CN114676632A - Method and device for predicting energy consumption of chip special for electric power and computer equipment - Google Patents
Method and device for predicting energy consumption of chip special for electric power and computer equipment Download PDFInfo
- Publication number
- CN114676632A CN114676632A CN202210318806.XA CN202210318806A CN114676632A CN 114676632 A CN114676632 A CN 114676632A CN 202210318806 A CN202210318806 A CN 202210318806A CN 114676632 A CN114676632 A CN 114676632A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- chip
- data
- consumption prediction
- electric power
- 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.)
- Pending
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 271
- 238000000034 method Methods 0.000 title claims abstract description 94
- 241000283153 Cetacea Species 0.000 claims abstract description 50
- 230000003044 adaptive effect Effects 0.000 claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 25
- 238000007781 pre-processing Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 5
- 230000003247 decreasing effect Effects 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 5
- 238000012843 least square support vector machine Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 230000005540 biological transmission Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 230000002431 foraging effect Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- YCKRFDGAMUMZLT-UHFFFAOYSA-N Fluorine atom Chemical compound [F] YCKRFDGAMUMZLT-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 229910052731 fluorine Inorganic materials 0.000 description 1
- 239000011737 fluorine Substances 0.000 description 1
- 230000019637 foraging behavior Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Primary Health Care (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting energy consumption of a chip special for electric power, which is characterized by comprising the following steps: the method comprises the following steps: determining target energy consumption data of the electric power special chip; constructing an energy consumption prediction model adaptive to the architecture of the special power chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; the expression of the nonlinear convergence factor α is: alpha-alphaint‑(αint‑αout)t2(ii) a In the formula, alphaintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; t is the predicted running time of the energy consumption prediction model; processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result; solution (II)The method solves the problem that the energy consumption of the chip special for electric power in the prior art cannot be effectively predicted by a prediction method of the energy consumption of the chip, so that an electric power enterprise cannot effectively deal with the sudden energy consumption of the chip.
Description
Technical Field
The invention belongs to the technical field of intelligent power grids, and particularly relates to a method and a device for predicting energy consumption of a chip special for electric power and computer equipment.
Background
The chip dedicated for power is widely applied to a core processor and a micro controller of a power enterprise, and is one of core hardware devices of a power system, and plays a vital processing role and a control role. The prediction of the energy consumption of the chip special for the electric power is mainly to scientifically predict the energy consumption by collecting historical energy consumption data and combining a related algorithm. In recent years, the use of novel hardware equipment causes the abnormal work of the power system, the traditional energy consumption prediction method cannot effectively predict the energy consumption of the chip, so that the power enterprise cannot effectively deal with the problem of sudden energy consumption of the chip and cannot ensure the stable operation of the power system.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the method, the device and the computer equipment for predicting the energy consumption of the chip special for the electric power are provided, so that the technical problems that the energy consumption of the chip cannot be effectively predicted by a method for predicting the energy consumption of the chip special for the electric power, an electric power enterprise cannot effectively deal with the sudden energy consumption of the chip, the stable operation of an electric power system cannot be guaranteed and the like are solved.
The technical scheme of the invention is as follows:
a method for predicting energy consumption of a chip special for electric power is characterized by comprising the following steps: the method comprises the following steps:
s1, determining target energy consumption data of the electric power special chip;
s2, constructing an energy consumption prediction model adaptive to the architecture of the special power chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; the expression of the nonlinear convergence factor α is:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; t is the predicted running time of the energy consumption prediction model;
and S3, processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
The determining the target energy consumption data of the power-dedicated chip comprises the following steps:
s11, acquiring initial energy consumption data of the special electric power chip, and processing the initial energy consumption data according to a preset preprocessing mode to obtain corresponding preprocessing data; the preprocessing mode comprises at least one of a vertical-horizontal contrast mode, a unification processing mode and a noise reduction processing mode;
and S12, determining target energy consumption data based on the obtained preprocessing data.
