WO2018083202A1

WO2018083202A1 - System and method for scheduling energy consumption in a network

Info

Publication number: WO2018083202A1
Application number: PCT/EP2017/078103
Authority: WO
Inventors: Richard LINGER; Juan Manuel ESCANO GONZALEZ; Adolfo Juan SANCHEZ DEL POZO FERNANDEZ
Original assignee: Cork Institute Of Technology
Priority date: 2016-11-02
Filing date: 2017-11-02
Publication date: 2018-05-11
Also published as: EP3536048A1

Abstract

A system and a method are disclosed for optimising power consumption in a network comprising at least one node powered by a power source and controllable by a terminal. The system comprises a plurality of networked terminals, including the node controlling terminal and at least one system controller terminal. The system controller terminal is adapted to model power consumption of a node according to characteristics of the node, and input and output constraints of the network. The system controller terminal is adapted is adapted to forecast power consumption of the or each node according to the modelled power consumption and/or meteorological data received from a meteorological data source as a first data feed and/or a second data feed, including at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes. A scheduler computes a schedule for the powering of the or each node according to the forecast power consumption and/or the meteorological data and/or the second data feed with a receding horizon function; and command the or each node controlling terminal to control powering of the or each node according to the optimised schedule.

Description

SYSTEM AND METHOD FOR SCHEDULING ENERGY CONSUMPTION IN A NETWORK

Field

[001] The present invention relates to a system and method for scheduling energy consumption in a network with power nodes. More particularly, the present invention relates to a system for optimising energy usage across the various electrically-powered nodes of a distributed processing architecture, such as a water processing network, a manufacturing plant, a building or the like.

Background

[002] The scheduling of energy consumption in large-scale materials processing and goods handling environments is a perennial issue and the subject of constant research and improvements, directed so much at the configuration and characteristics of electrical, hydraulic and other motors, as to the correcting and rectifying of utility-supplied electrical power, the distribution and arrangement of loads according to physical and quality of service constraints, and more.

[003] On the one hand, physical constraints and required output and/or supply levels of the networks which such large-scale materials processing and goods handling environments constitute, can be forecast reasonably accurately. On the other hand, the various capacities, operational parameters and other characteristics of the powered nodes in such networks, are known. The optimisation quandary facing the skilled person is how best to employ the powered nodes in the network, to match the required output or supply levels with the network constraints at any one time and at least operational cost and, having regard to more recent environmental imperatives, at least environmental impact or 'carbon footprint', wherein this environmental constraint is not necessarily synonymous nor compatible with the least cost imperative. [004] Considering the field of water treatment networks by way of non-limitative example, such networks typically include a potable water sub-network and associated potable water treatment plant(s), a wastewater sub-network and associated wastewater water treatment plant(s), with each sub-network including one or more tanks, pumps, conduits, inlets, outlets, collectors and more. [005] Water treatment plants traditionally have control systems in place, known as Supervisory Control and Data Acquisition ('SCAD A'). Such systems are known to record or otherwise collect data from pumps and processes around a plant and/or the network to which it belongs, and to communicate that data to a control location at which a plant operator may observe changes to critical systems and make control decisions. Recorded data typically includes flow rate, pressure, energy usage, temperature, water quality, and potentially many further parameters and characteristics.

[006] Building on this layer of data acquisition, the prior art contains a number of automated decision making systems, embodied as software packages, conferring a degree of intelligence to SCADA architectures and which monitor the recorded data against predetermined values, then output control decisions based on the comparison outcomes. Regardless of any specific implementation within a water treatment context, such automation can be distinguished as a plant optimisation system and/or a network management system. Such systems are focused on optimising electricity consumption and shifting loads from peak electricity tariffs, or focused on monitoring network health and security, for instance adapted to, in a water treatment context, detect leaks or bursts and predict storm water overflow. Numerous prior art systems exist, for example US Patent publication number US 2015/310461 discloses a system for determining and recommending an optimal electricity rate system. However this system is effectively restricted to choosing a preferred tariff rate in which a device can be powered.

[007] Given recent statutory imperatives to now actively mitigate environmental degradation, particularly to reduce emissions of carbon dioxide associated with just about all powered processing of any material or resource, none of these prior art systems are considered apt to provide an environmentally-focused optimisation. There is still no system which bridges the difference of approach between plant optimisation at a fine-grained time-step granularity and plant state monitoring, with the objective of minimising power draw at the optimum time. It is an object to provide a system for optimising scheduling of power consumption in a network to overcome at least one of the above problems.

Summary

[008] The present invention mitigates shortcomings associated with the prior art of reference by providing an automated system adapted to dynamically schedule the powering of network nodes (i.e. loads), through a combination of network and nodal constraints modelling, and/or weather correlating.

[009] According to an aspect of the present invention therefore, there is provided, as set out in the appended claims, a system for optimising power consumption in a network or energy system, wherein the network or energy system comprises at least one node powered by a power source and controllable by at least one data processing terminal, wherein the system comprises a plurality of networked data processing terminals including the at least one node controlling terminal, at least one system controller terminal and a meteorological data source; and wherein the at least one system controller terminal is configured with: modelling means for modelling power consumption of the or each node according to characteristics of the or each node, input constraints of the network and output constraints of the network; forecasting means for forecasting power consumption of the or each node according to the modelled power consumption and meteorological data received from the meteorological data source; and scheduling means for computing a schedule for the powering of the or each node according to the forecast power consumption and the meteorological data with a receding horizon function; and communication means for commanding the or each node controlling terminal to control powering of the or each node according to the optimised schedule. [0010] The technical problem which the above system solves, is how to implement technical requirements associated with environmental preservation policies, expressed as a reduction of carbon dioxide level associated with plant activity, into industrial network optimising systems. The problem is solved by generating a schedule optimising plant machinery activity, based on a determination of both the effect of weather on the plant activity and throughput and the level of renewable energy on the power supply grid, according to which industrial network node powering can be delayed or must be undertaken immediately. The optimising schedule is generated both to avoid network undersupply and to minimise carbon dioxide emissions associated with the power consumption of the network nodes.

