CN102662325B - Improved adaptive learning tree power supply management method - Google Patents
Improved adaptive learning tree power supply management method Download PDFInfo
- Publication number
- CN102662325B CN102662325B CN201210135838.2A CN201210135838A CN102662325B CN 102662325 B CN102662325 B CN 102662325B CN 201210135838 A CN201210135838 A CN 201210135838A CN 102662325 B CN102662325 B CN 102662325B
- Authority
- CN
- China
- Prior art keywords
- free time
- state
- node
- time length
- length
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a power supply management forecast method for a learning tree based on an idling time length, and relates to a network technology. An idling time length node is added on the basis of the conventional adaptive learning tree structure based on the probability; an idling time length value is used as a forecast basis, and an equipment entering mode at the idling time is controlled by using a corresponding low power consumption state; after the idling time, the idling time length forecast value is weighted and updated by adopting actual historical probability statistics of various power consumption states in a learning tree; and by adopting an N system method, the matching process for historical paths in the learning tree is avoided. According to the method, relatively high accuracy of the idling time length forecast value of the equipment is guaranteed, so that the equipment is relatively low in power consumption; and furthermore, the complexity of an adaptive learning tree forecasting and updating process is reduced.
Description
Technical field
The present invention relates to artificial intelligence field and built-in field, especially adopt Intelligent Forecasting to carry out the built-in field of device power supply (DPS) management.
Background technology
In today of embedded technology high speed development, along with the diversification of mobile terminal function, power problems has caused that people more and more pay attention to, and power management techniques has become the important indicator of weighing a mobile terminal performance.Dynamic power management (Dynamic Power Management, DPM) is as can be by operating system opertaing device electric power starting or close to reach the technology of saving electric power and obtained increasing research.Generally speaking, DPM strategy is divided into 3 classes: overtime strategy, predicting strategy and randomized policy.The basic thought of overtime strategy is to preset a series of timeout thresholds, once the continuous idle time exceedes this threshold value, is just switched to respective sleep mode, and threshold value can be fixed, and also can adjust with the variation self-adaptation of system loading.Once predicting strategy predicts this free time and can make up state and switch the power consumption penalty of bringing and just equipment is placed in to certain low power consumpting state.Randomized policy is using DPM as stochastic optimization problems, describes the behavior of equipment by setting up Stochastic Decision-making model.
Adaptive learning tree (Adaptive Learning Tree, be abbreviated as ALT) be a kind of predict model early proposing, this tree is made up of decision node (representing with justifying), predicted branches (dotting), historical branch (representing with solid line) and leaf node (representing with rectangle).Each decision node represents the residing state of equipment in one period of free time in the past, predicted branches is used for connecting decision node and leaf node, historical branch is used for connecting adjacent decision node, and leaf node represents the degree of confidence (seeing accompanying drawing 1) of equipment state.ALT has mainly adopted free time cluster IPC principle (Idle Period Clustering).
Free time is defined as: enter idle condition from system or equipment and exit from idle status duration to it.Similarly, holding time is its continuous in running order institute's duration.Therefore, the global behavior of system can be simulated by the time series of one group of working time and free time.In the time that a free time, length was enough offset the caused power consumption of device shutdown, system can obtain power consumption saving by closing device.
Suppose
nbe the low power consumpting state number that system has, system has altogether
n+ a kind of power supply status.With
represent power supply status set (
corresponding working state, all the other the like), system state rank is higher so, and its state power consumption is lower, and the power consumption of mutually switching with duty is higher.To definite free time
, its power consumption
be calculated as follows:
（1）
Wherein,
to be switched to from idle condition
the time of state,
be from
state is switched to the time of idle condition,
with
to switch accordingly power consumption levels,
represent state
iunder power consumption.The lower state power consumption of power consumption state loss based on lower has the hypothesis of higher switching power consumption simultaneously,
should be more than or equal to
and in the time that it is equal, obtain
represent state
iand state
i+ 1 threshold limit value.By like this, can ask for the threshold time of every a pair of adjacent power consumption state, use
represent power consumption state
with
between threshold value:
（2）
Thereby taking the threshold value that obtains as boundary, time coordinate can be divided into
n+ 1 interval, can be by a given free time and some intervals index value
associate, determine with this optimal power state that equipment should enter:
（3）
Therefore, one group free time sequence can be converted into the best low power consumpting state that represents each free time
value sequence (
), this process is called to IPC.
