CN116739292A

CN116739292A - Energy optimization scheduling method, system and storage medium of data center

Info

Publication number: CN116739292A
Application number: CN202310751668.9A
Authority: CN
Inventors: 于艳华; 刘雅熙; 李强; 刘迪; 郭少勇; 王婧; 李温静
Original assignee: State Grid Information and Telecommunication Co Ltd; Beijing University of Posts and Telecommunications
Current assignee: State Grid Information and Telecommunication Co Ltd; Beijing University of Posts and Telecommunications
Priority date: 2023-06-25
Filing date: 2023-06-25
Publication date: 2023-09-12

Abstract

The invention provides an energy optimization scheduling method, system and storage medium of a data center, wherein the method comprises the following steps: outputting solar predicted photovoltaic power generation output power based on the numerical weather prediction data and the pre-trained BiLSTM prediction network model; determining an energy requirement of the data center based on an amount of power consumed by the data center server; storing electric quantity to a data center energy storage device or acquiring electric quantity from a smart grid; determining average response times of various types of clients based on the queuing network model; obtaining an optimal energy demand configuration of the data center based on an energy dynamic optimization scheduling model aiming at minimizing the operation cost; and obtaining a daily time sequence prediction result of each decision variable in the energy dynamic optimization scheduling model based on an exponential smoothing method, and obtaining the optimal energy demand configuration in the data center day based on the daily time sequence prediction result. The invention can reasonably schedule resources and manage energy of the data center, and ensure the economical efficiency of operation.

Description

Energy optimization scheduling method, system and storage medium of data center

Technical Field

The invention relates to the technical field of energy management, in particular to an energy optimization scheduling method and system for a data center.

Background

In order to adapt to new service requirements, the data center continuously improves the performance of the server, meanwhile, the heat generation of the data center is increased, the power of various devices is in an ascending trend, and the energy crisis problem and the greenhouse gas emission problem are also highlighted. At present, renewable energy sources such as photovoltaics become a main direction of future global energy development, and mainly reduce carbon emission to strengthen environmental protection, reduce energy consumption cost to improve economic benefit, get rid of excessive dependence of fossil energy sources on cooling facilities to improve stability dependence of a data center system and the like.

The invention provides a Chinese patent with application number 202011102939.0 and the name of a comprehensive energy system optimization scheduling method of integrating distributed data central computing power and energy flow, which provides a comprehensive energy system optimization scheduling method of integrating distributed data central computing power and energy flow, and the method comprises the following steps: establishing a comprehensive energy system mathematical model of the integration of the distributed data center calculation force and energy flow based on graph theory; establishing a comprehensive energy system operation evaluation index system comprising three primary indexes of economy, safety and cleanliness and a plurality of secondary indexes; determining the comprehensive weight of each index by adopting a comprehensive evaluation method; constructing an optimized scheduling model of the comprehensive energy system by taking the lowest running cost, the highest safety and the lowest pollution emission as three objective functions; and accessing the comprehensive energy system optimal scheduling model into a mathematical model thereof to obtain an optimal scheduling method and an optimal scheduling result. The distributed data center is incorporated into the comprehensive energy system, and other energy sources such as calculation power, electric power and heat are comprehensively and optimally scheduled, so that the energy waste is reduced, the consumption of clean energy is increased, and the flexibility of the system is improved. However, the patent does not consider the characteristics of intermittence, randomness, mutation and the like of clean energy, which may lead to poor economic benefit, and focuses on the optimal scheduling of an energy supply system, and the management of an energy consumption system of a data center is lacking.

The invention discloses a data center real-time energy management method and a data center real-time energy management system, which belong to the field of electric energy management, and are disclosed in China patent application with the application number of 202110905770.0 and the invention name of 'a data center real-time energy management method and a data center real-time energy management system', wherein the energy management method comprises the following steps: establishing a real-time energy management model of the data center in a current preset period, and reconstructing the model into a Markov decision process; optimizing an optimal real-time energy management strategy of the data center by solving the Belman equation time by time, respectively performing value function approximation by adopting three-dimensional state variables of the batch load quantity, the electricity storage quantity and the cold storage quantity in a queue, overcoming the problem of difficulty in solving the value function, combining the batch load quantity approximation function, the electricity storage quantity approximation function and the cold storage quantity approximation function in the queue obtained by offline training in advance to obtain a three-dimensional approximation function, substituting the three-dimensional approximation function into the Belman equation in a Markov decision process, and solving the Belman equation to obtain an approximate global optimal decision variable set for managing the data center. The patent application may provide for real-time energy management of data centers operating in an uncertain environment. The problem with this patent application is that the characteristics of clean energy intermittence, randomness, variability, etc. in the energy system are still not considered, which may lead to poor service stability and the economy of the data center is only considered in terms of energy consumption costs.

Chinese patent application No. 202080003421.3 entitled "data center energy management system" discloses a technique that includes managing energy flow within a system, which describes a system that includes: a power generation system; a battery storage system having a state of charge attribute; and processing circuitry capable of accessing the power grid, the power generation system, and the battery storage system. In one example, the processing circuit is configured to: determining a prediction of energy utilization of the data center; monitoring energy availability factors; determining an energy flow configuration defining an energy flow involving the power grid, the power generation system, the battery storage system, and the data center based on the energy availability factors of the energy utilization predictions and monitoring, wherein the energy flow configuration includes information identifying one or more of the power grid, the power generation system, or the battery storage system as a source of power for the data center; powering a data center based on the energy flow configuration; and managing energy flow involving the battery storage system based on the energy flow configuration. This patent application considers various factors that affect the energy generation system and system management based on energy flow within the system, but does not consider factors that affect the cost effectiveness of the data center.

The data center is an energy system with a large amount of new energy access, is also a huge physical information system, and is a key point that the data center can achieve minimization of operation cost and safety and stability of operation process under the premise of guaranteeing operation economy of the data center, reasonably meets calculation task planning and arrangement and effectively manages energy consumption of the data center, is crucial for operation of the whole data center, and reasonably solves resource scheduling and energy management.

Disclosure of Invention

In view of the above, the embodiment of the invention provides an energy optimization scheduling method for a data center, so as to solve the problems of resource scheduling and energy management of an energy supply side and an energy demand side of the data center.

One aspect of the present invention provides a method for energy-optimized scheduling of a data center, the method comprising the steps of:

acquiring numerical weather prediction data, inputting the acquired numerical weather prediction data into a pre-trained BiLSTM prediction network model, and outputting a day-ahead prediction result of photovoltaic power generation output power;

determining an energy requirement of the data center based on the power consumed by the data center server for starting, the power consumed by the work and the power consumed by the cooling;

Determining the electric quantity stored in an energy storage device of a data center or the electric quantity required to be acquired from a smart grid based on the solar-day prediction result of the photovoltaic power generation output power and the energy demand of the data center;

determining average response times of various types of clients based on the queuing network model;

obtaining an optimal energy demand configuration of the data center based on an energy dynamic optimization scheduling model aiming at minimizing the operation cost; wherein the decision variables in the energy dynamic optimization scheduling model include at least one of the following variables: the arrival rate of the client request, the real-time electricity price of the smart grid and the photovoltaic power generation output power; the operating cost is obtained based on the following factors: the electric quantity exchanged by the data center and the intelligent power grid, the operation and maintenance price of the energy storage device and the average response time of various types of clients;

and obtaining a daily time sequence prediction result of each decision variable in the energy dynamic optimization scheduling model based on an exponential smoothing method, obtaining the optimal energy demand configuration in the data center day based on the daily time sequence prediction result, and progressively updating the energy dynamic optimization scheduling model according to the time sequence.

In some embodiments of the present invention, before the pre-training the BiLSTM predictive network model, the method further comprises:

And (5) processing the original photovoltaic power generation output power and the numerical weather prediction data by using standard fraction normalization.

In some embodiments of the invention, the amount of power consumed by cooling is determined based on an outside air cooling method.

In some embodiments of the present invention, the determining, based on the future prediction result of the photovoltaic power generation output power and the energy requirement of the data center, the amount of electricity stored in the data center energy storage device or the amount of electricity required to be obtained from the smart grid includes:

if the day-ahead prediction result of the photovoltaic power generation output power is smaller than the energy requirement of the data center, the data center acquires electric quantity from the intelligent power grid, and the energy requirement of the data center is equal to the sum of the day-ahead prediction result of the photovoltaic power generation output power and the electric quantity acquired by the intelligent power grid; if the day-ahead prediction result of the photovoltaic power generation output power is larger than the energy demand of the data center, the energy storage device of the data center stores redundant electric quantity, and the energy demand of the data center is equal to the difference between the day-ahead prediction result of the photovoltaic power generation output power and the electric quantity stored by the energy storage device.

In some embodiments of the invention, the queuing network model is built based on an exponential distribution of customer request arrival times and service times.

In some embodiments of the invention, the arrival rate of the client request is an effective arrival rate of the site of the client request determined based on the routing matrix.

In some embodiments of the invention, the average response time is determined based on litter's law.

In some embodiments of the invention, the exponential smoothing method comprises a seasonal processing scheme in Holt-windows three-parameter exponential smoothing.

Another aspect of the invention provides an energy-optimized data-center scheduling system comprising a processor and a memory, said memory having stored therein computer instructions for executing the computer instructions stored in said memory, which system, when executed by the processor, implements the steps of the method of any of the embodiments described above.

Another aspect of the invention provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method of any of the embodiments described above.

According to the energy optimization scheduling method and system for the data center, the energy management problem of multi-channel coordination of the data center is studied by combining the IT equipment such as renewable photovoltaic energy sources at the energy supply side, the energy storage device, the intelligent power grid, the energy demand side server and the like and cooling equipment. The invention has the advantages that the energy application scene studied is clean and needs multipath coordination, and accords with the current sustainable low-carbon complex data center environment; secondly, the future multi-step prediction method based on the BiLSTM time sequence predicts the photovoltaic power generation output power, so that the accuracy of the future prediction result of the photovoltaic power generation output power can be improved; in addition, an exponential smoothing method is introduced to predict each decision variable in an energy dynamic optimization scheduling model aiming at minimizing the operation cost in a daily way, the optimal energy consumption configuration of the data center is continuously predicted, optimized and updated according to the actual change, the output condition of multi-path equipment in the system is coordinated, the energy consumption of the data center can be effectively managed on the premise of meeting the calculation task planning and arrangement, the resource scheduling and the energy management are reasonably solved, the operation cost minimization of the data center is realized, and the safe and stable operation of the data center is ensured.

Additional advantages, objects, and features of the application will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and drawings.

It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present application are not limited to the above-described specific ones, and that the above and other objects that can be achieved with the present application will be more clearly understood from the following detailed description.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate and together with the description serve to explain the application.

Fig. 1 is a flow chart of an energy-optimized scheduling method of a data center according to an embodiment of the application.

FIG. 2 is a bar graph of the number of servers assigned to a network group, an application group, and a database group during a day in accordance with one embodiment of the present application.

FIG. 3 is a bar graph of the number of servers that are turned on and off during a day in accordance with one embodiment of the present application.

FIG. 4 is a graph of a comparison of predicted and actual values of renewable energy production during a day in accordance with one embodiment of the present application.

Fig. 5 is a line graph of the data center obtaining power from the smart grid during one day in an embodiment of the present invention.

FIG. 6 is a plot of renewable energy production and sales volume over a day in accordance with an embodiment of the present invention.

Fig. 7 is a line graph of the energy sold and used by the energy storage device during a day in an embodiment of the present invention.

FIG. 8 is a comparison plot of predicted and actual values of sales prices for renewable energy during a day in accordance with an embodiment of the present invention.

Fig. 9 is a comparison line graph of daily forecast values and true values of grid prices within a day in an embodiment of the invention.

FIG. 10 is a comparison line graph of the operation cost of the data center obtained by the simulation method and the optimization scheme according to the present invention in an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the following embodiments and the accompanying drawings, in order to make the objects, technical solutions and advantages of the present invention more apparent. The exemplary embodiments of the present invention and the descriptions thereof are used herein to explain the present invention, but are not intended to limit the invention.

It should be noted here that, in order to avoid obscuring the present invention due to unnecessary details, only structures and/or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details not greatly related to the present invention are omitted.

It should be emphasized that the term "comprises/comprising" when used herein is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

The data center is an energy system with a large amount of new energy access, is also a huge physical information system, and is a key point that whether the data center can reasonably and effectively manage the energy consumption of the data center on the premise of meeting the calculation task planning arrangement is critical to the operation of the whole data center, and the reasonable resource scheduling and the energy management are the data center to achieve the minimization of the operation cost and ensure the safety and stability of the operation process.

Based on the data center energy management optimization scheduling method, the data center energy management optimization scheduling method comprises the daily front and daily inner prediction. Firstly, considering the intermittent characteristic of clean energy, improving the accuracy of future multi-step energy prediction based on BiLSTM time sequence; in addition, the energy supply of the interaction of the energy storage device and the intelligent power grid is considered, the server meets the factors of multi-application service requests, energy consumption of cooling and the like, and an energy dynamic optimization scheduling model is built with the aim of minimizing the operation cost; and an exponential smoothing method is introduced to predict all decision variables in the model in a daily way, the optimal energy consumption configuration of the data center is continuously predicted, optimized and updated according to actual changes, the output conditions of multiple devices in the system are coordinated, the running cost of the data center is reduced, and adverse effects of high prediction difficulty and equipment power failure caused by uncertain factors of weather, electric power markets and user behaviors are reduced.

Fig. 1 is a flow chart of an energy-optimized scheduling method of a data center according to an embodiment of the invention. As shown in FIG. 1, the method may specifically include steps S110-S160.

Step S110: and acquiring numerical weather prediction data, inputting the acquired numerical weather prediction data into a pre-trained BiLSTM prediction network model, and outputting a day-ahead prediction result of the photovoltaic power generation output power.

The photovoltaic power generation has the characteristics of scattered physical positions, small production scale, intermittent power output, randomness, mutation and the like, so that the output power of the photovoltaic power generation is often influenced by multiple factors such as time, weather and the like, and the influence caused by other meteorological factors is ignored, so that prediction errors are very easy to generate. In addition, the accuracy of the photovoltaic power generation output power prediction increases the cost of the photovoltaic power generation output power prediction.

Numerical weather prediction (NWP, numerical Weather Prediction) refers to a method of predicting the atmospheric motion state and weather phenomenon for a certain period of time in the future using a mathematical model of the atmosphere and the current weather conditions as input data, and is generally performed using a supercomputer or a distributed computing cluster to perform calculations, including meteorological information such as irradiance, wind speed, wind direction, and the like.

More specifically, the pre-trained BiLSTM predictive network model can predict the response of data and update the network state, and then predict the photovoltaic power output power for a certain period of time in the future according to the set sliding window range.

Further, long and short term memory (LSTM, long Short Term Memory) networks have broader applications in future multi-step predictions. The LSTM model controls discarding or adding information through mechanisms such as forgetting gate, input gate and output gate, effectively solves the problem of long-term dependence, and is suitable for classification, processing and prediction based on time sequence data. However, the LSTM model can only predict the time sequence information of the previous time unit to obtain the information of the next time unit, and cannot encode the information from the back to the front, and the two-way long and short-term memory (BiLSTM, bi-directional Long Short Term Memory) network is composed of a forward LSTM and a backward LSTM, so that the information of the context data can be fully learned.

More specifically, the core idea of the LSTM model is as follows:

1) The forget gate f may be used to determine the information discarded from the state of the cell for controlling the extent to which the last cell was forgotten. Input x of current cell _t And the output h of the last cell _t-1 The information is selectively filtered by a sigmoid function. The sigmoid function has an output value of [0,1 ]In between, information can be selectively passed, 0 indicating complete discard, and 1 indicating complete pass.

f _t ＝Sigmoid(W _f [h _t-1 ,x _t ]+b _f )；

Wherein f _t Representing the output of the current forget gate, W _f And b _f Respectively representing the weight and bias of the forgetting gate.

2) Input deviceThe gate and tanh functions determine the portion of new information that is added in the cell state. The sigmoid function may be used to determine the retained input information i _t And x is _t And h _t-1 Part of the candidate value C 'is updated by the tanh function to generate a new candidate value C' _t . Binding i _t And C' _t To update the old cell state C _t-1 Obtaining a new cell state C _t 。

i _t ＝Sigmoid(W _i [h _t-1 ,x _t ]+b _i ]；

C _t '＝tanh(W _C [h _t-1 ,x _t ]+b _C )；

C _t ＝f _t C _t-1 +i _t ·C' _t ；

In which W is _i And b _i Respectively represent the weight and bias of the input gate, W _c And b _c The weights and biases of the tanh layers are shown, respectively.

3) The output gate is used to control the extent to which the current cell state is filtered. First x _t And h _t-1 Determining output part O of cell state by sigmoid function _t Setting the cell state to be tanh and combining with O _t Multiplying to obtain the predicted value h of the LSTM model _t 。

O _t ＝Sigmoid(W ₀ [h _t-1 ,x _t ]+b ₀ )；

h _t ＝O _t ·tanh(C _t )；

Wherein O is _t Representing the output of the current output gate, W ₀ And b ₀ Respectively representing the weight and bias of the output gate.

The single-layer BiLSTM model consists of one forward propagating layer (forward LSTM layer) and one backward propagating layer (backward LSTM layer), and the forward and backward propagating layers are commonly connected to the output layer. The input sequence is propagated in the forward direction from t ₁ To t ₂ Calculating forward time for one time to obtain and store the output of each time forward hidden layer; the input sequence is at the counter-propagating layer along time t ₂ To t ₁ Calculating time reversely to obtainAnd storing the output of the backward hidden layer at each moment; finally at [ t ₁ ，t ₂ ]And combining the output results of the forward propagation layer and the backward propagation layer at the corresponding moments in time to obtain a final output. The BiLSTM model can better capture the bi-directional semantic dependence, so that the photovoltaic power generation output power can be predicted more accurately in future and in a multi-step day-ahead mode.

As an example, specific building steps of the pre-trained BiLSTM predictive network model may include:

1) And (5) constructing a BiLSTM prediction network model. The BiLSTM network model comprises a sequence input layer, a BiLSTM layer, a full connection layer and an output layer, and relevant training parameters such as network weight, learning rate, iteration number and the like are configured.

2) Training of BiLSTM predictive network models.

Normalized training data set containing K sets of dataInputting the BiLSTM prediction network model constructed by the steps, wherein x is _k Is a self-variable data set, and contains related factors influencing the output power of photovoltaic power generation, such as NWP data; y is _k Is a dependent variable data set, namely a data set containing K photovoltaic power generation output powers. And after the training set and the test set sample are divided, performing repeated iterative training and correction to obtain a corresponding photovoltaic power generation prediction model, namely a pre-trained BiLSTM prediction network model.

The NWP data is used as an argument to predict the photovoltaic power generation output power, but the present invention is not limited thereto, and other factors affecting the photovoltaic power generation output power such as a solar cell module, dust loss, etc. may be used.

In some embodiments of the present invention, prior to pre-training the BiLSTM predictive network model, further comprising: and (5) processing the original photovoltaic power generation output power and the numerical weather prediction data by using standard fraction normalization.

More specifically, to enable data to have the same metrics while the neural network is able to converge quickly, a standard score (Z-score) method may be used to normalize the raw NWP data and photovoltaic power generation power that are processed input. The formula for the Z-score normalization method can be expressed as:

wherein x is _i Ith data, x 'in data samples representing raw NWP data and photovoltaic power generation' _i Representing the ith data in the normalized data sample, mean (x) represents the mean of the overall data sample, std (x) represents the standard deviation of the overall data sample.

Further, the photovoltaic power generation output power needs to be defined as follows:

0≤P _s (t)≤P _smax ；

wherein t.epsilon. {0,1,.. The. T represents a predicted time range (which may be divided into T-1 time intervals), P _s (t) represents the generated output power of the photovoltaic power plant in the t-th time interval, P _smax Representing the maximum generated output power of the photovoltaic power plant in the data center.

The following describes the accuracy of the future multi-step prediction of the photovoltaic power generation output power by the BiLSTM, by way of example, and includes:

in the day-ahead photovoltaic prediction stage, simulation comparison tests are performed based on BiLSTM single-step prediction, LSTM single-step prediction, BP prediction and BiLSTM multi-step prediction methods, and the prediction evaluation index results of the four simulation comparison methods are shown in Table 1. Data from 1 month in 2017 to 8 months in 2018 (data collected every 15 minutes for a total of 65760 time series related data) were selected to predict photovoltaic power generation power data from 10 months to 12 months in 2018. By analyzing the day-ahead prediction result of renewable photovoltaic power generation output power, a conclusion that the BiLSTM multi-step prediction method can obtain higher prediction precision can be obtained.

The training process between the neural network models is mostly similar, and is different in design of network structure and adjustment of parameters. The LSTM single-step prediction, biLSTM single-step prediction and BP prediction method is different from the BiLSTM multi-step prediction method in that: setting a predicted window as 1 in the BiLSTM single-step prediction model; setting a predicted window as 1 in an LSTM single-step prediction model, and replacing a BiLSTM layer in a neural network model with an LSTM layer; the neural network of the BP prediction model is designed to be an input layer, a hidden layer and an output layer. All four neural network models need four steps of preprocessing data, constructing models, training and predicting.

In addition, the data used in the prediction process comprises desensitized environmental data (namely environmental data obtained by removing data with larger deviation), actual irradiance of an electric field and electric field generating power, and the data fields comprise time, irradiance, wind speed, wind direction, temperature, humidity, pressure and actual power.

According to the invention, two conventional prediction evaluation indexes of average absolute error (MAE, mean Absolute Error) and mean square error (RMSE, root Mean Square Error) are selected, wherein the two conventional prediction evaluation indexes range from [0, + ], and when the predicted value of the photovoltaic power generation output power is completely matched with the true value, the predicted value is equal to 0, namely a perfect model, and the larger the error is, the larger the values of the MAE and the RMSE are. The evaluation index results of the four prediction methods are shown in the table, so that the MAE and RSME effects of the BiLSTM multi-step prediction are minimum in the four methods, and the difference between the photovoltaic power generation value predicted by the method and the real power generation value is minimum, namely the BiLSTM multi-step prediction method is optimal in the four photovoltaic power generation output power prediction methods, and the prediction accuracy is higher.

Table 1 predictive evaluation index results for four methods

Index\prediction method	BiLSTM multistep	BiLSTM single step	LSTM single step	BP
					MAE	1.1221	1.4216	1.7633	3.0305
RSME	2.0010	2.3520	2.7385	4.8421

Step S120: the energy requirements of the data center are determined based on the amount of power consumed by the data center server to turn on, the amount of power consumed by the work, and the amount of power consumed by the cooling.

Energy demand E of data center energy consumption system _total (t) depends mainly on the energy G consumed by the new server on-state in the t-th time interval _on (t) energy E consumed by the operating Server _c (t), and the energy E that the server cooling needs to consume _cool (t)。

As an example, the basic idea of data center energy demand calculation is as follows:

1. the determining of the power consumption of the starting server specifically comprises the following steps:

the increase in the number of servers operated by the data center increases the power consumption if each server consumes E when opened _on (t) starting up the energy G consumed by the new server in interval t _on 9 t) can be expressed as:

G _on (t)＝N _on (t)E _on (t)；

wherein N is _on (t) represents the number of servers that are operating increasing within interval t.

2. The determining of the power consumption of the server in the working state of the data center specifically comprises the following steps:

1) The on and off states of the server need to be captured for each successive time interval. The number of servers operating in interval t is equal to the number N of servers operating in the last time interval t-1 _w (t-1) adding the number N of servers in the off state that are on for the time interval t _on (t) subtracting the number N of servers in the on state after being closed _off (t). Number of servers N operating in interval t _w (t) can be expressed as:

N _w (t)＝N _w (t-1)+N _on (t)-N _off (t); or (b)

N _w (t)＝N _w,web (t)+N _w,app (t)+N _w,db (t)；

Wherein N is _w,web (t) the number of servers operating for the network group within the time interval t, N _w,app (t) the number of servers operating for the application group within the time interval t, N _w,db (t) the number of servers operating for the database group during the time interval t, N when t=1 _w (t-1) is a given constant.

2) If the energy consumed by each server operating in a time interval is denoted as E _w (t) the power E consumed by the servers operating in interval t _c (t) can be expressed as:

E _c (t)＝N _w (t)E _w ＝(N _web (t)+N _app (t)+N _db (t))E _w ；

3. the determination of the cooling power consumption of the server may specifically include the following steps:

the energy consumption of IT equipment such as cooling servers is directly related to the operating energy consumption of these IT equipment. If the IT equipment adopts different types of cooling processes, the energy consumption calculation modes are different.

An external air cooling method belongs to an air cooling technology. The method does not need an expensive cooler, and the cooling equipment uses the temperature difference between the inside and the outside of the IT equipment of the data center, takes the external air with lower temperature as a cooling air source, and sends the external air to the IT equipment for heat exchange so as to achieve the purpose of cooling.

More specifically, the energy consumption E of the server is cooled by external air cooling during the interval t _cool The calculation formula of (t) can be expressed as:

wherein α represents the internal and external relative temperature ratio of the IT device, b represents the calculated ratio of the internal and external temperatures of the IT device, and t _in Representing the internal temperature of the IT equipment, assuming 35 ℃, t _out The temperature outside the IT equipment is represented, depending on the geographic location in which the data center is located.

The determination of the amount of power consumed by cooling of the data center server using the outside air method is merely an example, but the present invention is not limited thereto and includes calculating the amount of power consumed by cooling of the data center server using other cooling methods.

More specifically, energy demand E of data center within interval t _total The calculation formula of (t) can be expressed as:

E _total (t)＝E _c (t)+E _cool (t)+G _on (t)。

step S130: and determining the electric quantity stored in the data center energy storage device or the electric quantity required to be acquired from the intelligent power grid based on the solar-powered electricity generation output power day-ahead prediction result and the energy demand of the data center.

The renewable photovoltaic energy source, the energy storage device and the intelligent power grid jointly form an energy system.

Because of the continuous expansion of the application field of renewable energy sources and the acceleration of the opening process of the electric power market, most of the energy systems of the data center currently have energy exchange with an external power grid, so that in order to ensure that the energy supply of the energy systems and the energy demand of the energy consumption systems are in a balanced state, when the renewable photovoltaic power generation output power is less than the energy consumption of the data center, the external power grid supplies power to the data center; when the renewable photovoltaic power generation output power is greater than the energy consumption of the data center, the data center transmits power to an external power grid.

In some embodiments of the present invention, determining the amount of electricity stored to the data center energy storage device or the amount of electricity required to be obtained from the smart grid based on the future prediction result of the photovoltaic power generation output power and the energy demand of the data center includes:

More specifically, the basic idea that the energy supply of the energy system and the energy demand of the energy consumption system are in an equilibrium state is as follows:

1) If the future predicted result of the photovoltaic power generation output power is greater than the energy requirement of the data center, the energy storage device of the data center is required to store the energy of the redundant energy.

In the process of delivering redundant electric quantity to an energy storage device by a photovoltaic power station of a data center, the operation variable of the energy storage device needs to be restrained. The constraint variables of the energy storage device mainly take into account the possible maximum and minimum capacities of the energy storage device, the charging and discharging power, the charging and discharging efficiency and the dynamic changes of the electric quantity stored by the energy storage device. Variables that affect the operation of the energy storage device may be defined as follows:

in the method, in the process of the invention,representing minimum and maximum charge power, minimum and maximum discharge power, minimum and maximum capacity level, respectively, of the energy storage device, +.>Respectively representing the charge power and the discharge power of the energy storage device in the t period (also called the t time interval or interval t), mu _c (t),μ _d And (t) represents the charge state and the discharge state of the energy storage device in the t period respectively, and is a 0/1 variable. E (E) _esd (t) represents the capacity of the energy storage device during the period t, the capacity of the energy storage device being E when t=0 _esd Is generally constant. />Indicating the state of charge to discharge transition, +.>The state of transition from discharge to charge is indicated as 0/1 variable. C (C) _esd Represents the upper limit of the charge and discharge times of the energy storage device in one day, alpha _esd ,β _esd At the level of the given percentage of the total,and->Representing the charging efficiency, delta, of a data center energy storage device _esd The self-discharge efficiency of the energy storage device is shown, and Δt is the time interval difference.

Wherein,,and->Is defined as follows:

2) If the future predicted result of the photovoltaic power generation output power is smaller than the energy requirement of the data center, the server is enabled to normally operate in order to meet the energy requirement of the data center, and the data center needs to acquire electric quantity from the intelligent power grid.

Variables involved in exchanging energy between the smart grid and the data center need to be defined as follows:

in the method, in the process of the invention,representing the minimum and maximum power of the smart grid to the data center and the minimum and maximum power of the data center to the smart grid, respectively, in actual situations, generally consider +> P _b (t) represents the power of the smart grid to the data center in the t period, P _se (t) represents the power of the data center to transmit electricity to the smart grid within the t period, mu _b (t) represents the state of the power transmission of the smart grid to the data center in the period of t, mu _se And (t) respectively representing the state of the power transmission of the data center to the intelligent power grid in the t period, wherein the t period and the t period are 0/1 variable.

Wherein the variable mu _b (t),μ _se (t) is defined as follows:

3) If the future predicted result of the photovoltaic power generation output power is equal to the energy requirement of the data center, the electric quantity does not need to be stored in an energy storage device of the data center or obtained from the intelligent power grid.

If the daily forecast result of the photovoltaic power generation output power is equal to the energy demand of the data center, the generated renewable energy just meets the energy demand of IT equipment such as a server, the data center does not need to acquire electric quantity from the intelligent power grid, and no redundant electric quantity is transmitted to the storage device for storage.

Step S140: an average response time for the various types of clients is determined based on the queuing network model.

The service system of the data center comprises all servers of the data center and a dispatcher dis.

The service system of the data center can be regarded as a queuing network model, the requests sent to the data center by each client are respectively distributed to the network group web, the application group app or the database group db by the dispatcher at the entrance of the data center, the requests are satisfied by the servers distributed by each group, and the number of the servers distributed by each group is continuously changed. If a request leaves a group, it may go to another group or leave the data center, depending on the probability of routing between groups of networks, applications, or databases.

In some embodiments of the invention, the queuing network model is built based on an exponential distribution of client request arrival times and service times. Assuming that the arrival time and service time of client requests are exponentially distributed, the scheduler may be modeled as an M/1 queue, and the network group, application group, or database group may be modeled as an M/k queue (where k=n _web 、N _app Or N _db ，N _web The number of servers that are a network group, N _app The number of servers that are the application group, N _db Is the number of servers in the database group, and the sum of the numbers of servers in the three groups is the total server in the data centerNumber N). Furthermore, in view of the complex inference of the average response time of each set of multi-server queues, a network set, application set, or database set may be modeled as separate server sets with the same arrival rate, respectively, and each server within each set may be modeled as an M/M/1 queue.

More specifically, assume that the service system is a Jackson queuing network. Jackson networks are a set of queues with different routing probabilities, and schedulers, network groups, application groups, and database group queues in the network are also referred to as sites.

In some embodiments of the invention, the arrival rate of the client requests is the effective arrival rate of the sites of the client requests determined based on the routing matrix.

More specifically, the arrival rate of client requests is the effective arrival rate of each site, which is calculated by the routing matrix P, and the client of each type i (also called client type i, assuming N _c Type of client, 1.ltoreq.i.ltoreq.N _c ) Will have a different matrix P _i The formula can be expressed as:

the arrival rate of each station is used as a performance metric for each station to evaluate the performance of the entire network. Vector lambda _i The calculation containing the effective arrival rate for each site is as follows:

wherein I represents an identity matrix,representing the effective site arrival rate for client type i.

Since the scheduler and each server within each group can be modeled as an M/M/1 queue, for client type i, the average number of requests L for the scheduler _i,dis An average number of requests L with each server within each group may be employed _i,k The same way of calculation (average number of requests also called queue length) is expressed as:

wherein s is _k Representing site transition probability ρ _k Or ρ _dis Lambda is an intermediate calculated variable _i,k Is the arrival rate of a request of client type i to each server in each group, lambda _i,dis Is the arrival rate of the request of client type i at the scheduler, μ is the service rate of each server in each group, μ _dis Is the service rate of the scheduler, assuming μ _dis ＞＞μ。

For client type i, the average number of requests for each group is the sum of the average number of requests for servers within each group, and the sum of the average number of requests for each site is the average number of requests L for client type i in the service system _i 。

L _i ＝L _i,dis +L _i,web +L _i,app +L _i,db 。

In some embodiments of the invention, the average response time is determined based on littlet's law.

The calculation formula of the average response time can be expressed as:

in terms of data center services, data centers and N _c Service level agreements (SLA, service Level Agreement) are signed between clients of different types, which have different arrival behaviour and different request sizes in the day, and the number of servers of each group is also the same as that of the requestsThe processing of the solution is relevant, so that each type of client has a different average response time. Since the average response time can be used as an evaluation index for SLA, each customer type i has a cost function M that depends on the average response time _i The formula can be expressed as:

if the data center performance can guarantee the average response time W of a certain type of client i in the interval t _avg,i Below the critical time W _c (i.e. average response time to customer type i as specified in SLA), no loss is incurred and the data center will receive benefit a _i . However, if the average response time W _avg,ii Beyond the critical time, the data center will have to pay the penalty slope k _i Proportionally increasing fines.

The invention determines the average response time of various types of clients based on the Jackson queuing network is only an example, and other types of queuing networks can be selected to determine the average response time.

Step S150: obtaining an optimal energy demand configuration of the data center based on an energy dynamic optimization scheduling model aiming at minimizing the operation cost; wherein the decision variables in the energy dynamic optimization scheduling model include at least one of the following variables: the arrival rate of the client request, the real-time electricity price of the smart grid and the photovoltaic power generation output power; the operating cost is obtained based on the following factors: the amount of electricity exchanged by the data center and the smart grid, the energy storage device operating maintenance price, and the average response time of various types of clients.

More specifically, the data center can determine the condition of exchanging energy with the smart grid or the surplus energy stored by the energy storage device according to the photovoltaic power generation output power and the energy demand of the data center; then, according to the average response time of the clients of different types, the income of the SLA between the data center and the clients can be known; by using the electric quantity exchanged between the data center and the intelligent power grid, the dynamic operation condition of charging/discharging of the energy storage device and SLA income as factors influencing the operation cost of the data center, an energy dynamic optimization scheduling model aiming at minimizing the operation cost can be constructed, the optimal energy consumption configuration of the data center can be obtained, and the number of servers which need to be opened and closed in an interval t is determined, so that the operation states of IT equipment and cooling equipment of the data center can be known.

Considering the economic benefits of data center operation, an energy dynamic optimization scheduling model targeting operational cost minimization can be expressed as:

in the formula, PR _g Is the real-time electricity price when the data center and the intelligent power grid trade, M _esd Is the price for the operation and maintenance of the energy storage device. From the expression of the average response time and the calculation mode of the energy consumption of the cooling method, the optimization problem is nonlinear, and the cost objective function focuses on the cost of the data center in terms of meeting the request and the energy consumption of the client, and further comprises the possible benefits generated in the case of selling energy. Based on the energy dynamic optimization scheduling model, the optimal energy demand configuration of the data center can be obtained.

The arrival rate of the client request, the real-time electricity price of the smart grid, the photovoltaic power generation output power and the like can be used as decision variables of an energy dynamic optimization scheduling model.

The arrival rate of the client request, the real-time electricity price of the smart grid and the photovoltaic power generation output power are taken as decision variables of the energy dynamic optimization scheduling model only as examples, and other indexes such as the energy sales price can be selected as the decision variables of the energy dynamic optimization scheduling model.

Step S160: and obtaining a daily time sequence prediction result of each decision variable in the energy dynamic optimization scheduling model based on an exponential smoothing method, obtaining the optimal energy demand configuration in the data center day based on the daily time sequence prediction result, and progressively updating the energy dynamic optimization scheduling model according to the time sequence.

More specifically, in the actual operation process of the data center, the arrival rate of the client request at the interval t, the real-time electricity price of the smart grid, the renewable photovoltaic power generation output power and other decision variables of the energy dynamic optimization scheduling model need to be predicted. The optimal energy consumption configuration of the data center is obtained based on an energy dynamic optimization scheduling model aiming at minimizing the operation cost, the number of servers which need to be opened and closed in an interval t is determined, so that the operation states of the IT equipment and the cooling equipment of the data center can be known, and the predicted value of the next time interval is replaced by a true value according to time sequence progression, so that the operation state of the data center equipment is continuously updated.

The prediction of the time series requires prediction of several parameters when solving the optimization problem, since the actual values of these parameters change continuously over time, which can only be known within the corresponding time intervals, but in order to ensure that the balance of energy supply and energy consumption and the operating costs of the data center are minimized, planning decisions must be made in advance. The parameters to be predicted comprise the arrival rate of the client request, the real-time electricity price of the smart grid, the renewable photovoltaic power generation output power and the like.

Exponential smoothing is a widely used prediction technique. The method is a time sequence Y _t Weights are assigned to different observations of (a), wherein newer observations get more weight than older observations. When the time series exhibits seasonal behaviour (seasonal period having a duration s), the method can be predicted by using the seasonal characteristics of the series.

In some embodiments of the invention, the exponential smoothing method comprises seasonal processing in the Holt-windows three-parameter exponential smoothing method.

Because renewable photovoltaic power generation presents seasonal behaviors, a seasonal processing means in the Holt-windows three-parameter exponential smoothing method can complete the prediction process of each decision variable in the energy dynamic optimization scheduling model, and two parameters are needed to be used: l (L) _t And S is _t The calculation formula can be expressed as:

wherein, 0 gamma, kappa is less than or equal to 1, kappa is a horizontal smoothing coefficient, gamma is a seasonal smoothing coefficient, s is the total time, L _t Represents the level estimate within interval t, Y _t Is the actual value of the time series within interval t, S _t For the season estimation value within interval t, S _t-s Is a seasonal estimate within interval t-s.

Next time interval t+1, predicted value F of time sequence _t+1 Can be expressed as:

F _t+1 ＝L _t *S _t-s+1 ；

the first t values of S sum L _t The initialization method of (1) can be expressed as:

more specifically, the process of the decision variable of the intra-day prediction energy dynamic optimization scheduling model by the exponential smoothing method is as follows:

First, according to the above formula, pair L _t And S is _t Initializing two parameters, and obtaining a predicted value F of the next time interval t+1 according to an exponential smoothing formula _t+1 Will predict the value F _t+1 And adding the service resource allocation and energy consumption configuration schemes into the constructed energy dynamic optimization scheduling model, and obtaining service resource allocation and energy consumption configuration schemes of the interval t through Matlab. F of the last interval prediction when proceeding to time t+1 _t+1 Updating the actual data Y replaced with the new time interval t+1 _t+1 Then utilize the actual data of the current time series, i.e. Y ₁ ,Y ₂ ,…,Y _t ,Y _t+1 Calculating the parameter L according to a formula _t+1 、S _t+1 、F _t+2 The above process is cycled until the end of the time.

Another embodiment for comparing predicted data with actual operation of a data center is shown below and mainly comprises the following steps:

the particular embodiment selects one day as the prediction time unit, and the prediction result is more accurate than long-time prediction. Tests were performed at 1 hour intervals, each requiring predicted energy prices, renewable energy generation, and customer arrival rates, which were used to perform the optimization scheme proposed by the present invention (at intervals), determining the number of servers assigned to each group and the values of other decision variables.

1) Dynamic operating conditions of data center service system:

the number of servers assigned to each group in a day varies with a trend similar to that of the objective function, as depicted in fig. 2. As can be seen from fig. 2, the server is most allocated to App types (application software), then Database types (databases), and at least Web types (websites), which accords with the situation that the service request amount of the mobile phone end is larger in real life.

Fig. 3 depicts the number of servers that are turned on and off during the day as a function of the objective function. As can be seen from fig. 3, there is no case where all servers are turned on at any one hour of the day, which can reduce the cost.

2) Data center energy supply system and smart grid power exchange:

fig. 4 is a comparison between the actual value of the renewable energy production in one day and the predicted value in one day, and as can be seen from fig. 4, there is a certain error between the actual value and the predicted value, but the current average error range is-1%, and the error range is smaller.

Fig. 5 shows the energy purchased from the smart grid by the data center during a day. As can be seen in fig. 5, at 10:00-14:00 time quantum is because the radiance is strong, and the renewable photovoltaic energy that produces is enough, and data center need not to purchase the electric quantity from smart power grids, and other time quantum are like 4:00-8:00 and 16:00-20: the 00 time period can not meet the energy consumption requirement of the data center due to limited photovoltaic power generation capacity, and more electric quantity needs to be purchased from the intelligent power grid.

Fig. 6 shows the throughput and sales variation of renewable photovoltaic energy during the day. As can be seen from fig. 6, at 7:00-18: the irradiation is enough in the 00 time period, and the radiation intensity is larger near noon, so that the renewable photovoltaic power generation output power is higher. During other periods of time there is no irradiance and little renewable energy is produced. As can be seen from fig. 5, at division 10:00-14: during other periods of the day than the 00 period, the data center needs to purchase electrical energy from the grid, which means that the production of renewable photovoltaic energy during other periods does not meet the energy requirements of the data center.

Fig. 7 shows the stored energy sales and capacity usage of the energy storage device during a day. If a certain number of servers are shut down, the data center will receive the remaining power generation and can sell it. The energy storage device provides energy sources in all time periods, and the stable and reliable operation of the system is ensured. According to renewable photovoltaic power generation conditions and server energy consumption, at 11:00-14: during the 00 period, a portion of the stored electrical energy is sold because the selling price is greater than the selling price of the renewable energy source, resulting in more benefits.