[0011] In one embodiment of the invention the system can implement energy cost optimisation for water pumping use-cases. For a given site or system, the system can optimise the performance of a specific parameter within the site-specific constraints that apply. The specific parameter that can be optimised is energy cost.

[0012] In one embodiment the cost of energy consumed can be significantly reduced for the given site or system by using a risk-managed pass through electricity market price. This is done by comparing predicted energy price against energy demand requirements on a ookahead' basis, for that system; and actively seeking the least-cost solution for the predicted energy demand. Suitably, a Model Predictive Control (MPC) strategy with receding horizon is applied to minimize a cost function along the horizon. This results in an optimised schedule for running the relevant energy-consuming assets, such as pumps; and for controlling any associated ancillary equipment such as valves.

[0013] In an embodiment of the system, the modelling means may comprise a modelling module, the forecasting means may comprise a demand prediction module and the scheduling means may comprises a model predictive controller operably interfaced with the modelling module, the demand prediction module and the communications means.

[0014] In an embodiment of the system, the meteorological data may be supplied by the meteorological data source to the scheduling means. [0015] In an embodiment of the system, the meteorological data may comprise one or more selected from wind speed, rainfall level and solar irradiation in a predetermined time period. In an alternative embodiment, the meteorological data may comprise instead, or also comprise, one or more selected from wind speed, rainfall level and solar irradiation monitored by the meteorological data source in real-time. [0016] In an embodiment of the system, the plurality of networked data processing terminals may further include a power cost monitoring terminal, and the system controller terminal may be further adapted to compute the schedule optimising the powering of the or each node according to the forecast power consumption, the meteorological data received from the meteorological terminal and power cost data received from the power cost monitoring terminal. In a variant of either of these embodiments, the power cost data may be supplied by the power cost monitoring terminal to the scheduling means.

[0017] An embodiment of the system may be implemented in a water network comprising one or more water tank(s), wherein the or each node is a water pump or valve or other item that is electrically powered associated with the one or more tank(s). [0018] According to another aspect of the present invention, there is also provided a method for optimising power consumption in a network, wherein the network comprises at least one node powered by a power source and controllable by at least one data processing terminal, and wherein the at least one data processing terminal is controllable by at least one remote system controller terminal associated with a meteorological data source; the method comprising the steps of, at the at least one remote system controller terminal, modelling power consumption of the or each node according to characteristics of the or each node, input constraints of the network and output constraints of the network; forecasting power consumption of the or each node according to the modelled power consumption and meteorological data received from the meteorological data source; computing a schedule for the powering of the or each node according to the forecast power consumption and the meteorological data with a receding horizon function; and commanding the or each node controlling terminal to control powering of the or each node according to the optimised schedule.

[0019] An embodiment of the method may comprise the further step of receiving monitored network data at the at least one remote system controller terminal from the node-controlling data processing terminal. In a variant of this embodiment, the step of forecasting may further comprise sampling the monitored network data at a predetermined time interval and limiting the receding horizon with the or another predetermined time period. [0020] In an embodiment of the method, the meteorological data may again comprise one or more selected from wind speed, rainfall level and solar irradiation in a predetermined time period, and variations may provide for the meteorological data to be monitored by the meteorological data source in real-time.

[0021] According to still another aspect of the present invention, there is also provided a computer program product, comprising a computer usable medium having computer readable program code instructions embodied therein, said computer readable program code adapted to be executed to implement a method for optimising power consumption in a network as described herein.

[0022] In one embodiment there is provided a system and method for optimising scheduling of power consumption in a network,

wherein the network comprises at least one node powered by a power source and controllable by at least one data processing terminal ;

the system comprises a plurality of networked data processing terminals including the at least one node controlling terminal and at least one system controller terminal; and

the at least one system controller terminal is configured with:

modelling means for modelling power consumption of the or each node according to characteristics of the or each node, input constraints of the network and output constraints of the network; forecasting means for forecasting power consumption of the or each node according to the modelled power consumption and a second data feed, including at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes; scheduling means for computing a schedule for the powering of the or each node according to the forecast power consumption and the second data feed with a receding horizon function or other mathematical mapping function; and

communication means for commanding the or each node controlling terminal to control powering of the or each node according to the optimised schedule.

[0023] In another embodiment there is provided a system and method for optimising scheduling of power consumption in a network,

wherein the network comprises at least one node powered by a power source and controllable by at least one data processing terminal;

characterised in that

the at least one system controller terminal is configured with:

modelling means for modelling power consumption of the or each node according to characteristics of the or each node, input constraints of the network and output constraints of the network; forecasting means for forecasting power consumption of the or each node according to the modelled power consumption and/or meteorological data received from a meteorological data source as a first data feed and/or a second data feed, including at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes;

scheduling means for computing a schedule for the powering of the or each node according to the forecast power consumption and/or the meteorological data and/or the second data feed with a receding horizon function or other mathematical mapping function; and

[0024] Other aspects of the present invention are as stated in the appended claims. Brief Description of the Drawings

[0025] The invention will be more clearly understood from the following description of an embodiment thereof, given by way of example only, with reference to the accompanying drawings, in which: -

[0026] Figure 1 is a block diagram of an embodiment of an optimising system according to the invention, including a model predictive controller.

[0027] Figure 2 is a simplified and modular representation of the functionality of the model predictive controller. [0028] Figure 3 illustrates a network with powered nodes of the prior art, in the example a water processing network having a plurality of water pumps controlled by a data processing terminal, capable of optimisation with the model predictive controller of Figures 1 and 2. [0029] Figure 4 shows a plurality of data processing terminals in a data communication network, including at least the node controlling terminal of Figure 3, a meteorological terminal, and at least one system controller terminal configured into the model predictive controller of Figures 1 and 2 to optimise the network of Figure 3. [0030] Figure 5 details components of the system controller terminal of Figure 4, including memory means.

[0031] Figure 6 illustrates data processing steps performed by the system controller terminal of Figure 4. [0032] Figure 7 shows the contents of the memory means of Figure 5 when the system controller terminal performs the steps of Figure 6.