ALT strategy hypothesis equipment has under idle condition
n(total status number is to plant low power consumpting state
n+ a kind, comprise a kind of normal operating conditions), then based on IPC (Idle Period Clustering) principle by each section of idle condition time span of equipment corresponding to an integer numerical value
represent, and each
value is corresponding the optimal operational condition of equipment again.For
n+ a kind of state, can use state set
represent all states, for length be like this
ma string historical free time path sequence, just can use
represent, wherein
,
represent recent free time
value, then according to this historical series
predict that next contingent free time is corresponding
value.So ALT utilizes historical free time path sequence to predict the process of the length of free time next time.
ALT has following advantage: the one, and this tree root border free time historical record and setting up factually has good assurance to equipment behaviour in service in macroscopic view; The 2nd, ALT can upgrade automatically, reasonable changing condition (the Eui-Young Chung that has reflected that equipment uses, Benini, L., De Micheli G. Dynamic power management using adaptive learning tree, Computer-Aided Design, 1999,7 (11): 274-279.).For the predicting strategy research that adopts ALT, mainly contain based on the prediction of device physical status historical record and the method based on free time length prediction at present.
Predict that based on device physical status historical record existing method is to carry out predicted state selection according to probability about using, then accurately whether upgrade respective path probability according to predicting (He Kejia. the adaptive learning predicting strategy based on probability, computer engineering, 2010,36 (10): 215-220.).The method has maintained actual free time situation of upgrading prediction window and recording historical equipment under given path, in selection window, occur that maximum records is as predicting the outcome next time, when renewal, abandon the oldest record, obviously, a kind of increase of state probability will inevitably cause the reduction of other several state probabilities.
Aspect free time length Forecasting Methodology, existing document proposes to predict free time length upgrade the predicted time (Qi Longning on this path based on time expectation value, Hu Chen, Zhang Zhe etc. expect the DPM predicting strategy of table based on the time, Circuits and Systems journal, 2007,2 (12): 89-93.).The method to equipment at every turn in low power consumpting state
itime energy consumption and free time
trelationship description (Sandy Irani, Sandeep Shukla, Rajesh Gupta. Competitive Analysis of Dynamic Power Management Strategies for Systems with Multiple Power Saving States. Proceedings of the 2002 Design, Automation and Test in Europe Conference and Exhibition, 2002,2:1530-1591.), suppose that free time distribution has probability density function
, obtain the energy consumption of equipment in the time of state i and expect as shown in Equation (1):
（4）
Wherein
that equipment is in state
itime power consumption,
that equipment is from state
ibe switched to the power consumption penalty of normal operating conditions,
to expect free time.Thereby draw using free time expectation value as basis for forecasting, can obtain in theory minimum power consumption penalty.In renewal process, this strategy is taked exponential average method pair
value is weighted renewal, to carry out prediction next time.The computing formula of exponential average forecast updating algorithm is as follows:
（5）
Wherein
aa constant between 0 to 1, represent with last time actual value degree of closeness,
alarger, represent that predicted value more approaches the actual value of last time.
There is corresponding defect in above-mentioned two kinds of methods: the former exists higher power consumption, and that the prediction free time length update method that the latter uses can cause with actual deviation is larger, also causes it to have higher power consumption.
Summary of the invention
The present invention is directed to the high power consumption of prior art, defect that predicated error is larger, propose a kind of adaptive learning tree power management Forecasting Methodology of improving, for mobile terminal peripheral power supply management provides the method that can obtain more low-power consumption loss.