The power-dedicated chip comprises a three-layer architecture,
the first tier architecture includes an AXI bus; the AXI bus is provided with at least one of a dynamic memory and a digital signal processor;
a second tier architecture comprising an AHB bus connected to the AXI bus; the AHB bus is provided with at least one of an instruction interface and an interrupt controller;
the third layer architecture includes an APB bus connected to the AHB bus; at least one of a timer, a serial port, a controller and an execution module is arranged on the APB bus.
The processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result comprises the following steps:
s31, in the current iteration process, when target energy consumption data are processed according to the energy consumption prediction model, based on the optimal solution mode of whale swarm, predicting the energy consumption of the special electric power chip through the following formula to obtain a corresponding first energy consumption prediction value:
in the formula (2), f is the predicted value of the obtained energy consumption, N is the number of collected samples of the target energy consumption data, and xiAn actual energy consumption value determined according to the target energy consumption data; x is the number ofi' is a preset average predicted energy consumption value;
s32, when the next iteration is started, updating the weight of the energy consumption prediction model and redistributing the position information of the whale colony according to the maximum adaptive weight and the minimum adaptive weight of the model, and performing energy consumption prediction again based on the energy consumption prediction model obtained by corresponding updating to obtain a corresponding second energy consumption prediction value;
and S33, outputting a corresponding energy consumption prediction result when the first energy consumption prediction value is determined to be matched with the second energy consumption prediction value.
Before step S31, the method further includes:
initializing model parameters and randomly distributing whales; wherein:
the model parameters comprise at least one of the moving range of the whale colony, the number of whales, the maximum data iteration number of the model, the maximum adaptive weight and the minimum adaptive weight;
the distribution position information of each whale corresponds to a standard kernel function respectively, and the standard kernel functions are used for correcting the calculation errors of the model so as to ensure the accuracy of the predicted structure.
Step S32, adjusting the weight of the energy consumption prediction model according to the maximum adaptive weight and the minimum adaptive weight of the model, includes:
s321, according to the self-adaptive weight calculation method, adjusting the weight w of the energy consumption prediction model by the following formula:
w＝wmin+(wmax+wmin)×kμ； (3)
in the formula, wmaxIs a preset maximum adaptive weight, wminK is the adjustment ratio of the adaptive weight, and μ is the coefficient of relationship of decreasing weight.
An energy consumption prediction device of a power-dedicated chip, the device comprising a data acquisition module, a model construction module and an energy consumption prediction module, wherein:
the data acquisition module is used for determining target energy consumption data of the special power chip;
the model construction module is used for constructing an energy consumption prediction model adaptive to the architecture of the electric power special chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; wherein the expression of the nonlinear convergence factor α is:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; t is the predicted running time of the energy consumption prediction model;
and the energy consumption prediction module is used for processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
The data acquisition module is further used for acquiring initial energy consumption data of the special power chip and processing the initial energy consumption data according to a preset preprocessing mode to obtain corresponding preprocessing data; the preprocessing mode comprises at least one of a vertical-horizontal contrast mode, a unification processing mode and a noise reduction processing mode; and determining target energy consumption data based on the obtained preprocessing data.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method of any one of claims 1 to 6 when the computer program is executed by the processor.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
The invention has the beneficial effects that:
according to the method, the whale algorithm with the introduced nonlinear convergence factor is used as a basis, the energy consumption prediction model of the electric power special chip is constructed, and under the condition that the numerical value of the nonlinear convergence factor is larger and the convergence is slower, the searchable range of the energy consumption prediction model can be expanded, and the global searchability of the model is improved. And the nonlinear convergence factor is in a descending trend on the value along with the increase of the iteration times of the target energy consumption data. Compared with the prior art, the method has higher prediction accuracy and stability, can effectively reduce the difficulty of chip energy consumption prediction, and provides theoretical data support for power allocation of a power system. The method solves the technical problems that the energy consumption of the chip cannot be effectively predicted by the method for predicting the energy consumption of the special power chip in the prior art, so that power enterprises cannot effectively deal with the sudden energy consumption of the chip, the stable operation of a power system cannot be guaranteed, and the like.
