DE102008036246A1 - Electrical power consumption optimizing method for e.g. personal computer, involves using inactivity phase for disconnection of components of data processing device in condition of reduced power consumption of components - Google Patents

Abstract

The method involves continuously and periodically querying individual components of a data processing device. The components and /or the processing device are disconnected in an operating condition. Inactivity of a user of the processing device and/or components of the processing device is automatically detected. Time period of the inactivity is determined and inactivity phase is determined from the determined time period. The inactivity phase is used for disconnection of the components of the processing device in a condition of reduced power consumption of the processing device. Independent claims are also included for the following: (1) a device for optimizing an electrical power consumption of a data processing device (2) a data carrier for storing software for optimizing an electrical power consumption of a data processing device.

Description

The The invention relates to a method for optimizing the electrical Energy consumption of at least one data processing device, in particular a mobile data processing device, in which from consecutive, periodic or event-driven chronologically logged Usage information of the data processing device or parts the data processing device derived information about future use and to optimize the energy consumption of this system.

power consumption represents in today's PC systems from technical, environmental and meanwhile also economic establish play an important role in typical desktop systems in private Use rather secondary effects like warming and noise by cooling a role, so the rising demand for energy in notebooks directly a problem because the current battery development is no longer able to meet the demand. A shortening the term is a direct consequence. Power management is here the solution. the hardware side Support, which makes it possible to switch off unused components of the PC already today. The algorithms required for use are subject but a need for improvement.

In currently used operating systems have become profile-based Method with static timeouts of the devices to be controlled (typically: Screen, the hard disk and the system state). There are a number of such profiles available, each certain applications (Presentation, Laptop mode, desktop mode). Switching to the respective matching profile is done manually by the user of the system. Problematic On the one hand, there is the need for manual switching and on the other hand the generalization of use cases by only a small number of existing profiles. The user must frequently between Energy savings and annoyance due to wrong decisions in power management weigh. First improvements shows Microsoft in the commercial sector with the integration of the adaptive display timeout control in the power management from Windows Vista. Are wrong decisions of power management detected, the timeout is corrected accordingly upward.

research side Within the framework of a project, a fuzzy-controlled reactive power management system was developed developed, which under observation of different input quantities the Decision to switch off the device applies. Disadvantage here is the purely reactive character, which only allows for the current usage the system, but not on future needs. This is but necessary, there shutdowns especially the hard disk or the entire system only after some time in the off state actually Save energy. Remedy here is learning or adaptive Procedures that allow learn the behavior of the user and the system over time. On the basis of recurring Pattern leaves then determine the potential for switching off a device very accurately. Currently relate known learning methods for controlling the power management exclusively on individual devices. So there exists a procedure which rules in a learning phase for optimal selection of a spin-down strategy of a hard disk determined. In a further approach is taking into account the accessing applications on the basis of a trained Markov model the optimal Shutdown time of the hard disk determined. Adaptive method to Control of hard disk timeout are also known. All these Approaches are However, strong on the field of application (the control of power management the hard drive) and are therefore not suitable for implementation Power management of entire PC systems.

Of the Invention is based on the object, a method of the initially genus to show the prediction of potential to shut off or reduce the power consumption of all unused resources and this then to optimize the power consumption of the data processing device uses.

These The object is achieved according to claim 1 according to the invention in that inactivity the user of the data processing device and / or of Components of the data processing device itself detected be determined, the periods of inactivity and off the determined periods possible Inactivity phases determined be used to shutdown and / or to change to a state reduced power consumption of components of the data processing device and / or the data processing device itself be used.

These The check according to the invention can be continuous, periodically or controlled by events. Can appropriate Potential are determined, this is used to the data processing device or parts of the data processing device to turn off or in to put a state of lower power consumption and thus to optimize the energy consumption of the entire system. Hereinafter the solution becomes detailed described.

The inventive method searches for Components of the data processing device or the entire data processing device optimal times for switching off or to change to states with reduced power consumption. These times are called timeouts below. If the inactivity time t _idle of input variables relevant for a component of the data processing device exceeds this timeout t _o , then this component is switched off or put into a power-reduced state. An optimally selected timeout t _o causes a shutdown or a change to a reduced power state takes place so that the respective component of the data processing device is switched off for a long time or is in the reduced power state and there are no further consequences such as a user or malfunction a short time later held request to the affected part of the data processing device. Due to the sufficiently long residence of the component of the data processing device in the off or in the reduced power state energy is saved compared to the previous operating state.