The present invention increases free time length node as basis for forecasting in the learn trees based on probability, adopts the probability statistics of historical free time to upgrade it, adopt "
nsystem " index value of number keeping track of history path sequence.Concrete technical scheme is as follows:
In the learn trees structure based on probability, on each decision node of the bottom, increase a leaf node that represents free time length node, leaf node is the brotgher of node with the state probability node that decision node in learn trees structure based on probability is connected;
According to free time length and corresponding power consumption state probability, the last time of in learn trees, preserving predicted free time length, call formula:
（6）
Calculate this prediction free time length value, according to this prediction free time length value, learn trees is upgraded.Wherein,
represent the probability that power consumption state corresponding to actual free time length preserved in learn trees,
for actual free time length,
for predicting free time length value last time,
for this prediction free time length value; In the time that the available free time occurs, opertaing device enters low power consumpting state; Equipment free time finishes, and opertaing device enters the use request of mode of operation with response user.
By corresponding each historical path sequence one unique "
nsystem " (whole historical path sequence are (to be assumed to be under the length of setting
l), the length of the power consumption state obtaining is
lall combinations; In implementation procedure, historical path sequence is what close on most in the past
lthe length that in individual free time, the residing actual power loss state of equipment forms is successively
lsequence), the historical path sequence 210 that 2 kinds of low power consumpting state length are 3 as having can be converted to " three-shift " number
=2*3*3+1*3=21.
The effect of this index value is to save complicated route matching process, each index value has just represented a historical path sequence, has preserved state probability and the prediction free time length value of corresponding historical path sequence in the structure during program realizes under each index value.With the relation of prediction free time length be exactly, obtained index value corresponding to historical path sequence, just can directly get the prediction free time length value under this index value.
The present invention uses historical outline statistics to upgrade free time length, can be more accurate to following free time length prediction, thus the low power consumpting state of better controlling free time equipment should entering while arriving is realized more power consumption and is saved; In learn trees renewal process, adopt "
nsystem " form, each historical path can be represented with a shaping numerical value, thereby avoided the complicacy of process, promote the efficiency realizing.
Brief description of the drawings
Fig. 1 is original adaptive learning Tree-structure Model;
Fig. 2 is based on probability adaptation learn trees structural model;
Fig. 3 is the model that the present invention uses structure;
Fig. 4 is schematic flow sheet of the present invention.
Embodiment
Below in conjunction with accompanying drawing, enforcement of the present invention is illustrated.
Fig. 1 is original adaptive learning Tree-structure Model.
According to free time equipment status, equipment state degree of confidence set up learn trees structural model.
Learn trees is made up of decision node (representing with circle), predicted branches (dotting), historical branch (representing with solid line) and leaf node (representing with rectangle).Decision node represents equipment status in one period of free time in the past, and leaf node represents the degree of confidence of equipment state, and predicted branches connects decision node and leaf node, and historical branch connects adjacent decision node.
Fig. 2 is based on probability adaptation learn trees structural model.
This tree is made up of decision node (representing with circle), predicted branches (dotting), historical branch (representing with solid line) and leaf node (representing with rectangle).Decision node represents equipment status in one period of free time in the past, and leaf node represents equipment state probability, and predicted branches is used for connecting decision node and leaf node, and historical branch is used for connecting adjacent decision node.
Fig. 3 is the learn trees structural model that the present invention sets up.
On each decision node of learn trees of the present invention bottom in the learn trees structure based on probability, increase a leaf node that represents free time length node, the state probability node that leaf node is connected with decision node is the brotgher of node.This leaf node comprises state probability node and prediction free time length node.
Learn trees structure of the present invention comprises decision node (representing with circle), predicted branches (dotting), historical branch (representing with solid line) and leaf node (representing with rectangle) equally, except the structure retaining based on probability adaptation learn trees, leaf node is also set and (uses field
represent) the next free time length of predicting of expression.In the learn trees structure based on probability, on each decision node of the bottom, increase a leaf node that represents free time length node, leaf node is the brotgher of node with the state probability node that decision node in learn trees structure based on probability is connected.As shown in the figure, decision node e connects rectangle leaf node part (comprise state probability node 1/5,1/5,3/5 and
) representative carries out certain state in pilot process, decision node
mthe rectangle leaf node part that connects (comprise state probability node 1/3,1/3,1/3 and
) expression original state.