[0033] Figure 8 illustrates data processing steps performed by the system controller terminal of Figure 4, when configured according to an alternative embodiment.

[0034] Figure 9 shows the contents of the memory means of Figure 5 when the system controller terminal performs the steps of Figure 8.

[0035] Figure 10 shows graphs charting optimised reservoir levels against real reservoir levels.

[0036] Figure 11 shows graphs charting optimised pump activation against real pump.

[0037] Figure 12 shows graphs charting the level, supply demand and pumping rate of a test low reservoir.

[0038] Figure 13 shows graphs charting the level, supply demand and pumping rate of a test intermediate reservoir.

[0039] Figure 14 shows graphs charting the level, supply demand and pumping rate of a test high reservoir.

Detailed Description of the Drawings

[0040] Referring now to the figures and initially Figures 1 and 2, an embodiment of the operational architecture of an optimising system according to the invention is shown as a diagram of functional blocks, associated with a network 10 comprising a plurality of powered nodes 11, for instance a site or installation plant 10 configured with a plurality of electricity grid-powered loads 11 such as, in a specific and non- limitative example described in further detail hereafter, a water network.

[0041] The core and complementary functions delivered by the system of the invention include plant network optimisation, network throughput management and optimisation, network demand forecasting and weather data factoring for reduction network-associated carbon dioxide, on the basis that meteorological and/or forecast data can be used both to calculate its impact on the network processes and throughput. In another aspect the invention can determine the level of renewable energy on the power grid, with the aim of minimising carbon dioxide consumption as opposed to energy consumption or energy cost. In one embodiment a carbon dioxide intensity profile can be generated and input into the forecast model. For example green or renewable generated power can be used to supply nodes in the network and an optimum schedule can be calculated to power the nodes based on the amount of green or renewable energy. This can be important for a network owner to increase carbon credits and minimise any possible Government fines that could be imposed for using carbon based energy. Suitably, at least one battery in the network can be charged at optimum charging times using renewable energy. The computed optimised schedule can take account of times when a charged battery can be used to power one or more of the power nodes in the network when it is more efficient than powering from a main grid supply, whilst at the same time reducing the amount of carbon energy that can be used. [0042] Given the example context of the water network 10, the scope of the system and method of the invention is to optimise the energy usage across the entire water processing spectrum, effectively determining whether pumping can be delayed or must be undertaken immediately. The system is compatible with existing monitoring and control SCADA systems, allowing easy interfacing with legacy node- and network-controlling and monitoring hardware.

[0043] The system must initially identify where flexibility exists within the water treatment processes, identifying bottle necks, critical processes and underused assets such as tank and/or pumping capacity. The system assesses the various processes which take place during water treatment, identifying both non-critical processes and flexibility in critical processes.

[0044] Once identified, this information can be used both for immediate optimisation and for network updating and redesigning, through the identification of topology and/or asset upgrades apt to deliver most benefit to the network, in terms of e.g. increased flexibility, reduced cost, maximised capacity.

[0045] The system uses both the determined flexibility within the processes and the weather impact on water treatment requirements, which will vary from plant to plant (due to hinterland size, geology, and population variations), using computer learning algorithms, to optimise the operation of the various controllable nodes, i.e. pumps 1 I_A-_C of the water network, reduce the overall power consumption of the nodes and increase the accuracy of the network model over time. [0046] The system can use both energy market price analysis and weather impact modelling on such electricity prices (e.g. impact of renewable energy on the grid) to output still more accurate predictions of future energy price, on the basis of which processes already optimised on the basis of process flexibility and meteorological data can be further optimised to run at the lowest possible energy costs. For instance, on the basis of day-ahead electricity tariffs, load scheduling can be generated to balance future demand with costs minimisation potential. A data feed can be provided to the forecasting module (14) for forecasting power consumption of each node (1 I_A-_C). The data feed can include at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes 1 I_A-_C- In such a scenario the meteorological data is not needed and not included in the optimisation. The data feed may comprise a profile of future electricity prices at a granularity that matches the market settlement periods - namely half-hourly or hourly - for periods of time ranging from a point in time, such as real time (time equals zero) to 7 days in advance of the time of consumption of the electricity in question. Examples include day-ahead electricity price indices from power brokers such as Nordpool in the United Kingdom; or Ex- Ante and Within-Day price profiles from the Single Electricity Market Operator in Ireland; or other relevant and suitable types and sources of such data. For a given time period of consumption, such data typically becomes available at predictable times. It is typically polled from suitable internet locations and merged / processed by the relevant elements of the invention herein described to create an appropriate input for optimisation. [0047] In one embodiment the system of the invention primarily focuses on optimising load activation to avoid high carbon dioxide production, therefore node powering by the power supply grid, through a combination of network modelling, network throughput and power demand forecasting, and load scheduling: the core purpose of the system of the invention is to generate and apply an optimised control strategy for the network 10, wherein the generation includes performing an accurate prediction of the network state, calculating the prediction error along a time horizon, and generating a set of commands for the network nodes that minimizes a carbon dioxide function. Accordingly, in this embodiment the optimising system comprises a modelling module 12 and a demand prediction module 14 associated with a model predictive controller ('MPC') module 16. [0048] The purpose of the modelling module 12 is both to model an installation or plant as a network comprising powered nodes, on the basis of the plant topography, powered machinery, characteristics and operating parameters of such machinery, inflows of materials and resources into the network, characteristics and operating parameters of the plant as a whole; and to constrain the spectrum of optimising solutions of the MPC 16 for the network, on substantially the same the basis as for the modelling, since the network output cannot exceed the operating parameters of such machinery, the maximum inflows of materials and resources into the network, or the maximum output level of the plant as a whole. In the water network 10 of the example, powered machinery typically include pumps; the network topography will extend to water storage tanks and any pressurising network; characteristics and operating parameters of the plant will encompass water standing time in the network, water aeration regimes, wastewater water processing rates, and more. [0049] The purpose of the demand prediction module 14 is to compute and forecast external demand level for the network output(s), on the basis of inflows of materials and resources into the network, network throughput, characteristics and operating parameters of the plant as a whole, and meteorological data to the extent that same influences the demand, which can also be predicated or real-time data. For instance, in the case of the example water network in which powered pumps constitute the network nodes, weather data analysis enables the prediction of network loads and nodal loads due to high temperatures and rainfall: during periods of warm weather, people generally use more water for personal hydration and hygiene, as well as for plant watering, but less water during periods of cold weather; and during periods of high rainfall, the volume of collected waste water increases, pumping of storm storage tanks may require optimisation, and infiltration can also occur. The demand prediction module 14 effectively provides the MPC 16 with an accurate prediction of the network state, complete with a predicted disturbance parameter.