Drawings
FIG. 1 is a flow chart of a power-specific chip energy consumption prediction method in an embodiment of the invention;
FIG. 2 is a flow diagram of foraging of a whale in the standing position in an embodiment of the invention;
FIG. 3 is a diagram of the subject architecture of a power specific chip according to an embodiment of the present invention;
FIG. 4 is a detailed flow chart illustrating energy consumption prediction for a power-specific chip according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating comparison between the prediction accuracy of the prediction method of the present application and the prediction accuracy of the conventional task scheduling-based power-specific chip energy consumption prediction method and the least square support vector machine-based power-specific chip energy consumption prediction method in an embodiment of the present invention;
FIG. 6 is a schematic diagram comparing the prediction error rate of the prediction method of the present application with the conventional task scheduling-based power-specific chip energy consumption prediction and the least-squares support vector machine-based power-specific chip energy consumption prediction method in an embodiment of the present invention;
FIG. 7 is a schematic diagram illustrating comparison between the prediction method of the present application and the prediction stability of the conventional task scheduling-based power-specific chip energy consumption prediction method and the least square support vector machine-based power-specific chip energy consumption prediction method in an embodiment of the present invention;
fig. 8 is a system configuration diagram of an energy consumption prediction apparatus of a power-dedicated chip according to an embodiment of the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
In one or more embodiments of the present invention, as shown in fig. 1, a method for predicting energy consumption of a power-specific chip is provided, which is described by taking as an example that the method is applied to a computer device (the computer device may specifically be a terminal or a server, and the terminal may specifically be, but is not limited to, various personal computers, laptops, smartphones, tablet computers, and portable wearable devices).
And S1, determining the target energy consumption data of the power special chip.
Specifically, the computer device is connected to the energy consumption data acquisition device and acquires initial energy consumption data of the power dedicated chip. Due to the influence of the working environment of the electric power special chip and the energy consumption data acquisition equipment, the initial energy consumption data often has the problems of data loss, noise interference and the like, so that the workload of energy consumption prediction can be increased, and the phenomenon of inaccurate prediction result is easily caused. Therefore, before the energy consumption data acquired by the computer device is processed by using the subsequently constructed energy consumption prediction model, the initial energy consumption data needs to be preprocessed to avoid the above problems. It should be noted that the preprocessing method adopted by the computer device may be flexibly set, for example, in an embodiment, when it is determined that noise interference exists, a moving average algorithm is adopted to perform noise reduction processing on the energy consumption data, and the preprocessing method is not limited in the embodiment of the present application.
In one embodiment, the target energy consumption data includes various data indexes capable of reflecting energy consumption, such as power consumption per unit time, and the like, which is not limited in this embodiment.
S2, constructing an energy consumption prediction model adaptive to the architecture of the special power chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; wherein, the expression of the nonlinear convergence factor alpha is as follows:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; and t is the predicted running time of the energy consumption prediction model.
In one aspect, it is noted that the whale algorithm is a search algorithm which takes the foraging mode of the whale as a prototype, wherein the foraging behavior of the whale comprises prey circling, prey implementation, random prey and the like. The included algorithm flow comprises the following steps: the present invention relates to a method for determining the optimal prey for a target prey group, and more particularly, to a method for determining the optimal prey for a target prey group by using a sitter whale group, which can be understood by referring to fig. 2 without being described in more detail in the embodiments of the present invention.
However, the traditional whale algorithm has certain limitation in the process of finding the optimal solution, and the optimal solution finding process cannot be presented well. Therefore, on the basis of the traditional whale algorithm, aiming at the structural characteristics of the electric power special chip, the embodiment of the application introduces the nonlinear convergence factor alpha to improve the traditional whale algorithm, constructs an energy consumption prediction model based on the improved whale algorithm, and sequentially realizes the optimal solution for predicting the energy consumption of the electric power special chip.
On the other hand, when the initial power-dedicated chip energy consumption is predicted, the larger the value of the nonlinear convergence factor α is, the slower the model converges, the wider the searchable range of the prediction model is, and the stronger the global search performance is. And with the increase of the iteration times of the energy consumption data, the convergence factor alpha is in a descending trend on the numerical value, and accordingly, the prediction efficiency of the model is higher.