Fig. 4 is schematic flow sheet of the present invention.Specifically comprise the steps:
Initialization; In learn trees, obtain once free time length value according to the historical path providing; Free time arrives, and opertaing device enters corresponding low power consumpting state; Free time finishes, according to this actual free time length value and learn trees in historical probability statistics renewal learning tree in the predicted value of respective paths, and opertaing device enters mode of operation; Repeat above step, in the time that equipment turn-offs, finish whole flow process.
(1) initialization study tree tree structure
Set up historical series index, set historical path sequence length, (whole historical path sequence are (to be assumed to be under the length of setting to the structure variable such as predicted state probability and predicted time length information of the whole historical path sequence of initialization
l), the length of the power consumption state obtaining according to IPC principle is
lall combinations; In implementation procedure, historical path sequence is what close on most in the past
lthe length that in individual free time, the residing actual power loss state of equipment forms is successively
lsequence).
(2) determine free time length according to historical series
In the time that equipment is in running order, the index value corresponding according to the historical path sequence in learn trees obtains the length value of next time predicting free time: in learn trees
(initial historical path sequence is chosen as full 0 to the save value of field, and the historical path sequence that is 3 as length is
).
(3) in the time that the available free time occurs, opertaing device enters low power consumpting state
According to the free time length predicted value of obtaining, by being relatively divided into equilibration time a low power consumpting state that adopts the equipment that IPC principle obtains, thereby in the time that free time arrives, the device power supply (DPS) management function opertaing device providing by calling system enters this state model (as called SetDevicePower () function under WinCE system).
(4) learn trees upgrades
Actual free time of equipment finishes, the historical path sequence of calculating preseting length "
nsystem " (be similar to binary computing method, the historical path sequence 210 that 2 kinds of low power consumpting state length are 3 as having can be converted to " three-shift " number to index value
=2*3*3+1*3=21), (the probability denominator of all state probability nodes adds 1 to upgrade the predicted state probability node of preserving in the learn trees structure under this index value, the state probability node molecule that equipment actual power loss state is corresponding adds 1), upgrade prediction free time length according to new predicted state probability, method is as follows: adopt actual free time length to be multiplied by probability that power consumption state corresponding to actual free time length preserve in learn trees and add prediction free time length and be multiplied by 1 and deduct the probable value that actual power consumption state corresponding to free time length preserved in learn trees, call formula:
calculate this prediction free time length value, learn trees is upgraded.Wherein,
represent the probability that power consumption state corresponding to actual free time length preserved in learn trees,
for actual free time length,
for predicting free time length value last time,
for this prediction free time length value.
"
nsystem " effect of index value is can save complicated route matching process in the realization of specific procedure; and each index value has just represented a historical path sequence, has preserved state probability and the prediction free time length value of corresponding historical path sequence in the structure during program realizes under each index value.With the relation of prediction free time length be exactly, obtained index value corresponding to historical path sequence, just can directly get the prediction free time length value under this index value.
(5) opertaing device enters mode of operation
Equipment free time finishes, and the device power supply (DPS) management function opertaing device providing by calling system enters the use request (as WinCE system under called SetDevicePower () function) of mode of operation with response user.
(6) repeat the 2nd to the 5th step until equipment enters off state.
Below illustrate:
At one's leisure, length is obeyed in the situation that is uniformly distributed conditional probability the inventive method is tested.Design parameter arranges as follows: the power consumption of each state is respectively
,
,
, switch power consumption and be respectively
,
,
, historical series length is set to 3, and moving window length is set to 10, and the exponential average coefficient a value based in expectation value strategy is 0.5; Evaluate the performance of dynamic power management Forecasting Methodology, the common main contention that uses, Forecasting Methodology produces the ratio of power consumption and the desirable power consumption of off-line, and contention is lower, shows that the effect of power consumption saving is more remarkable.Adopt above parameter arrange the present invention and existingly carry out contrast test based on time expectation value strategy, to obtain the evaluation of objective.