[0050] The MPC module 16 receives plant-independent data that is relevant to computing the optimising schedule through a first data feed, including at least meteorological data 17 from a meteorological data source that is representative of historical, real-time and/or forecast meteorological conditions proximate the network 10 and preferably beyond. The meteorological data 17 is thus used in the system to predict both throughput of the network 10, in particular potable water usage levels and wastewater treatment loads, and renewable energy output, in particular wind and solar energy generation. [0051] In an alternative embodiment also described herein, such plant-independent data may further include a second data feed, including at least power supply grid data 19 from one or more utility suppliers, that is representative of historical, real-time and/or forecast unit cost of grid energy supplied to the nodes 1 I_A-_C- and may also be used to determine when power generation sources, for instance combined heat and power ('CHP') units or generators, are switched on and/or demand is off.

[0052] The optimising system further comprises a communication module 18, which operably interfaces the MPC module 16 with the network 10 through a real-time and bilateral data communication within which one or more data processing terminals, e.g. SCADA systems, of the network 10 communicate monitored plant and nodal data to the MPC 16 and receive optimising schedules from the MPC 16.

[0053] The optimisation performed by the system of the invention application is highly dependent upon the particulars of each plant configuration. Several simulations have been carried out by the inventors to test model predictive control strategy implementation, based on a real plant configuration and associated real plant data, and validated with a test rig of the real plant configuration. Figure 3 is a simplified diagram of a real life water network 10 installation.

[0054] The network 10 is made up of three water reservoirs 13_{a c} that supply water to different parts of a geographic region. The reservoirs are located at different heights, with a low reservoir 13_a, an intermediate reservoir 13b, and a high reservoir 13_c. [0055] Reservoirs 13_{a c} are assumed cylindrical with an inlet water flow (supplied by pumps) and an output flow (demand). The network 10 has several pumps 1 l_a._c and valves assigned to each of the water reservoirs and water transport is done through piping segments separately for each of the reservoirs. The flow rate of low reservoir 13_a is variable because it is formed by one or more VSD pumps 1 l_a, whereas the pumps 1 lb,c associated with the high and intermediate reservoirs 13b,_c are ON/OFF pumps.

[0056] The different water reservoirs are therefore sub-networks that are not coupled by the pumps or by distribution through the pipes. However, there is a filtered water tank 13_d from which pumps extract water to fill the reservoirs 13_a._c. This tank 13_d has a capacity of approximately 900m³ and an inflow rate of circa 1900m³/h. This tank 13_d cannot be emptied by explicit specification of the plant. So it cannot, generally, exceed a total pumping rate higher than 1900m³/h. Therefore, the three sub-networks are coupled through the filtered water tank 13d.

[0057] The network 10 is controllable by at least one data processing terminal (100) operated by the water utility agency and is a personal computer device which emits and receives data encoded as digital signals over wired and/or wireless data transmissions that are routed by a local router device 111 implementing a wired local network 110 operating according to the IEEE 802.3-2008 Gigabit Ethernet transmission protocol and/or again a high-bandwidth wireless local network operating according to the IEEE 802.11 Wi- Fi wireless transmission protocol. Such wired data transmissions include local incoming transmissions associated with monitoring network data within the network 10, such as respective water levels in the water tanks 13_A D, and local outgoing transmissions for controlling respective operation of the pumps 1 I_A-_C-

[0058] Such wired and/or wireless data transmissions further include data transmissions with terminals remote from the computer 100 and the network 10 itself, which can also be relayed respectively to or from the computer 100 via the router 111, which is itself connected to a wide area network ('WAN') 104 via a conventional ADSL or optical fibre connection over a wired telecommunication network 112, wherein the local router device 111 interfaces the computer 100 to the WAN communication network 104., an example of which is the Internet. [0059] With reference to Figure 4 now, an embodiment of the system according to the invention is shown interfaced with the water network terminal 100 over the WAN 104. The system includes at least one system controller terminal 101 and a meteorological data source 102, in the example a terminal operated by a meteorological agency. An embodiment of the system may optionally further include a power cost monitoring terminal 103 operated an electricity utility agency, which is also shown in Figure 4.

[0060] Each of the system controller terminal 101, the meteorological data terminal 102 and the optional power cost monitoring terminal 103 is remote from the water network controlling terminal 100 and from one another. Each of the system controller terminal 101, the meteorological data terminal 102 and the optional power cost monitoring terminal 103 is a personal computer device or PLC device which emits and receives data encoded as digital signals over wired and/or wireless data transmissions that are routed by a respective local router device 111 implementing a wired local network 110 operating according to the IEEE 802.3-2008 Gigabit Ethernet transmission protocol and/or again a high-bandwidth wireless local network operating according to the IEEE 802.11 Wi-Fi wireless transmission protocol. Such wired and/or wireless data transmissions include data transmissions with remote terminals, as each router 111 is again connected to the WAN 104 via a wired telecommunication network 112. Network connectivity and interoperable networking protocols of each terminal 100, 101, 102, 103 allow the terminals to connect to one another and communicate data to and receive data from one another according to the methodology described herein.

[0061] A typical hardware architecture of the node controlling terminal 100, the system controller terminal 101, the meteorological terminal 102 and the optional power cost monitoring terminal 103 is shown in Figure 5 in further detail, by way of non-limitative example. As skilled persons will readily understand, the hardware architecture of the terminals 100-103 may be substantially identical or similar to one another, with components designed for durability and redundancy of operation.

[0062] Each data processing terminal 100-103 is a computer configured with a data processing unit 201, data outputting means such as video display unit (VDU) 202, data inputting means such as HiD devices, commonly a keyboard 203 and a pointing device (mouse) 204, as well as the VDU 202 itself if it is a touch screen display, and network data inputting/outputting means such as the wired network connection 110 to the communication network 104 via the router 112, a magnetic data-carrying medium reader/writer 206 and an optical data-carrying medium reader/writer 207.

[0063] Within data processing unit 201, a central processing unit (CPU) 208 provides task co-ordination and data processing functionality. Sets of instructions and data for the CPU 208 are stored in memory means 209 and a hard disk storage unit 210 facilitates non- volatile storage of the instructions and the data. A wireless network interface card (NIC) 211 provides the interface to the network connection 110. A universal serial bus (USB) input/output interface 212 facilitates connection to the keyboard and pointing devices 203, 204.

[0064] All of the above components are connected to a data input/output bus 213, to which the magnetic data-carrying medium reader/writer 206 and optical data-carrying medium reader/writer 207 are also connected. A video adapter 214 receives CPU instructions over the bus 213 for outputting processed data to VDU 202. All the components of data processing unit 201 are powered by a power supply unit 215, which receives electrical power from a local mains power source and transforms same according to component ratings and requirements. [0065] Within the context of the system shown in and described with reference to Figures 3 to 5, the methodology of the invention is implemented as a data processing logic processed at the system controller terminal 101 for optimising power usage within the network 10, and now described with reference to Figure 6. Data processing steps of the methodology are described as a discrete group of chronological data processing tasks repeated iteratively at runtime. It will be readily understood by the skilled person that such steps may be optimised and, where appropriate, processed substantially in parallel, as the architecture of the CPU 201, and the basic instructions set and libraries for same allows. [0066] After powering up the system controller terminal 101 conventionally, an operating system (OS) is loaded into the terminal memory 209 and started locally at step 601, including communications subroutines of the OS and a user interface is eventually instantiated on the display 202.

[0067] To perform the optimization of the plant, one or more models that approximate the dynamics of each of the network's reservoirs and pumps are needed, and therefore generated at step 602. A model has to be designed specifically for each water network 10. Generally, but not always, modelling a water network with reservoir is based on a set of discrete time equations and, to perform the optimization of the network, models that approximate the dynamics of each of the reservoirs are needed. For example, although all reservoirs can be modelled with an inlet water flowrate, supplied by pumps, and an output water flowrate according to demand, not all reservoirs have the same dimensions, surface, countenance, elevation, or even shape. [0068] Thus, although the equations hereafter are generally applicable to each water network with reservoirs 13, they may differ for distinct networks 10 not only in the parameters values, but also in the model itself. In the example, the models used are simple models governed by the following equations: dV_y {t)

- <lu (t) - q_l0 (0

dt

dV₂ (t)

(t) - q_2o {t)

dt dV_n (t)

dt

Wherein V_i is the volume of a reservoirs, q is the water inlet flow to a reservoir (income) and q -_Q is the water outlet flows of a reservoir (demand).

[0069] In the example still, constraints for the water reservoirs 13A-D can be expressed as:

A₁ < h₁ < B₁

A₂ < h₂ < B₂

wherein ¾, are the heights of the reservoirs.

[0070] Additional constraints on the flowrate and the terminal constraints of the levels can be expressed as:

+ ·· · + q_n < ^c

terminal constraint h = D₁

terminal Constraint h₂ = D₂

terminal constraint h_n = D_n [0071] Thus, for each relevant component of the network 10, in particular the its reservoirs 13 _A D and pumps 11 _A-c, the modelling module 12 obtains data representative of its parameters, such as dimensions, elevation, inlet and outlet flowrates for a reservoir, and activation type, power rating and flow rate for a pump; and data representative of its constraints, such as limits of maximum and minimum heights of operation for a reservoir, and limits on time acting on a pump.

[0072] At step 603, the modelling module 12 validates the model of step 602 and estimates power consumption by each pump H_A-_C based on the parameters recorded at step 602 and the validated model. The validation of the model of step 602 is done in an iterative way: inflow and outflow of each of the reservoirs 13_{a c} are taken from the real plant data, and that data is sampled at 5 minutes to integrate the model with the data for 5 minutes, then a height is calculated on each reservoir 13_{a c}. The model is closed loop with the previous height computed through prediction, to obtain a validated prediction. [0073] The strategy of choice for solving the optimisation problem is a Model Predictive Control (MPC) strategy, generally preferred due to its numerous advantages that include ease of formulation of a multi- variable case, minimizing data processing costs, maximising network performance, consideration of physical limitations in each actuator, operation near restrictions-controlled variables and delay compensation. MPC strategies can be distinguishes as linear and non-linear.

[0074] MPC based on a linear model is one of the preferred control strategies for a large number of industrial processes, however situations exist where nonlinear effects justify a non-linear approach, because strongly non-linear processes are subject to large disturbances and constitute set-point tracking problems in which the operating point changes frequently.

[0075] MPC based on a non-linear model is a relatively recent technique and there are few references of industrial applications. However it is expected that it will become a promising short-term option due to its great potential. Problems associated with a non-linear MPC strategy are obtaining a model, deriving a solution to the problem, a loss of convexity and stability (because a global optimum is not guaranteed for many optimisers), and the fact that data processing associated with calculation increases significantly due to repeatedly evaluating the objective function, wherein each evaluation must solve the system of nonlinear equations that constitute the prediction model. The MPC algorithm can calculate the gradient of the function and the next points search, in addition to checking for any violation of restrictions and algorithm ending criteria. A nonlinear approach is nevertheless recommended in the example case of the water network 10, due to the nonlinear dynamics of the reservoirs and demand consumption.

[0076] Constraints have to be set by way of restrictions and algorithm ending criteria, for bounding the spectrum of optimising solutions. Accordingly, constraints are set for the optimization of the network 10 at step 604, including limits of maximum and minimum heights of operation in each reservoir, limits on time acting on the ON/OFF pumps and a limit on the maximum total throughput (m³/h) that can be extracted from the filtered water tank 13_d. [0077] In the example, the filtered water tank 13D does not have enough water resources (volume and inlet flowrate) to keep the 3 pumps 1 l_{a c} fully active at the same time, and there is thus a total maximum flow constraint, which must be above the sum of the flow rates of the 3 pumps 1 l_a._c at any given time. If the sum of the flow rates of the 3 pumps 11 _a._c exceeds that constraint, the filtered water tank begins to empty and could eventually be wholly emptied, a situation which cannot be allowed to happen. A constraint in the filtered water tank volume or water level could be used, but the inflow rate of the reservoir is not known; only the raw water inflow rate, so the most restrictive constraint is used, being the maximum inlet flowrate into the filtered water reservoir, at 1900m³/h.

[0078] The MPC module 16 next receives monitored plant and nodal data from the remote network controlling terminal 100 at step 605, and plant-independent meteorological data 17 from the remote meteorological data terminal 102 at step 606. [0079] The MPC module 16 next processes the model of step 602, demand forecast of step 603, constraints of step 604, captured network data of step 605 and meteorological data of step 606 with an MPC algorithm with a receding horizon of, in the example, 24 hours, to calculate a sequence of control signals for the pumps 1 I_A-_C, wherein the sequence effectively minimises an objective function along that horizon. A 24 hours receding horizon has been chosen as corresponding to a period representative of a reasonably accurate weather prediction, advantageously corresponding also to a period representative of a reasonably accurate prediction for energy costs, if market data should be added to weather data.

[0080] Configuring the MPC controller 16 with a receding horizon of 24 hours and the constraints of step 604 causes a problem of optimisation of output variables. In the example, the plant operates at two fixed rates during the day and night respectively. The aim is, as previously stated, to reduce the carbon dioxide emissions associated with nodal powering, i.e. pump activation in the network 10 of the example, and accessorily also to reduce the cost of operation based on knowledge of the MW/h cost expectations offered by the utility provider, by optimization. [0081] Since two of the pumps are OFF/ON, a Mixed-Logic-Integer Optimization problem has to be solved, and an optimisation algorithms such as real-interior points cannot be used. The use of optimisation algorithms that allow the restriction of integers variables is necessary. For the low reservoir pump l l_a, which is VSD, an algorithm of real numbers can be used, since it has a wide operating range. However, the three systems are coupled, in principle, by the total maximum flow that can be extracted from the filtered water reservoir 13 d-

[0082] Optimisation is thus done using the coupled model and a genetic algorithm, since it will predict the model with both ON/OFF pumps. The genetic algorithm is a method for solving both constrained and unconstrained optimization problems that is based on natural selection: the genetic algorithm repeatedly modifies a population of individual solutions. At each step, the genetic algorithm selects individuals at random from the current population to be parents and uses them to produce the children for the next generation. The genetic algorithm uses three main types of rules at each step to create the next generation from the current population: selection rules select the individuals, called parents that contribute to the population at the next generation; crossover rules combine two parents to form children for the next generation; and mutation rules apply random changes to individual parents to form children. In that context, an individual represents the set of actions, modelled as integers (e.g. ON as 1 and OFF as 0), for every OFF/ON pump 11C,_B and a value between 0 and 100 for the VSD pump 11_A- The fittest or best individual is selected based on a fitness function.

[0083] A skilled person will readily understand that the parameters of the genetic function can be adjusted in order to get a compromise between a better accuracy to reach the global optimum, and computation time. The output of the function, and therefore of the calculation of step 607, is a complete sequence of actions for pumps 1 l_a-c, consisting variously of flowrate level(s) expressed in m³ per hour for the VSD pump 11_A and start and end activation times for each of the ON/OFF pumps 11C,_B, output as the optimising schedule at step 608 The MPC module 16 subsequently invokes the communications module 18 to communicate the optimising schedule to the network-controlling terminal 100 at step 609, whereby the network-controlling terminal 100 may then control the pumps 1 I_A-C in the network 10 according to the received schedule. Logic control at the system terminal 101 returns to step 605, wherein the MPC may then process an updated schedule at a next iteration of step 607, based on updated monitored network data and updated meteorological data 17 received at next iterations of steps 605, 606, and so on and so forth.

[0084] The contents of the memory means 209 of the system controller terminal 101 at runtime, when steps 602 to 609 are processed by the CPU 208, are illustrated in Figure 6, wherein an operating system is shown at 701 which, if the system controller terminal 101 is for instance an desktop computing device as described in Figures 3 and 4 and manufactured e.g. by DELL® Inc. of Round Rock, Texas, USA, is Windows Server 2012 R2 distributed by Microsoft® Inc. of Redmond, Washington, USA. The OS 701 includes instructions for governing the basic data processing, interdependence and interoperability of the server hardware components e.g. as described in Figure 5, and communication subroutines 702 to configure the system controller terminal 101 for bilateral network communication via the NIC 211 interfaced with the wired connection 110 to the local router 111. The OS 701 also includes input subroutines for reading and processing input data variously consisting of user direct input to human interface devices, namely the keyboard 203 and computer mouse 204.

[0085] The data processing terminal 101 is configured into the system controller terminal 101 of the system by a set of instructions 703 embodying the MPC 16 and its associated modules 12, 14, 18, and residing in the memory 209 at runtime. The set of instructions 703 is interfaced with the OS 601 through one or more Application Programmer Interfaces (API) 704 and comprises and coordinates the data processing activity of further function-specific data processing subroutines, and a user interface 705 updated and output to the display 202 in real-time. [0086] The modeller module 12 is shown as a first subroutine 706, the demand prediction module 14 is shown as a second subroutine 707 and the communications module 18 is shown as a third subroutine 708.

[0087] The modeller module 12 maintains a list 709 of nodes 11 _A-C in the network 10. Data received from the remote node-controlling terminal 100, or input by the user of terminal 101, and processed by the modeller module 12 at step 602, 603 includes nodal parameter data 710 for each listed node, inclusive of e.g. pump type (VSD or ON/OFF), rated power (kWh), inlet and outlet valve aperture dimensions; nodal constraints data 711 for each listed node, inclusive of e.g. minimum and maximum flowrates, maximum safe operation time; network parameter data 712, inclusive of e.g. reservoir types, dimensions, elevations, inlet and outlet valve aperture dimensions; and network constraints data 713, inclusive of e.g. minimum and maximum reservoir levels, minimum and maximum flowrates.

[0088] Data received from both the remote node-controlling terminal 100 and the remote meteorological terminal 102, and processed by the demand prediction module 14 and the MPC 16 at steps 604 to 608 additionally includes monitored network data 714, inclusive of e.g. current water level, inflow rate and outflow rate for each reservoir, pump activity status and flowrate for VSD pumps 11_A; meteorological data 17 shown at 715 inclusive of e.g. current and forecast solar irradiation level, air temperature, rainfall level; and the optimising schedule output at each step 608 and shown at 716. [0089] An important embodiment of the system shown and described with reference to Figures 1 to 7 is shown in Figures 8 and 9, wherein like reference numerals relate to like features and data processing steps. In this embodiment, the task imparted to the MPC 16 of minimising carbon dioxide generation is combined with a task of reducing the operational cost of the network 10, expressed as the total cost of the predicted power grid energy supplied to the nodes 1 I_A-C for operating same according to the optimising schedule 716 over time.

[0090] In this embodiment therefore, grid energy cost data 19 is supplied to the MPC 16 by the utility- operated power cost monitoring terminal 103 at new step 801, after and in addition to the monitored network data supplied by the node-controlling terminal 100 at step 605 and the meteorological data 17 supplied by the meteorological source 102 at step 606. In this embodiment also, the contents of the memory means 209 of the system controller terminal 101 further includes the energy cost data feed 19 received from the power cost monitoring terminal 103, and the function minimised at steps 604 and 607 is a cost function.

[0091] With reference to Figures 10 and 11 now, applying the methodology of steps 602-609 to the example water network of Figure 3, the constraints being taken into account in the resolution at step 604 are:

3m < h_LR < 8.85m

1.5m < h_1R < 4.55m

2.8m < h_HR < 9.95m

iR + 1900 m³/h wherein ILR is the water level in the low reservoir 13_a, 1¾R is the water level in the intermediate reservoir 13b, IHR is the water level in the high reservoir 13_c, qin is the flow rate of the VSD pump 1 l_a associated with the low reservoir 13_a, qm is the flow rate of the ON/OFF pump l i_b associated with the intermediate reservoir 13b and (¾HR is the flow rate of the ON/OFF pump 1 l_c associated with the high reservoir 13_c.

[0092] For the sake of operational safety, the minimum height of the reservoirs 13_{a c} is increased by an offset to optimize the system remotely from the operational limits, updating the constraints as:

5.3m < h_LR < 8.85m

3.5m < h_IR < 4.55m

[0093] The parameters imparted to the MPC 16 at step 604 are a sample time of 1 hour, a receding horizon of 24 hours, a network output (water demand) profile for a week day and a weekend day and power costs. Figures 10 and 1 1 show the results 608 of the predictions and optimisation of the pumping rates for the real plant of the example over a period of 30 days. The results are shown with, and contrasted by, real flowrates of the plant for the same period of time, based on 2014 data for the real water network 10.

[0094] With reference to Figures 8 and 9, when applying the methodology of steps 602-609 and 801 to the example water network of Figure 3 on the above basis, a prediction over a period of 30 days yielded an energy cost saving of 14.45% and the same prediction over a period of 60 days yielded an energy cost saving of 17.48%.

[0095] With reference to Figures 12 to 14 now, the inventors elaborated a physical laboratory test rig of the water plant of the example water treatment network shown and described with reference to Figure 3, as an approximated scaled system thereof. The main objective with this test rig was to prove that the optimisation procedure using a Model Predictive Control Strategy is capable of maintaining the system working within system constraints (maximum and minimum levels of the reservoirs 13A-D). [0096] This test was run in a test rig predicting the water network 10 layout illustrated in Figure 3. The test rig was formed of a low reservoir 13A, an intermediate reservoir 13B, a high reservoir 13c, a tower reservoir 13D extracting water from the high reservoir 13c, pumps 1 1A-C, valves, programmable logic controllers ('PLC') and an Open Platform Communications (OPC) computer. [0097] As in the water network 10 on the basis of which predictions were run, the tower reservoir pumping system drains water from the high reservoir 13c and the control algorithm is the same as for the water network 10. The tower reservoir 13D pumps water with an ON-OFF pump until it reaches a maximum operation level, then the pump stops until the tower reservoir reaches a minimum operation level, at which the pump starts again. This control algorithm is taken into account in the optimization, but it is not part of the MPC control algorithm to calculate the control actions for the pumps 1 IA-C of the low, intermediate and high reservoirs 13A-C- [0098] The constraints of the reservoirs (plant operation) are:

0.35m < h_LR < 0.7m

0.35m < h_IR < 0.7m

0.4m < h_HR < 1.2m

0.4m < h_TR < 1.2m

[0099] In order both to run the test for 3 days (72 hours) of simulation and to take the small dimensions of the test rig reservoirs into account, the time was compressed. As the intention was to replicate the system 10 of the water network, a 1 hour sample time was used, but time compressed to 6 minutes, so that 24 hours would correspond to 144 minutes (2.5 hours). This way, the only difference was the running time of the test rig.

[00100] Figures 12, 13 and 14 respectively relate to the low reservoir 13_a, the intermediate reservoir 13_b and the high reservoir 13_c of the test rig. In the figures, the respective levels of the reservoirs are shown, with the horizontal axis representing the hours (3 days equivalent to 72 hours), proving that the control algorithm is working as expected. It is keeping the level between the maximum and minimum safe operational points. In the figures, the respective supply demands and pumping rates of the reservoirs are shown, with the horizontal axis representing the minutes (3 days, 72 hours, 432 minutes). The graphs clearly show that the optimisation system with the MPC control algorithm is capable of avoiding some of the high peaks of cost of the SEM Prices, when possible, because the minimum safe operational points of the reservoirs are hard constraints and have to be accomplished.

[00101] In another embodiment of the invention the system may comprise at least one battery in the network that can be charged at optimum charging times. The computed optimised schedule can take account of times when a charged battery can be used to power one or more of the power nodes in the network when it is more efficient than powering from a main grid supply.

[00102] It will be appreciated that the node hereinbefore described can relate to any power generating equipment such as pumps, valves and the like. It will be appreciated the invention can be used in any system for optimising energy usage across the various electrically-powered nodes of a distributed processing architecture, such as a manufacturing plant, a building or the like.

[00103] In the specification the terms "comprise, comprises, comprised and comprising" or any variation thereof and the terms include, includes, included and including" or any variation thereof are considered to be totally interchangeable and they should all be afforded the widest possible interpretation and vice versa.

[00104] The invention is not limited to the embodiments hereinbefore described but may be varied in both construction and detail.

Claims

Claims 1. A system for optimising scheduling of power consumption in a network,

wherein the network (10) comprises at least one node (11_A) powered by a power source and controllable by at least one data processing terminal (100);

characterised in that

the system comprises a plurality of networked data processing terminals including the at least one node controlling terminal (100) and at least one system controller terminal (101); and

the at least one system controller terminal (101) is configured with:

modelling means (12) for modelling power consumption of the or each node (H_A-_C) according to characteristics of the or each node, input constraints of the network and output constraints of the network; forecasting means (14) for forecasting power consumption of the or each node (1 I_A-_C) according to the modelled power consumption and/or meteorological data (17) received from a meteorological data source (102) as a first data feed and/or a second data feed, including at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes 1 I_A-_C;

scheduling means (16) for computing a schedule (716) for the powering of the or each node (11_A- c) according to the forecast power consumption and/or the meteorological data and/or the second data feed with a receding horizon function; and

communication means (18) for commanding the or each node controlling terminal (100) to control powering of the or each node (1 I_A-_C) according to the optimised schedule (716).

2. The system according to claim 1, wherein the modelling means (12) comprises a modelling module, the forecasting means (14) comprises a demand prediction module operable to provide the second data feed and the scheduling means (16) comprises a model predictive controller operably interfaced with the modelling module, the demand prediction module and the communications means (18).

3. The system according to claim 1 or 2, wherein the meteorological data (17) is supplied by the meteorological data source (102) to the scheduling means (16) and/or modelling means (12).

4. The system according to any of claims 1 to 3, wherein the meteorological data (17) comprises one or more selected from wind speed, rainfall level and solar irradiation in a predetermined time period.

5. The system according to any of claims 1 to 3, wherein the meteorological data (17) comprises one or more selected from wind speed, rainfall level and solar irradiation monitored by the meteorological data source (102) in real-time.

6. The system according to any of claims 1 to 5, wherein the plurality of networked data processing terminals further includes a power cost monitoring terminal (103), and wherein the system controller terminal is further adapted to compute the schedule optimising the powering of the or each node according to the forecast power consumption, the meteorological data (17) received from the meteorological terminal and power cost data (19) received from the power cost monitoring terminal (103).

7. The system according to claim 6, wherein the power cost data (19) is supplied by the power cost monitoring terminal (103) to the scheduling means (16).

8. The system according to any of claims 1 to 7, wherein the network is a water network comprising one or more water tank(s) (13_A-C) and the or each node is a water pump (1 I_A-C) associated with the one or more tank(s) or valve or other piece of equipment.

9. A method for optimising scheduling of power consumption in a network (10),

wherein the network comprises at least one node (11_A) powered by a power source and controllable by at least one data processing terminal (100), and

wherein the at least one data processing terminal ( 100) is controllable by at least one remote system controller terminal (101);

the method comprising the steps of, at the at least one remote system controller terminal (101): modelling power consumption of the or each node (H_A-_C) according to characteristics of the or each node, input constraints of the network and output constraints of the network;

forecasting power consumption of the or each node (H_A-_C) according to the modelled power consumption and/or real-time or predicted meteorological data received from a meteorological data source (102) as a first data feed and/or a second data feed, including at least power supply grid data from one or more utility suppliers that is representative of a historical, real-time and/or forecast unit cost of grid energy for a period of time supplied to the nodes 1 I_A-C;

computing a schedule (716) for the powering of the or each node (1 I_A-_C) according to the forecast power consumption and the real-time or predicted meteorological data and/or the second data feed with a receding horizon function; and

commanding the or each node controlling terminal (100) to control powering of the or each node (1 I_A-C) according to the optimised schedule (716).

10. The method according to claim 9, comprising the further step of receiving monitored network data at the at least one remote system controller terminal (101) from the node-controlling data processing terminal (100).

11. The method according to claim 10, wherein the step of forecasting further comprises sampling the monitored network data at a predetermined time interval and limiting the receding horizon with the or another predetermined time period.

12. The method according to any of claims 9 to 11, wherein the meteorological data (17) comprises one or more selected from wind speed, rainfall level and solar irradiation in a predetermined time period.

13. The method according to any of claims 9 to 12, wherein the meteorological data (17) comprises one or more selected from wind speed, rainfall level and solar irradiation monitored by the meteorological data source (102) in real-time.

14. A computer program product downloadable from a communication network (104, 112) and/or stored on a computer-readable and/or microprocessor-executable medium (206, 207), characterized in that it comprises program code instructions (703) for implementing a method for optimising power consumption in a network (10) according to any one of Claims 9 to 13.

15. A network comprising at least one node (11A) powered by a power source and controllable by at least one data processing terminal (100), characterised in that the data processing terminal (100) is controlled by at least one remote system controller terminal (101) associated with a meteorological data source (102) and configured by the computer program product of claim 14.