And S3, processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
Specifically, in the corresponding iteration process, the computer device may directly output the energy consumption prediction result obtained in the current iteration process. Alternatively, to ensure the accuracy of the output result, the computer device may further output the result while ensuring the accuracy of the energy consumption prediction result obtained in the current iteration process based on the energy consumption prediction results obtained in the current iteration and the previous iteration. Specific implementation, reference may be made to fig. 3, and since fig. 3 will be described later, the embodiments of the present application are not described more herein.
According to the energy consumption prediction method for the electric power special chip, the energy consumption prediction model of the electric power special chip is constructed on the basis of the whale algorithm introducing the nonlinear convergence factor, and under the condition that the numerical value of the nonlinear convergence factor is larger and the convergence is slower, the searchable range of the energy consumption prediction model can be expanded, and the global searchability of the model is improved. And the nonlinear convergence factor is in a descending trend on the value along with the increase of the iteration times of the target energy consumption data. Compared with the prior art, the method has higher prediction accuracy and stability, can effectively reduce the difficulty of chip energy consumption prediction, and provides theoretical data support for power allocation of a power system.
In one or more embodiments of the invention, the determining the target energy consumption data of the power-dedicated chip in step S1 includes:
s11, acquiring initial energy consumption data of the special electric power chip, and processing the initial energy consumption data according to a preset preprocessing mode to obtain corresponding preprocessing data; the preprocessing method includes at least one of an aspect contrast method, a normalization processing method, and a noise reduction processing method.
Specifically, the computer device eliminates redundant data in the initial energy consumption data by means of cross-contrast, and selectively performs data unification after eliminating the redundant data, and in one embodiment, the data unification may be performed with reference to the following formula:
wherein h isi' denotes initial energy consumption data h after elimination of redundant dataiCarrying out unification treatment and then obtaining a corresponding numerical value; h ismaxAs initial energy consumption data hiThe maximum value that can be taken; h isminAs initial energy consumption data hiThe minimum value that can be taken.
Based on the foregoing embodiment, after the normalization processing is completed, in consideration of the problem of noise interference, in the current embodiment, the computer device uses a moving average algorithm to perform the normalization processing on the energy consumption data hi' noise reduction processing is performed by taking the currently processed data as the center, calculating the arithmetic mean value before and after the data, and obtaining the mean value as the arithmetic mean value of the currently processed dataAnd (4) average value.
And S12, determining target energy consumption data based on the obtained preprocessing data.
Specifically, in the current embodiment, the computer device uses the preprocessed data obtained in step S11 as the target energy consumption data, inputs the target energy consumption data into the constructed energy consumption prediction model, and processes the target energy consumption data by the energy consumption prediction model, so as to predict the energy consumption of the power-dedicated chip.
In the embodiment, the acquired energy consumption data is preprocessed before energy consumption prediction, so that the problems of noise interference and data loss are avoided, the accuracy of the prediction result is improved, and the application requirements of power enterprises are met.
In one or more embodiments of the present invention, in step S2, the power-dedicated chip includes a three-layer architecture, where: the first tier architecture includes an AXI bus; the AXI bus is provided with at least one of a dynamic memory and a digital signal processor; the second layer architecture includes an AHB bus connected to an AXI bus; the AHB bus is provided with at least one of an instruction interface and an interrupt controller; the third layer architecture includes an APB bus connected to the AHB bus; at least one of a timer, a serial port, a controller and an execution module is arranged on the APB bus.
Specifically, please refer to fig. 4, wherein:
(1) the architecture layer of the power-dedicated chip is an AXI bus, which includes two sets of dynamic memory eDRAN0 and core processing modules such as a DSP (i.e., a digital signal processor), wherein each core processing module has a separate instruction interface and a separate data transmission interface. It should be noted that the above-mentioned command interface is used for internal transmission of the chip, and the data transmission interface may be connected to an external device (e.g., an energy data acquisition device) for data sharing.