Concrete contention comparing result is in table one.In table, method 1 represents the contention that the present invention can reach, method 2 represents to adopt prior art, the contention that can reach based on time expectation value method, as can be seen from the table, the power consumption of two kinds of methods is respectively higher than the desirable power consumption 20.9% and 23.69% of off-line, the effect that power consumption of the present invention is saved, lower than based on 2.79 percentage points of expectation value strategies, has better reached the saving of power consumption.
Table 1 is uniformly distributed lower prediction contention contrast
| Contention | Method 1 | Method 2 |
| Experiment number 1 | 1.21 | 1.2314 |
| Experiment number 2 | 1.2127 | 1.2394 |
| Experiment number 3 | 1.2105 | 1.2409 |
| Experiment number 4 | 1.2065 | 1.2365 |
| Experiment number 5 | 1.2055 | 1.2365 |
| Mean value | 1.209 | 1.2369 |
Claims (2)
1. one kind is improved adaptive learning tree power management Forecasting Methodology, it is characterized in that: comprise step: (1) sets up historical path sequence index, set historical path sequence length, the predicted state probability of the whole historical path sequence of initialization and predicted time length information, whole historical path sequence are that the length of the power consumption state that obtains according to free time cluster IPC principle is all combinations of whole historical path sequence preseting length L; (2), in the time that equipment is in running order, the index value corresponding according to the historical path sequence in learn trees obtains the length value of next time predicting free time; (3) according to the free time length predicted value of obtaining, by being relatively divided into equilibration time a low power consumpting state that adopts the equipment that IPC principle obtains, in the time that free time arrives, the device power supply (DPS) management function opertaing device that calling system provides enters this state model; (4) finish when actual free time, calculate " N system " index value of the historical path sequence of preseting length, upgrade the predicted state probability node of preserving in the learn trees structure under this index value, the denominator of all state probability nodes adds 1, the state probability node molecule that equipment actual power loss state is corresponding adds 1, according to actual free time length T
actual, and the corresponding power consumption state probability P of preserving in learn trees
actual, prediction free time length T
old, call formula: T
new=T
actual× P
actual+ T
old× (1-P
actual) calculate new prediction free time length value T
new, use T
newupgrade prediction free time length; (5) equipment free time finishes, and the device power supply (DPS) management function opertaing device that calling system provides enters mode of operation; (6) repeating step (2) is to (5) until equipment enters off state.
2. method according to claim 1, it is characterized in that, described learn trees is to increase a leaf node that represents free time length node on each decision node of the bottom in the learn trees structure based on probability, and the state probability node that leaf node is connected with decision node is the brotgher of node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210135838.2A CN102662325B (en) | 2012-05-04 | 2012-05-04 | Improved adaptive learning tree power supply management method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210135838.2A CN102662325B (en) | 2012-05-04 | 2012-05-04 | Improved adaptive learning tree power supply management method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102662325A CN102662325A (en) | 2012-09-12 |
| CN102662325B true CN102662325B (en) | 2014-10-15 |
Family
ID=46771831
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210135838.2A Active CN102662325B (en) | 2012-05-04 | 2012-05-04 | Improved adaptive learning tree power supply management method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102662325B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103853918B (en) * | 2014-02-21 | 2017-01-04 | 南京邮电大学 | A kind of cloud computing server dispatching method based on free time prediction |
| CN105468823B (en) * | 2015-11-16 | 2019-01-25 | 广州广电运通金融电子股份有限公司 | A kind of self-service device energy-saving control method and device |
| CN114048197B (en) * | 2022-01-13 | 2022-04-19 | 浙江大华技术股份有限公司 | Tree structure data processing method, electronic equipment and computer readable storage device |