Fig. 8 depicts the difference between the daily predicted result and the actual value of the selling price of renewable energy in one day. The average error is-0.01% and the error is small, and the deviation from the true value is small in a few time periods. From the results, the performance of the prediction method is good enough to obtain good prediction data.

Fig. 9 depicts the difference between the intra-day prediction result and the actual value of the energy cost of the smart grid within a day, and the predicted trend is approximately the same as fig. 8. As can be seen from fig. 9, the actual value deviates slightly from the actual value for several time periods, the average error is-0.01%, and the error is small. From the results, the performance of the prediction method is good enough to obtain good prediction data.

3) The prediction effect of the optimization scheme of the invention is as follows:

FIG. 10 shows the results of comparing the objective costs of the simulation method with the optimization scheme of the present invention. The simulation method is randomly scheduled by the data center, and does not consider a cost data center management scheme. The result of the optimization scheme of the invention is different from the simulation result, and the optimization scheme is far superior to the simulation method. The simplified calculation of the average response time is always given by the upper limit of the actual average response time given by the M/k queue, the main assumption of the simplified model being that servers belonging to a group equally distribute their workload and work independently. This is in contrast to the case of M/k queues, where the system dynamically allocates workload among servers, making them use of resources in a better way.

In summary, the optimization method combining the daily prediction and the later prediction is better than the simulation method, and the daily prediction of various decision variables is not greatly different from the real data, so that a certain service protocol can be ensured, and the aims of reducing the cost and the energy consumption are fulfilled.

The data center energy optimization scheduling method comprising the daily forecast and the daily forecast, provided by the invention, utilizes a large amount of historical data to construct a dynamic nonlinear optimization model aiming at minimizing the operation cost based on the daily forecast, and more accurately configures the optimal energy of the data center at a certain time sequence through the daily forecast, and has the advantages that:

1. the scheduling problem is optimized for complex data centers in uncertain environments. In the aspect of energy supply, the invention not only considers the photovoltaic output and energy storage charge and discharge of renewable energy sources, but also introduces a smart grid to perform energy interaction of real-time electricity price, thereby fully utilizing the renewable energy sources; and secondly, the external cooling consumption is considered, and the problem of optimal scheduling of complex multipath data centers in actual operation is comprehensively solved.

2. Aiming at the characteristics of intermittence, randomness and the like of renewable resources, the invention provides a future multi-step day-ahead prediction method based on BiLSTM time sequence, which predicts the output power of photovoltaic in the day-ahead; the method is characterized in that the daily decision variables such as the request arrival rate, the power grid price, the renewable power generation capacity and the like are requested based on an exponential smoothing method, the allocation of server resources and the energy consumption scheme are obtained based on a constructed optimal scheduling model, the decision variables such as the number of servers and the energy consumption configuration of each group are predicted, the service request of a system data center and the renewable energy source output capacity are considered in multiple aspects, the smart power grid price real-time change is realized, and the running safety and economical efficiency of the data center are ensured.

3. Because of the real-time change of service requests, the invention constructs a supply-side photovoltaic power generation, energy storage device and intelligent power grid interaction 3-type energy system unit model, constructs an energy demand model for consumption and cooling consumption of a demand side server, combines two sides to construct a dynamic nonlinear optimization model aiming at minimizing operation cost, adopts queuing theory to solve service transfer in a network, then utilizes intra-day exponential smoothing to conduct time sequence prediction, conducts optimal scheduling according to the result of certain time sequence prediction, obtains the optimal energy configuration of a data center at the time sequence, continuously conducts prediction, optimization and updating on the model according to the progressive of the time sequence, and verifies the effectiveness and superiority of the method in actual operation of the data center through result analysis.

The invention can effectively manage the energy consumption of the data center on the premise of reasonably meeting the calculation task planning and arrangement, reasonably solve the problems of resource scheduling and energy management, and realize the minimization of the operation cost and the stable operation of the data center.

Correspondingly, the invention also provides an energy optimization scheduling system of the data center, which comprises computer equipment, wherein the computer equipment comprises a processor and a memory, the memory is stored with computer instructions, the processor is used for executing the computer instructions stored in the memory, and the energy optimization scheduling system of the data center realizes the steps of the method when the computer instructions are executed by the processor.

The embodiments of the present invention also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the edge computing server deployment method described above. The computer readable storage medium may be a tangible storage medium such as Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disk, a removable memory disk, a CD-ROM, or any other form of storage medium known in the art.

Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein can be implemented as hardware, software, or a combination of both. The particular implementation is hardware or software dependent on the specific application of the solution and the design constraints. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave.

It should be understood that the invention is not limited to the particular arrangements and instrumentality described above and shown in the drawings. For the sake of brevity, a detailed description of known methods is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present invention are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between steps, after appreciating the spirit of the present invention.

In this disclosure, features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and various modifications and variations can be made to the embodiments of the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An energy optimized dispatching method for a data center is characterized by comprising the following steps:

2. The method of claim 1, further comprising, prior to pre-training the BiLSTM predictive network model:

3. The method of claim 1, wherein the amount of power consumed by cooling is determined based on an outside air cooling method.

4. The method of claim 1, wherein determining the amount of power stored to a data center energy storage device or the amount of power required to be harvested from a smart grid based on the pre-day prediction of the photovoltaic power generation output power and the energy demand of the data center comprises:

5. The method of claim 1, wherein the queuing network model is built based on an exponential distribution of client request arrival times and service times.

6. The method of claim 1, wherein the arrival rate of the client requests is an effective arrival rate of the sites of the client requests determined based on a routing matrix.

7. The method of claim 1, wherein the average response time is determined based on litter's law.

8. The method of claim 1, wherein the exponential smoothing method comprises seasonal processing in Holt-windows three-parameter exponential smoothing.

9. An energy-optimized dispatching system for a data center, comprising a processor and a memory, wherein the memory has stored therein computer instructions for executing the computer instructions stored in the memory, which system, when executed by the processor, implements the steps of the method of any of claims 1 to 8.

10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 8.