(2) The second layer of the structure of the power special chip is an AHB bus which is connected with the first layer of main line and is constructed by adopting a matrix structure and a full communication mode. The Memory comprises 4 instruction interfaces, wherein the instruction interfaces are responsible for processing and analyzing instructions from the first layer architecture, support custom program compilation and can read and write the instructions of the first layer architecture in a Memory mode. Meanwhile, the chip also comprises an interrupt controller, and when the chip is high in energy consumption, unstable in operation or has a fault, the interrupt controller can be started to stop the operation of the chip so as to ensure that other structures of the chip are not damaged.
(3) The three layers of the architecture of the special power chip are APB buses, which comprise a plurality of components such as a timer, a serial port, a controller, an execution module and the like, are important data processing modules, and are compatible with a data communication link. A user can access the ethernet to acquire data, and in order to ensure data security, in the current embodiment, mechanisms such as a watchdog and a firewall are used to perform malicious attack defense.
In one or more embodiments of the present invention, in step S3, the processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result includes:
s31, in the current iteration process, when target energy consumption data are processed according to the energy consumption prediction model, based on the optimal solution mode of whale swarm, predicting the energy consumption of the special electric power chip through the following formula to obtain a corresponding first energy consumption prediction value:
in the formula (2), f is the energy consumption predicted value, N is the number of collected samples of the target energy consumption data, and xiAn actual energy consumption value determined according to the target energy consumption data; x is the number ofi' is a preset average predicted energy consumption value.
And S32, when the next iteration is started, updating the weight of the energy consumption prediction model and redistributing the position information of the whale colony according to the maximum adaptive weight and the minimum adaptive weight of the model, and performing energy consumption prediction again based on the energy consumption prediction model obtained by corresponding updating to obtain a corresponding second energy consumption prediction value.
Specifically, the selection of the parameter α determines, to some extent, the predictive performance of the energy consumption prediction model. In the current embodiment, the influence of whale location vectors in an improved whale algorithm on a prediction result is considered, and the prediction weight of the model is calculated by adopting a self-adaptive weight calculation method, so that model parameters are adjusted in data iteration, and the flexibility of energy consumption prediction of the electric power special chip is improved.
And S33, outputting a corresponding energy consumption prediction result when the first energy consumption prediction value is determined to be matched with the second energy consumption prediction value.
Specifically, referring to fig. 3, the computer device compares the first energy consumption prediction value with the second energy consumption prediction value, and outputs the first energy consumption prediction value or the second energy consumption prediction value as a prediction result when the first energy consumption prediction value and the second energy consumption prediction value are determined to be equal to each other. Otherwise, the process returns to step S31 to resume the energy consumption prediction.
In one embodiment, before outputting the prediction result, the computer device further determines whether the first energy consumption prediction value or the second energy consumption prediction value meets a preset output format, and if so, directly outputs the original data; otherwise, the corresponding energy consumption prediction value is subjected to numerical value conversion, and the converted numerical value obtained through conversion is used as a prediction result to be output.
In one or more embodiments of the invention, before step S31, the method further includes: initializing model parameters and randomly distributing whales; wherein: the model parameters comprise at least one of the moving range of the whale colony, the number of whales, the maximum data iteration number of the model, the maximum adaptive weight and the minimum adaptive weight; the distribution position information of each whale corresponds to a standard kernel function respectively, and the standard kernel functions are used for correcting the calculation errors of the model so as to ensure the accuracy of the predicted structure.
In one or more embodiments of the invention, step S32, adjusting the weight of the energy consumption prediction model according to the maximum adaptive weight and the minimum adaptive weight of the model includes:
s321, according to the self-adaptive weight calculation method, adjusting the weight w of the energy consumption prediction model by the following formula:
w＝wmin+(wmax+wmin)×kμ； (3)
in the formula, wmaxIs a preset maximum adaptive weight, wminK is the adjustment ratio of the adaptive weight, and μ is the coefficient of relationship of decreasing weight.
It should be noted that the value of μ represents a decreasing trend of the adaptive weight, and the larger the value of μ is, the larger the decreasing degree of the weight w is, and when μ approaches to 0, it indicates that the weight w has no significant change.
In one embodiment, in order to verify the actual prediction performance of the method, a comparison experiment is performed, energy consumption data of a certain power-dedicated chip in a week is used as an experiment data sample set in an experiment environment, the prediction method of the application is selected to perform experiment comparison with a traditional power-dedicated chip energy consumption prediction method based on task scheduling and a power-dedicated chip energy consumption prediction method based on a least square support vector machine, and set experiment parameters are as shown in table 1 below.
TABLE 1 Experimental parameters
| Item | Parameter(s) |
| Data transmission interface | IO |
| Command interface | SOI |
| Chip model special for electric power | AMD tachylon X4860k |
| Operating system | Windows |
| Protocol stack | Z-Stack |
| Maximum data acquisition in a single pass | 10G |
| Control module | CNA |
According to the designed experimental parameters, the prediction method and the traditional prediction method are selected to predict the energy consumption of the power chip, the prediction interval is 0.5h, and in order to ensure the authenticity and reliability of the experimental result, the energy consumption prediction precision standard commonly used by the national power grid is selected: the prediction accuracy and the prediction error rate are the basis for measuring the prediction accuracy of the three prediction methods, wherein the calculation modes of the prediction accuracy and the prediction error rate are as follows:
in the formula, EMAPEIndicating the accuracy of the prediction, EFAIndicating the prediction error rate, xiActual energy consumption values, x, determined for the target energy consumption datai' is a preset average predicted energy consumption value, and N is the number of collected samples of the target energy consumption data. The comparison results of the prediction accuracy and the prediction error rate of the three prediction methods are shown in fig. 5 and 6. As can be seen from the above two figures, as the number of data iterations increases, the accuracy of the three prediction methods decreases, and at the same time, the error rate begins to increase. Wherein the accuracy rate of the prediction method based on task scheduling has obvious descending trend, the lowest accuracy rate is 51 percent,the average accuracy was 75% and the average error rate was 23%. The accuracy rate of the prediction method based on the least square support vector machine is relatively smooth in descending trend, the minimum accuracy rate is 69%, the average accuracy rate is 86%, and the average error rate is 15%. The accuracy of the prediction method based on the improved whale algorithm has no obvious downward-sliding trend, the general accuracy is higher than 93%, the error rate is lower than that, the average accuracy of 10% is 96%, the average error rate is 6%, and the prediction accuracy is far higher than that of the other two traditional prediction methods.
In the current embodiment, in order to further verify the stability of the three power-dedicated chip energy consumption prediction methods, a time-series description method is used to describe the data fluctuation states of the three prediction methods in the energy consumption prediction process, wherein the faster the data fluctuation frequency is, the larger the amplitude transformation range is, the worse the stability of the prediction method is, and the comparison results of the prediction stability of the three prediction methods obtained can be referred to fig. 7. As can be seen from the above figure, within 175 hours of observation, the data fluctuation of the prediction method based on task scheduling is severe, the fluctuation transformation has no regularity, and the data fluctuation amplitude transformation range is +/-3. As can be seen from FIG. 7, the least square support vector machine-based prediction method has high data fluctuation frequency at the initial stage and is relatively gentle at the later stage, but the overall amplitude has no obvious change and is always maintained within the variation range of +/-2. In comparison, the prediction method based on the improved whale algorithm has regular data fluctuation, consistent transformation frequency, small overall fluctuation and a data fluctuation amplitude transformation range of +/-1.
In summary, the prediction method for the energy consumption of the special electric power chip based on the improved whale algorithm has high prediction accuracy and stability, and can accurately predict the energy consumption of the special electric power chip, while the traditional prediction method cannot properly process acquired energy consumption data, so that the prediction result accuracy is low, the flexibility of the adopted calculation mode is low, the calculation process is complex, the overall stability is affected, and therefore a conclusion can be drawn.
Referring to fig. 8, the present application discloses an apparatus 800 for predicting energy consumption of a chip dedicated for electric power, the apparatus 800 includes a data obtaining module 801, a model building module 802, and an energy consumption predicting module 803, wherein:
a data obtaining module 801, configured to determine target energy consumption data of the power-dedicated chip;
the model construction module 802 is configured to construct an energy consumption prediction model adapted to the architecture of the power-dedicated chip based on a whale algorithm with a nonlinear convergence factor α introduced; wherein, the expression of the nonlinear convergence factor alpha is as follows:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha (alpha) ("alpha")outIs a preset final value of α; t is the predicted running time of the energy consumption prediction model;
and the energy consumption prediction module 803 is configured to process the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
In one embodiment, the data acquisition module is further configured to acquire initial energy consumption data of the power dedicated chip, and process the initial energy consumption data according to a preset preprocessing mode to obtain corresponding preprocessed data; the preprocessing mode comprises at least one of a vertical and horizontal contrast mode, a unification processing mode and a noise reduction processing mode; and determining target energy consumption data based on the obtained preprocessing data.
In one embodiment, the power-specific chip comprises a three-tier architecture, wherein: the first tier architecture includes an AXI bus; the AXI bus is provided with at least one of a dynamic memory and a digital signal processor; the second tier architecture includes an AHB bus connected to an AXI bus; the AHB bus is provided with at least one of an instruction interface and an interrupt controller; the third layer architecture includes an APB bus connected to the AHB bus; at least one of a timer, a serial port, a controller and an execution module is arranged on the APB bus.
In one embodiment, the energy consumption predicting module 803 is further configured to predict the energy consumption of the electric power special chip according to the following formula based on an optimal solution searching manner of the whale swarm when the target energy consumption data is processed according to the energy consumption predicting model in the current iteration process, so as to obtain a corresponding first energy consumption predicted value:
in the formula (2), f is the predicted value of the obtained energy consumption, N is the number of collected samples of the target energy consumption data, and xiAn actual energy consumption value determined according to the target energy consumption data; x is a radical of a fluorine atomi' is a preset average predicted energy consumption value; when the next iteration is started, updating the weight of the energy consumption prediction model and redistributing the position information of the whale colony according to the maximum adaptive weight and the minimum adaptive weight of the model, and performing energy consumption prediction again on the basis of the energy consumption prediction model obtained by corresponding updating to obtain a corresponding second energy consumption prediction value; and outputting a corresponding energy consumption prediction result when the first energy consumption prediction value is matched with the second energy consumption prediction value.
In one embodiment, the apparatus 800 further comprises an initialization module, wherein:
the initialization module is used for initializing model parameters and randomly distributing whales; wherein: the model parameters comprise at least one of the moving range of the whale colony, the number of whales, the maximum data iteration number of the model, the maximum adaptive weight and the minimum adaptive weight; the distribution position information of each whale corresponds to a standard kernel function respectively, and the standard kernel functions are used for correcting the calculation errors of the model so as to ensure the accuracy of the predicted structure.
In one embodiment, the energy consumption prediction module 803 is further configured to adjust the weight w of the energy consumption prediction model according to an adaptive weight calculation method by the following formula:
w＝wmin+(wmax+wmin)×kμ； (3)
in the formula, wmaxIs a preset maximum adaptive weight, wminIs a preset minimum adaptive weight, k is the adjustment ratio of the adaptive weight, mu is the weightThe relationship coefficient of the heavy decline.
According to the energy consumption prediction device for the power special chip, the energy consumption prediction model of the power special chip is constructed on the basis of the whale algorithm introducing the nonlinear convergence factor, and under the condition that the numerical value of the nonlinear convergence factor is larger and the convergence is slower, the searchable range of the energy consumption prediction model can be expanded, and the global searchability of the model is improved. And the nonlinear convergence factor is in a descending trend on the value along with the increase of the iteration times of the target energy consumption data. Compared with the prior art, the method has higher prediction accuracy and stability, can effectively reduce the difficulty of chip energy consumption prediction, and provides theoretical data support for power allocation of a power system.
In one or more embodiments of the present invention, there is also provided a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.