In order for such an energy saving and no disturbance of the user, different criteria apply for an optimal timeout t _o . If a timeout is used, then it is assumed that in the case of a monitored input variable, which signals inactivity, if a timeout t _{o is reached} by the idling time t _idle , ie at _{id id} > t _o , no activity will probably continue to be detected at least in the short term. A monitored input quantity is considered to be inactive if no activity is detected at this input value for at least a time t _min . Thus, for an optimal timeout t _o, the criterion t _o ≥ t _min applies. As an activity, for example, in the case of the user, an input to the data processing device can be evaluated. If a monitored input parameter indicates inactivity, it is not possible to predict how long this inactivity phase will last or when the next activity will be detected. In order to save energy compared to the current operating state, a component of the data processing device must be switched off for longer than the break-even time t _BE or be in a power-reduced state. The break-even time t _BE is the time after which the additional energy required for the shutdown or the state change corresponds to the energy saved in the switched-off or power-reduced state. This results in the criterion t _idle > t _o + t _BE . The break-even time differs depending on the component of a data processing device and the type of state change. In addition, it is possible that after some time an activity is detected at a monitored input, because the user is disturbed by the shutdown or the change in the power-reduced state (annoyance). A component of a data processing device should therefore at least one time .DELTA.t be _ok off or be in a reduced-power state in order to avoid from the user's perspective too frequent switching off or switching to a reduced-power state. This yields the criterion t _idle > t _o + Δt _ok , where Δt _ok t _BE .

It can be deduced from the stated criteria that, given a Δt _ok, an activity of a monitored input variable after t _idle during Δt _ok , ie at t _o <t _idle ≦ t _o + Δt _{ok, represents} an error case. Since the inactivity time t _{idle of} a monitored input variable can not usually be predicted, the optimal timeout t _{o is} the lowest time for which the error case occurs with a maximum likelihood p _max . Thus, for the optimal timeout t _o : W (t _o <t _idle ≤ t _o + Δt _ok ) ≤ p _max

Around the probability of determining that an activity is a supervised one Input size in one specific time interval occurs after a first training Inactivity times detected in the invention together with a Timestamp captured, stored in datasets and as part of a Learning component set In the following, for such a data set the Term "histogram" used. As inactivity time In this case, the time duration between two activities that have occurred is monitored Input size defined. In the case of the user, this time corresponds to the time interval between two inputs or the time after which he returned.

Dividing the time axis into several discrete areas (eg seconds intervals), one can count in the histogram how often the return to a certain time has taken place. Using the histogram, the absolute frequency of certain return times is determined. Dividing the absolute frequencies by the number of recorded values gives the relative frequency h (t). Stochastically, this can be understood as an empirical survey and thus as a random experiment with the random variable X ≈ "return after time t _idle ". The sample of the experiment includes the inactivity times of the monitored input size. Thus the relative frequency, which is determined with the help of the histogram, corresponds to the probability function f (t).

A Probability W (X ≤ t) is determined by the function of the probability distribution F (t). This one gets one by integration of the probability function f (t). Since that Histogram contains discrete values, applies to the distribution function:

With the distribution function F (t), the probability W (t _o <X ≦ t _o + Δt _ok ) can be determined the. So the equation holds: W (a <X ≦ b) = F (b) - F (a)

Thus, the said probability can be calculated as follows: W (t _o <X ≦ t _o + Δt _ok ) = F (t _o + Δt _ok ) - F (t _o )

The criterion for the aforementioned optimal timeout t _o is thus as follows: F (t _o + Δt _ok ) - F (t _o ) ≤ p _max

The optimal timeout t _o must therefore satisfy the inequality F (t _o + Δt _ok ) - F (t _o ) ≦ p _max with a given distribution F (t), the time interval Δt _ok and the maximum probability p _max . Since this may apply to several timeouts, the smallest value is taken to be greater than or equal to the minimum timeout t _min . Due to the distribution F (t), this timeout depends on the past and thus on the knowledge of the behavior of the monitored input variable. So it is very likely that from this timeout, a started inactivity takes so long that a switch-off or a change of the affected component in a power-reduced state is worthwhile. Thus, this value provides a very high probability of energy savings and low user disturbance.

• 1. so viel wissen, dass empirisch ein passender Timeout bestimmt werden kann.
• 2. nur so viel wissen, dass die histogrammbasierte Timeout-Bestimmung dynamisch auf Verhaltensänderung reagiert.

The learning component is adaptive. The described histogram and thus the distribution of inactivity times change over time. Initially, the histogram is empty. There is no knowledge about past inactivity. This knowledge arises slowly with each user return greater than or equal to the minimum time-out t _min . However, user and system behavior can be very different and change over time. This change is characterized by the fact that certain return times occur in a different frequency. The histogram eventually contains old values that describe the earlier behavior. They are at the beginning of a behavioral change in the majority and thus dominate the statistically determined timeout. This is no longer necessarily the new behavior. The consequence is an increased disturbance of the user. Only with increasing number of new values, the histogram adapts, so that the timeout changes. This process can take a long time for a histogram with many values. This raises the question of how far the knowledge of the histogram may reach into the past. The histogram should be

• 1. know so much that empirically a suitable timeout can be determined.
• 2. Knowing only enough that the histogram-based time-out determination reacts dynamically to behavioral change.

In order to provide for a dynamic adaptation of the histogram, it is provided in a further development of the invention to consider only a specific time window of length Δt _valid . Older values fall away. So old information is lost through oblivion.

to Determine the age of inactivity times in the histogram allowed to not all time ranges are used. For example the data processing device temporarily not in operation or the method of the invention described so far will not be for some time used, such time gaps are not relevant Innaktivitätszeiten and allowed have no influence on the age of a value in the histogram. The omission of certain time intervals in the determination of a Age is achieved by a development of the invention so, that a global time and a local time for the histogram values are introduced the above Reference timestamps are related. The age of one in the histogram contained inactivity time is determined on the local time axis. When adding a inactivity a global is converted to a local timestamp. Certain Time intervals are omitted by the length of the respective time interval the global reference time stamp is added. This can be a new Global timestamp converted into a local timestamp be that for the calculation of an age omitted specific time intervals become.

According to a next development of the invention, it is provided that at least one correction option for timeouts is contained in the learning component. The adapted histogram arises over time. At first it is empty and contains no knowledge. This knowledge grows with added inactivity times. By contrast, forgetting old values can cause the number of known inactivity times to decrease again. With a small sample size, a timeout determined by this knowledge for the next inactivity period is not very meaningful. Thus, additional mechanisms are needed to complement the histogram within the learning component. Such a mechanism will be described below. For the learning method, the period Δt _{ok has been} introduced. An activity of the monitored input size during this time is a mistake both for energy saving and annoyance reasons. In the case of such a disturbing error, it makes sense to correct the timeout upwards. The range Δt _ok , in which the decision of the power management is interpreted as annoying, may be greater than the break-even time. The larger Δt is _ok , the more likely it is that a return of the user towards the end of the period Δt _{ok will} not occur due to irritation. Therefore, the corrected timeout is also dependent on the return time with respect to Δt _ok . Too high dialed timeout is also not always useful. Even a timeout which leads to a shutdown or to a state change of the respective component of the data processing device with energy saving and without interference can be too high. Therefore, if there is no disturbing error case, the timeout can be changed downwards in the direction of the minimum time-out t _min . The described timeout correction is a dynamic response to the current activity of the monitored input data. It ensures a very high timeout with a disturbing error and otherwise an adjustment in the direction of the minimum timeout.

According to a next development of the invention, it is provided to connect the already described concepts of the histogram and the timeout correction in order to determine a common timeout. Both concepts receive the same input data. These consist of the current inactivity times of the monitored input size, the time and also the last calculated timeout. Both have a timeout as a result of their calculation, the histogram the timeout t _oHist and the correction the timeout t _oKorr . The expressiveness of the histogram depends on the number of inactivity times included. Therefore, the histogram should not only account for the resulting timeout. Instead, the timeout determined using the histogram may only have a weighted portion w _hist . Since the validity of the histogram depends on the number of contained values, this must also apply to the weight w _Hist . The remainder of the resulting timeout is the correction timeout with w _Korr . _{We have} w _Korr = 1 - w _Hist . The resulting timeout t _o finally results from the weighted sum of the two timeouts: t O = t oHist · w hist + t oKorr · w corr = t oHist · w hist + t oKorr · (1 - w hist )

The Histogram realizes the memory of the timeout calculation learning behavior. It takes care of a timeout based on the knowledge of past inactivity times the supervised Input data based. The correction module takes into account the dynamic behavior. In this way, the timeout calculation determines a resulting timeout both on the basis of learned knowledge and reactive due of current behavior the supervised Input data.