| CN116610204B (en) * | 2023-07-19 | 2023-11-14 | 碳丝路文化传播(成都)有限公司 | Power management method, system, electronic equipment and medium for electric equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1696847A (en) * | 2005-06-14 | 2005-11-16 | 上海理工大学 | Power supply management method for electronic apparatus |
| CN101754437A (en) * | 2008-12-19 | 2010-06-23 | 英特尔公司 | Handling sensors in a context aware platform |
| CN102208028A (en) * | 2011-05-31 | 2011-10-05 | 北京航空航天大学 | Fault predicting and diagnosing method suitable for dynamic complex system |
| CN102436630A (en) * | 2010-07-28 | 2012-05-02 | 株式会社东芝 | Power supply-demand planning device and method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7368918B2 (en) * | 2006-07-27 | 2008-05-06 | Siemens Energy & Automation | Devices, systems, and methods for adaptive RF sensing in arc fault detection |
-
2012
- 2012-05-04 CN CN201210135838.2A patent/CN102662325B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1696847A (en) * | 2005-06-14 | 2005-11-16 | 上海理工大学 | Power supply management method for electronic apparatus |
| CN101754437A (en) * | 2008-12-19 | 2010-06-23 | 英特尔公司 | Handling sensors in a context aware platform |
| CN102436630A (en) * | 2010-07-28 | 2012-05-02 | 株式会社东芝 | Power supply-demand planning device and method thereof |
| CN102208028A (en) * | 2011-05-31 | 2011-10-05 | 北京航空航天大学 | Fault predicting and diagnosing method suitable for dynamic complex system |
Non-Patent Citations (1)
| Title |
|---|
| JP特开2012-34444A 2012.02.16 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102662325A (en) | 2012-09-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sedghi et al. | Distribution network expansion considering distributed generation and storage units using modified PSO algorithm | |
| CN113379110B (en) | Medium-and-long-term runoff forecast result trend testing method | |
| CN102629106B (en) | Water supply control method and water supply control system | |
| CN103164742B (en) | A kind of server performance Forecasting Methodology based on particle group optimizing neural network | |
| CN110930016A (en) | Cascade reservoir random optimization scheduling method based on deep Q learning | |
| Si et al. | Mobility prediction in cellular network using hidden markov model | |
| CN104009494B (en) | A kind of environmental economy power generation dispatching method | |
| CN102662325B (en) | Improved adaptive learning tree power supply management method | |
| CN103702401B (en) | Mobile network's power-economizing method based on user behavior analysis | |
| CN102395146B (en) | Multiple-target monitoring oriented method for sensing topology construction in wireless sensor network | |
| CN106315319A (en) | Intelligent pre-dispatching method and system for elevator | |
| CN110198042B (en) | Dynamic optimization method for power grid energy storage and storage medium | |
| CN110490429A (en) | Based on SSA algorithm intelligent building micro-capacitance sensor household loads fast dispatch method | |
| CN102855524A (en) | Carry-over storage year-end fluctuating level prediction method and system | |
| CN116805198A (en) | Method and system for power grid planning dynamics based on reinforcement learning and predictive analysis | |
| CN104732298A (en) | Method for achieving EMS load prediction based on decision tree and linear regression | |
| CN113139341A (en) | Electric quantity demand prediction method and system based on federal integrated learning | |
| CN113887809A (en) | Power distribution network supply and demand balance method, system, medium and computing equipment under double-carbon target | |
| Li et al. | A hybrid learning method for the data-driven design of linguistic dynamic systems | |
| CN110380407B (en) | Power distribution network operation optimization method considering agricultural electric irrigation and drainage loads | |
| Gastaldi et al. | Short-term forecasting of municipal load through a Kalman filtering based approach | |
| CN116963461A (en) | Energy saving method and device for machine room air conditioner | |
| CN102509177B (en) | Locally weighted linear regression projection operation method based on cloud platform | |
| CN113839795B (en) | Peak eliminating processing method and device for CDN node, electronic equipment and storage medium | |
| Chen et al. | Short-term power load model based on combined optimization of cuckoo algorithm and lightGBM |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |