (historic) behavior of the energy trading systems and/or grid balancing systems (33). Furthermore the processing unit is connected to a grid behavior model (34), which contains relations between measured parameters (such as phase shift of the grid) and future behavior of the grid.

A particular implementation of this invention is for distributed computing. Cloud computing prices fluctuate but are very much coupled to demand and cost of computing. An owner of distributed computing hardware can choose to lower its prices when electricity prices go down. Therefore offering services at a lower price and cost than others. For this the platform can send price information to other servers that make decisions for their own price structure. There are several markets where electric installation control can be monetized. A single electrical installation can be used for various markets. This requires the electrical installation to be split up into several virtual installations to be controlled as separate installations. Figure 2 shows an embodiment wherein a meter 10 has to measure the electrical installation 5. This meter data is transmitted to the platform 3. The platform or the meter splits the measurements into several virtual installations that can be controlled separately. The virtual installations are communicated as single installations to servers of different markets 21, 22, 23. Control signals from the markets for the virtual installations are aggregated by the platform and sent to the electrical installation.

This is also possible where the electrical installation is a group of installations.

Typically, the electrical installations are connected through a connection to a platform. In a particular embodiment, one or more electrical installations are connected to a sub- optimizer, that optimizes the installation(s) and present them to the platform as an aggregated controllable system. The platform checks the availability of the electrical installation, but also the quality of the connection to the electrical installation.

Besides directly reacting to the platform's instructions, the electrical installation receives pre-programmed control actions or strategies. This increases reliability of the system. If the connection to the platform fails, the electrical installation can resume functionality on its own. For example the day-ahead prices of electricity can be used to control the electrical installation in the case that the connection fails. Also weather data can be used to predict price fluctuations. A sensor that measures the phase and/or the frequency of the mains can be used to control the device, when the connection is not present. That sensor can give its measurements to the platform. The platform can then use multiple measurements to create a more accurate model of the state of the grid. Particularly when the grid interconnections are separated and the grid breaks into several stand-alone sub grids. The electrical installation buffers measurements and control actions to send to the platform when the connection resumes.

Based on the grid behavior model several strategies can be made for the automated energy trading. The trading strategies can be determined and optimized by back testing of strategies on historic data. The trading can be performed by a trained system, using historic data. For example a neural network, fuzzy logic or subspace identification. The frequency or the phase shift of the mains is used as a prediction of the imbalance market. With regards to the optimization of the grid, the system can measure electricity use of one or more of the electrical installations.

Figure 3 shows an example wherein a site (48) has an electrical connection and gets electricity from the grid (44) receiving electricity from BRPl (41). However BRP2 (42) wants to control an electrical installation (46) on the site. The location/site of the electrical installation will usually have a grid connection. That grid connection may be in the portfolio of a certain Balance Responsible Party (BRP). These parties (usually electricity companies) are responsible for the balance between the electricity production and electricity use. However if the electrical installation is in the portfolio of another BRP. This means that the BRP may not be able to get the full potential out of controlling the electrical installation. Also this could result in a suboptimal balancing of the grid on a country level.

Therefore the platform can use the measurement data to make a payment happen from one BRP to another. This way the BRP of a site does not have to be the same BRP of the electrical installation on the site.

This figure shows an example wherein a site has an electrical connection and a meter (44) and gets electricity from BRPl. However BRP2 wants to control an electrical installation on the site. That is fine, however every steering action of the electrical installation is credited to BRPl. The proposed system corrects this fault by adding an additional meter (45) to the electrical installation on the site. This way the steering actions can be measured and a financial correction can be made between BRPl and BRP2 by the control system (10).

A special case of this is when there are profiled customers. Profiled customers are not measured by themselves. Usually this is done in large urban areas: Several households are combined, an estimate is made how many households get electricity from BRPl and how many from BRP2 (and/or BRP3...A...etc). Then a ratio is agreed. The problem with this is that a steering action in a profiled customer credits not only the appropriate BRP, but also its competitors (according to the ratio). This prevents most BRPs to do this. The proposed system changes this by using the meter at a site to measure steering actions and to make financial corrections. Figure 4 shows a graph of the phase shift over a period of a year. The horizontal axis shows the time over a year and the vertical axis shows the amount of seconds deviation. This shows that over a year's period, the phase shift is steered towards a phase close to 0. But the deviation can be as big as 100 seconds. A structural deviation of the phase shift indicates a structural power shortage or surplus. Therefore a change in the phase shift reflects a shortage or a surplus and therefore predicts if grid stabilizing parties (e.g. TSO/DSO) or energy trading parties will start to contract electricity assets.

An embodiment uses the following rules:

1. If the phase shift is behind (meaning the average frequency has been low in the past) and there is another phase shift that puts the grid even more behind, then the grid has a shortage. The result will be that it is likely that grid balancing systems and energy trading systems will contract, deploy or buy more energy generation (or decrease use).

2. If the phase shift is behind (meaning the average frequency has been low in the past) and there is another phase shift that puts the grid less behind, then the grid is restoring. The result will be that it is likely that grid balancing systems and energy trading systems will do nothing (or less than in the previous situation).

3. If the phase shift is preceding (meaning the average frequency has been high in the past) and there is another phase shift that puts the grid even more in the lead, then the grid has a surplus. The result will be that it is likely that grid balancing systems and energy trading systems will contract, deploy or buy more energy use (or decrease generation).

If the phase shift is preceding (meaning the average frequency has been high in the past) and there is another phase shift that puts the grid less in the lead, then the grid is restoring. The result will be that it is likely that grid balancing systems and energy trading systems will do nothing (or less than in the previous situation). Figure 5 shows an embodiment of the system where a control system 3 is used to control several ADR assets 5 through several regional or local service APIs/control systems 4. Control system 3 aggregates the assets to one or more services to grid balancing 2 systems and/or energy trading systems 1. The individual assets 5 may not be distinguishable by the grid balancing or trading systems 2, 1.

Figure 6 shows an embodiment of the method where information, such as state of the asset, about one or more assets 5 is communicated 8 to a control system, the availability (amount of power, duration of availability) of the group of assets is received by the control system 3. Control system 3 combines assets into groups that have most value for a certain trading systems 1 or grid balancing systems 2. The groups and their aggregated states are then communicated 6 with the trading systems 1 or grid balancing systems 2. Then the trading systems 1 or grid balancing systems 2 respond 7 what amount of power they want to deploy.

The control system 3 determines which assets need to change their power consumption of generation and communicates 9 that to the individual assets (or though sub controllers, not depicted).

Figure 7 shows an embodiment of the system where a control and/or metering device 10 is coupled to the control system 3 and a trading system 1 or grid balancing system 2. The metering device is used to prove control actions. The metering data can be sent directly 11 to one or more trading systems 1 or grid balancing systems 2 or through the control system 6. The data can be aggregated and sorted to improve processability for the trading system 1 or grid balancing system 2. Figure 8 shows an embodiment of a device that measures and determines the phase shift and actions from the measurements. The system is connected to one or more grids 86 through a grid connection 85. The coupling may be electric and/or magnetic or any type it can determine the grid frequency. One or more counters 84 count the cycles per second. This can be done by measuring the time between cycles or counting the number of cycles in a time period. For example a Schmitt-trigger can be used to turn the analog wave signal into a digital signal. From there the number of cycles per second can easily be determined. The system requires at least one particularly accurate clock 83. This at least one clock can, for example, be a tuned quartz clock, an atomic clock, a GPS receiver, a DCF77 receiver, an internet time server or alike. If the clock does not receive absolute time (such as GPS or DCF77) but needs to be set on time, this may be done in operation. A way to do this is to use the grid frequency as a guide. In some regions the grid frequency is increased by lOmHz in the phase shift is more than 20s behind and decreased if the phase shift is more than 20s leading. So a structural change in the grid frequency can be used to determine how much the grid is lagging. For example: A grid frequency of 50.01 Hz changes to 50Hz. This means that the grid is lagging by 20s. When the phase shift is measured this is communicated with a processing unit (82) and at least one grid behavior model. The model and/or processing unit may contain historic data about the phase shift.

Claims

System for balancing an electricity network with a grid, at least one power source and multiple loads, the system comprising:

At least one load control unit, for

• determining a regular energy consumption of one of the power loads;

• determining a current state of said load;

• storing and/or communicating the determined values;

• switching the load on and off the network based on the determined state and a switching signal;

At least one grid behavior model unit, for:

• Modelling grid behavior based on grid parameters and/or pricing

parameters;

• storing a grid behavior model in a database;

At least one central controller, having access to the database, for:

• calculating a switching strategy for matching the power requested by the loads and the power available from the power source, based on:

^■ the grid behavior model from the database;

^■ the current state of the at least one load;

• sending a switching signal to the load.

System according to claim 1, wherein the grid behavior model unit is coupled to at least one data gathering unit, for gathering:

• current grid parameters such as frequency and phase shift;

• current pricing data;

• public usage data and environmental parameters such as weather data.

System according to claim 2, wherein the data gathering unit is configured for combining the gathered data with behavioral models of electricity consumption in the grid to reach a prediction of energy consumption and resulting energy prices.

4. System according to any of the preceding claims, wherein the grid behavior model uses a phase shift of the grid to determine if there is a surplus or a shortage in the electricity grid.

System according to any of the preceding claims, wherein the load control unit

• determines a regular energy consumption of one of the power loads;

• determines a current state of said load;

• at regular intervals in order to measure the effect of switching.

System according to any of the preceding claims, wherein the grid behavior model unit comprises at least one neural network that combines the grid parameters and/or pricing parameters to a prediction for the grid behavior model.

System according to claim 5, wherein the neural network is improved by reinforcement learning where gathered grid parameters and/or pricing parameters are fed back into the system and compared to the prediction, which is then used to optimize the neural network model to improve future predictions.

Method of determining the state of an electricity grid balance, comprising the following steps:

a. Determining the phase shift over a longer time period such as a year; b. Determining the phase shift over a shorter time period such as an hour; c. Comparing the two phase shifts and determining the state of the grid based on a stored grid behavior model that relates a phase shift to a state of a grid;

d. Switching electrical loads or generation according to the determined state of the grid.

Method according to claim 8, where the following happens:

a. If the phase shift over said longer period is behind and the phase shift over the shorter period is behind, then switch off electrical load and/or switch on generation.

b. If the phase shift over said longer period is leading and the phase shift over the shorter period is leading, then switch on electrical load and/or switch off generation.