To one next Further training is provided that using the previously described Method determined timeouts for shutdown or change to a reduced power state of components of the data processing device or the entire Data processing device itself combined into a vector become. Contains this the vector maximum for every respect the power consumption relevant control option in the data processing device a separate timeout. Represented in terms of the overall system This vector shows the timeouts that match the current usage scenario Know about the future is best. Therefore, this vector will be below referred to as reference scheme.

To Alternatively, the method of the invention provides several reference schemes for different Operating modes such. B. battery or mains operation and / or others Characteristics of the usage environment, eg. Stationary, mobile, Network connection, location before. Background here is the additional consideration different user behavior in different operating modes or Use environments. It is assumed that use cases for Example in battery operation often in principle different from those in network operation. For example, in the office many times longer Interruptions of work on the data processing device due to meetings, breaks or phone calls the work traveling in z. B. trains and aircraft rather characterized by continuous work. Lies one of the considered operating modes of the data processing device before or is a considered characteristic of the usage environment recognized, this will be the appropriate reference scheme with the in this time supplied data. Tax decisions are made accordingly derived from this scheme.

Switching off or changing over to a reduced-power state of a component of the data processing device or the data processing device itself can lead to greater energy savings, but at the same time have more serious consequences than switching off or changing the state of other components. A consequence of the control decision is the time required until the corresponding component of the data processing device or the data processing device itself is again in the original operating state. The greater this time, the more disturbing the shutdown or the state change in a wrong decision by the user perceived. Some turn-off decisions may be related because, for example, they are being monitored for the same input. Ensuring that the shutdown or change of state of such components of the data processing device in the order of severity of their consequences, starting from the least serious happens so little disturbing can be tested for the user, whether components of the data processing device off or in Leistungsredu ciated states can be offset. For this purpose, a development of the invention is provided so that timeouts for Abschalt- or state change decisions that are related, are also corrected so that components of a data processing device in the order of severity of the consequences off or in a reduced power Condition be offset.

• 1. Die Timeout-Werte des Referenzschemas werden direkt zur Abschaltung oder zum Wechsel in einen Zustand reduzierter Leistungsaufnahme der Bestandteile einer Datenverarbeitungseinrichtung oder der gesamten Datenverarbeitungseinrichtung selbst verwendet. Besteht eine verhältnismäßig homogene Nutzung der zu steuernden Ressourcen mit häufig wiederkehrenden leicht zu erkennenden Mustern, eignet sich diese Art der Anwendung des Referenzschemas gut. Es ist eine minimale Anzahl an gesammelten Daten notwendig um zuverlässige Aussagen zu erhalten.
• 2. Häufig besteht gerade im Büroalltag eine höchst heterogene Nutzung einer Datenverarbeitungseinrichtung unter Verwendung vieler verschiedener Applikationen. Die Art der Applikation bestimmt hierbei häufig, wie die Datenverarbeitungseinrichtung genutzt und welche Bestandteile wie oft verwendet werden. Eine im Rahmen dieser Weiterbildung der Erfindung vorgeschlagene und erfasste Anwendungsmöglichkeit sieht daher vor, ein eigenes Referenzschema für jede Applikation aufzubauen. Befindet sich eine Applikation im Anwendungsfokus des Nutzers, so wird das zugehörige Schema mit den in diesem Zeitraum gesammelten Daten versorgt und entsprechende Steuerentscheidungen für die leistungsreduzierten Zustände werden aus diesem abgeleitet. Hierfür nötig ist die bereits beschriebene Weiterbildung der Erfindung nach Anspruch 6 um die jeweils zu einer Applikation gehörenden Histogramme nur bei aktiv genutzter Applikation altem zu lassen.
• 3. Liegt eine Nutzung der Datenverarbeitungseinrichtung vor, die heterogen jedoch unabhängig von Applikationen ist oder in der mehr als eine Applikation häufig involviert sind, spiegeln die unter 1. und 2. beschriebenen Anwendungsmöglichkeiten die Nutzung der Datenverarbeitungseinrichtung oder von Teilen der Datenverarbeitungseinrichtung möglicherweise schlecht wider. So kann beispielsweise zur Beantwortung einer E-Mail eine kurze Recherche im Internet nötig sein. Der hierfür verwendete Browser wird in diesem konkreten Fall jedoch häufig anders geartet als beispielsweise beim Ersteigern eines Artikels in einem Auktionshaus benutzt. Die Bildung eines unscharfen Profils für den Browser ist in diesem Fall die Konsequenz. Die dritte Anwendungsmöglichkeit der Weiterbildung dieser Erfindung, sieht für eine solche Art der Nutzung daher vor, vergleichbar zu den Anwendungsmöglichkeiten unter 1. und 2. nur ein Referenzschema für die gesamte Datenverarbeitungseinrichtung und/oder mehrere Referenzschemata für verschiedene Applikationen der Datenverarbeitungseinrichtung (weitere im Fall von verschiedenen Betriebsmodi) aufzubauen. Abweichend wird jedoch ein Referenzschema nicht direkt zur Steuerung verwendet, sondern es wird das am besten dazu passende Schema unter Verwendung eines geeigneten Verfahrens aus einer Sammlung fester Schemata ausgewählt. Hintergrund hierbei ist die Tatsache, dass im Fall von heterogener Nutzung der Datenverarbeitungseinrichtung das verhältnismäßig träge Referenzschema sich nur allmählich anpasst und daher häufig nur unscharf das derzeitige Nutzungsszenario widerspiegelt. Zur Beseitigung dieser Unschärfe wird daher ein scharf auf ein Nutzungsszenario zugeschnittenes Schema ausgewählt. Die Wahl des passenden Schemas erfolgt in geeigneter Weise je nach Ausprägung der Erfindung durch das Verfahren selbst, durch ein anderes Verfahren/eine andere Vorrichtung und/oder durch den Benutzer der Datenverarbeitungseinrichtung.
• 4. Die vierte im Rahmen dieser Erfindung vorgestellte Anwendungsmöglichkeit sieht vor, das aktuelle Referenzschema abspeichern zu können. Hierdurch lassen sich die derzeit adaptierten Timeouts definiert durch den Nutzer einem konkreten Nutzungsszenario zuordnen. Auf diese Weise lässt sich während des Betriebs der Datenverarbeitungseinrichtung Wissen in Form von scharfen Schemata sammeln, welches in Verbindung mit der unter 3. vorgestellten Anwendungsmöglichkeit die Qualität der Entscheidung des Powermanagements im Fall heterogener Nutzung stark erhöhen kann.

The invention provides in an additional development to use the reference scheme constructed with the means described so far of the method according to the invention in various ways in order to change into the corresponding states with reduced power consumption of the data processing device or components of the data processing device. Developments cover the four proposed in the context of the invention and described below options.

• 1. The time-out values of the reference scheme are used directly for switching off or changing to a state of reduced power consumption of the components of a data processing device or the entire data processing device itself. If there is a relatively homogeneous use of the resources to be controlled with frequently recurring, easily recognizable patterns, this type of application of the reference scheme is well suited. A minimum amount of collected data is necessary to get reliable information.
• 2. Often there is a highly heterogeneous use of a data processing device using many different applications, especially in everyday office life. The type of application often determines how the data processing device is used and which components are used and how often. A proposed in the context of this embodiment of the invention and recorded application possibility, therefore, to build a separate reference scheme for each application. If an application is in the application focus of the user, the associated schema is supplied with the data collected during this period and corresponding control decisions for the reduced-power states are derived therefrom. For this purpose, the already described development of the invention according to claim 6 is necessary in order to allow the respective histograms belonging to an application to be aged only when the application is actively used.
• 3. If there is a use of the data processing device that is heterogeneous, however, independent of applications or in which more than one application is frequently involved, the application possibilities described under 1 and 2 may reflect poorly the use of the data processing device or parts of the data processing device. For example, a short search on the Internet may be necessary to answer an e-mail. However, the browser used for this purpose is often different in this specific case than, for example, when bidding on an item used in an auction house. Creating a fuzzy profile for the browser is the consequence in this case. The third possible application of the development of this invention, therefore, provides for such a type of use, comparable to the applications under 1. and 2. only a reference scheme for the entire data processing device and / or multiple reference schemes for different applications of the data processing device (further in the case of different operating modes). By way of derogation, however, a reference scheme is not used directly for control, but the best-fit scheme is selected from a collection of fixed schemes using a suitable method. The background here is the fact that in the case of heterogeneous use of the data processing device, the relatively sluggish reference scheme adjusts only gradually and therefore often blurred reflects the current usage scenario. To eliminate this blur, therefore, a scheme tailored to a usage scenario is selected. The choice of the appropriate scheme is carried out in a suitable manner, depending on the form of the invention by the method itself, by another method / device and / or by the user of the data processing device.
• 4. The fourth application proposed in the context of this invention envisages being able to save the current reference scheme. In this way, the currently adapted timeouts defined by the user can be assigned to a concrete usage scenario. In this way, during the operation of the data processing device, knowledge can be collected in the form of sharp diagrams, which, in conjunction with the application option presented under 3, can greatly increase the quality of the decision of power management in the case of heterogeneous use.

ever in the manner of the usage scenario of the data processing device let yourself one of the possibilities described above make optimal use of the application of the reference schema

For a device for optimizing the electrical energy consumption of at least one data processing device, separate protection is claimed. This device is characterized by having means for detecting Inaktivnahmen of the user of the data processing device and / or the data processing device itself has that it has means for determining the periods of inactivity and that means for determining each matching energy saving states of parts of the data processing device or the data processing device itself is provided.

The Device has the necessary hardware and software for the implementation of inventive method on. There are means of checking the activity or inactivity provided by the user and / or the data processing device. The apparatus according to claim 17 can be fully implemented in hardware Software or realized by a combined hardware / software solution become. A software solution The invention can be stored on a data carrier according to claim 18 and transfer become.

An embodiment of the invention, from which further inventive features result, will be described below and in 1 - 4 supportive illustrated. Show it:

1 : Schematic illustration of the method according to the invention in a data processing device

1

The entire method according to the invention, stored on a disk for use with an operating system

2

One Quantity of collected and stored data, eg. B. a histogram.

3

The Error correction component

4

The Component for connecting the output of the histogram and the correction component

5

The Component that has one or more uses of the reference schema implemented

6

A system clock

7

One Time-based power management

8th

The Control layer of the hardware

9

The Monitoring layer of the hardware

2 : Schematic illustration of the determination of the optimal timeout from the distribution function of the inactivities

10

exemplary graphical representation of a distribution function of collected periods of inactivity

11

exemplary Marking a useful area for saving energy

3 : Formation of reference schemes from the calculated timeouts

12

One exemplary reference scheme for the battery operation

13

One exemplary reference scheme for the network operation

14

Timeout calculation the individual components of the data processing device to be controlled

4 : Uses of the reference schemes depending on the type of use of the data processing device

15

education a reference schema per application.

16

use the reference scheme for direct control.

17

selection a sharp scheme based on the current reference scheme.

18

storage of the current reference scheme to then z. B. as a sharp scheme choose to be able to.

1 schematically shows the embedding of the method according to the invention 1 in a data processing device. On the input side - indicated by the block HW monitoring 9 - Provided by the data processing device or the running operating system necessary data such as user activity. On the output side, the method provides calculated timeout parameters, which are a time-based power management 7 serve as input. The power management can be done using a control layer 8th set the power-reduced states of the considered components of the data processing device or data processing device itself accordingly or turn them off. If inactivity phases of a component of the data processing device or of the entire data processing device occur, then these become the corresponding histogram 2 entered and also the error correction component ( 3 ).

An exemplary histogram is in 2 shown. Here, the distribution function is the inactivity time 10 displayed in the histogram over time and valid at the current time. With the aid of this distribution function, the timeout t _o is determined using the condition F (t _o + Δt _ok ) - F (t _o ) ≤ p _max , which is best for controlling the change in power-reduced states or the shutdown of the supplied hearing-impaired component of the data processing device or the data processing device itself and the most useful for saving energy energy useful area 11 limited to the bottom.

The two parameters determined from the histogram and the error correction component become a following component 4 summarized. As in 3 For each component of the data processing device that allows the control of the change of power-reduced states or the shutdown, a plurality of such parameters are determined for the different supply modes 14 , These become several so-called reference schemes (eg 12 and 13 ) summarized.

Depending on the type of use of the data processing device, one of the four types of use presented (visualized by the block "image" 5 ) of the reference schemes are selected to the parameters for the timeout-based power management 7 to determine. The system clock 6 is used to determine inactivity times for both the method of the invention 1 as well as the timeout-based power management 7 needed.

4 , refines the picture block 5 , First of all, it is possible to set the parameters of the reference scheme directly for time-based power management 7 to use 16 , If there is a heterogeneous use of the data processing device, the formation of a reference scheme per application offers itself 15 at. If several applications are involved in a usage scenario or if an application is used differently in different usage scenarios, then the selection of the sharp scheme, which fits best to the currently determined reference scheme, offers itself 17 , If the data processing device has been used for a long time in the context of a typical usage scenario, the now well-adapted reference scheme can be stored as a sharp schema suitable for this scenario 18 ,

Claims (18)

Method for optimizing the electrical energy consumption of at least one data processing device, in particular a mobile data processing device, in which the use of the individual components of the data processing device and / or the data processing device as a whole is interrogated continuously, periodically or event driven, and components and / or the data processing device itself in itself characterized in that inactivities of the user of the data processing device and / or components of the data processing device itself are determined, the periods of inactivities are determined and from the determined time periods possible inactivity phases determined to shut down and / or change to a state of reduced Leistungsauf acquisition of components of the data processing device and / or the data processing device itself. Method according to claim 1, characterized in that that determined periods of inactivity in a learning component be adjusted, wherein the learning component at least a lot recorded data via Duration of inactivities contains. Method according to claim 2, characterized in that that frequencies and / or distributions of detected Data are determined. Method according to claim 2 or 3, characterized that the recorded data times and / or each assigned an age become. Method according to one of claims 2 to 4, characterized that collected data whose age exceeds a specified value, be removed from the respective dataset. A method according to claim 3, 4 or 5, characterized that in determining the age of registered in the dataset Data at least one time interval does not affect the age. Method according to one of claims 2 to 6, characterized that in the learning component at least one correction option for ascertained Inactivity, which for switching off and / or to change to a power-reduced state of components of the data processing device and / or the Data processing device itself can be used included is. Method according to one of the preceding claims, characterized characterized in that the determined from the collected data and a correction possibility inactivity combined by combination into a common inactivity phase, which for switching off and / or switching to a reduced power Condition of components of the data processing device and / or the data processing device itself is used. Method according to one of the preceding claims, characterized in that determined Inactivity phases, which can be used to switch off and / or change to a power-reduced state of components of the data processing device and / or the data processing device itself, are combined to form a reference scheme. Method according to claim 9, characterized in that that for the reference scheme at least one operating mode and / or at least a characteristic of the usage environment of the data processing device considered become. Method according to claim 9 or 10, characterized that some or all of the determined inactivity phases resulting decisions to shutdown and / or change in a state of reduced power consumption of components the data processing device and / or the data processing device even chronologically according to the order of severity of their consequences respectively. Method according to one of claims 9 to 11, characterized that the inactivity phases of the reference scheme directly for switching off and / or to change to a state of reduced power consumption of components of the data processing device and / or the data processing device itself be used. Method according to one of claims 9 to 12, characterized that for single or multiple applications, which on the data processing device accomplished be used, each with its own reference scheme. Method according to one of claims 9 to 13, characterized that one of a collection of schemas fits one to a reference schema Scheme for switching off and / or switching to a reduced state Power consumption of components of the data processing device and / or the data processing device itself is used. Method according to claim 14, characterized in that that the user of the data processing device, another Device, another method and / or the method itself selects a schema suitable for a reference schema. Method according to one of claims 9 to 15, characterized that a reference scheme can be stored. Device for implementing a method according to at least one of the preceding claims, characterized that the device in hardware or in software or in combination hardware and software. disk for storing the software according to claim 17.