According to the computer equipment, the whale algorithm introducing the nonlinear convergence factor is used as a basis, the energy consumption prediction model of the power special chip is constructed, the searchable range of the energy consumption prediction model can be expanded under the condition that the numerical value of the nonlinear convergence factor is larger and the convergence is slower, and the global searchability of the model is improved. And the nonlinear convergence factor is in a descending trend in numerical value along with the increase of the iteration times of the target energy consumption data. Compared with the prior art, the method has higher prediction accuracy and stability, can effectively reduce the difficulty of chip energy consumption prediction, and provides theoretical data support for power allocation of a power system.
In one or more embodiments of the invention, a computer-readable storage medium is provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
The storage medium constructs the energy consumption prediction model of the chip special for electric power on the basis of a whale algorithm introducing the nonlinear convergence factor, and can expand the searchable range of the energy consumption prediction model and improve the global searchability of the model under the condition that the larger the numerical value of the nonlinear convergence factor is, the slower the convergence is. And the nonlinear convergence factor is in a descending trend on the value along with the increase of the iteration times of the target energy consumption data. Compared with the prior art, the method has higher prediction accuracy and stability, can effectively reduce the difficulty of chip energy consumption prediction, and provides theoretical data support for power allocation of a power system.
Claims (10)
1. A method for predicting energy consumption of a chip special for electric power is characterized by comprising the following steps: the method comprises the following steps:
s1, determining target energy consumption data of the electric power special chip;
s2, constructing an energy consumption prediction model adaptive to the architecture of the special power chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; the expression of the nonlinear convergence factor α is:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; t is the predicted running time of the energy consumption prediction model;
and S3, processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
2. The method for predicting the energy consumption of the electric power special chip according to claim 1, wherein: the determining the target energy consumption data of the power-dedicated chip comprises the following steps:
s11, acquiring initial energy consumption data of the special electric power chip, and processing the initial energy consumption data according to a preset preprocessing mode to obtain corresponding preprocessing data; the preprocessing mode comprises at least one of a vertical and horizontal contrast mode, a unification processing mode and a noise reduction processing mode;
and S12, determining target energy consumption data based on the obtained preprocessing data.
3. The method for predicting the energy consumption of the electric power special chip as claimed in claim 1, wherein: the power-dedicated chip comprises a three-layer architecture,
the first tier architecture includes an AXI bus; the AXI bus is provided with at least one of a dynamic memory and a digital signal processor;
a second tier architecture comprising an AHB bus connected to the AXI bus; the AHB bus is provided with at least one of an instruction interface and an interrupt controller;
the third layer architecture includes an APB bus connected to the AHB bus; at least one of a timer, a serial port, a controller and an execution module is arranged on the APB bus.
4. The method for predicting the energy consumption of the electric power special chip as claimed in claim 1, wherein: the processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result comprises the following steps:
s31, in the current iteration process, when target energy consumption data are processed according to the energy consumption prediction model, based on the optimal solution mode of whale swarm, predicting the energy consumption of the special electric power chip through the following formula to obtain a corresponding first energy consumption prediction value:
in the formula (2), f is the predicted value of the obtained energy consumption, N is the number of collected samples of the target energy consumption data, and xiAn actual energy consumption value determined according to the target energy consumption data; x is the number ofi' is a preset average predicted energy consumption value;
s32, when the next iteration is started, updating the weight of the energy consumption prediction model and redistributing the position information of the whale colony according to the maximum adaptive weight and the minimum adaptive weight of the model, and performing energy consumption prediction again based on the energy consumption prediction model obtained by corresponding updating to obtain a corresponding second energy consumption prediction value;
and S33, outputting a corresponding energy consumption prediction result when the first energy consumption prediction value is determined to be matched with the second energy consumption prediction value.
5. The method for predicting the energy consumption of the electric power special chip as claimed in claim 4, wherein: before step S31, the method further includes:
initializing model parameters and randomly distributing whales; wherein:
the model parameters comprise at least one of the moving range of the whale colony, the number of whales, the maximum data iteration number of the model, the maximum adaptive weight and the minimum adaptive weight;
the distribution position information of each whale corresponds to a standard kernel function respectively, and the standard kernel functions are used for correcting the calculation errors of the model so as to ensure the accuracy of the predicted structure.
6. The method for predicting the energy consumption of the electric power special chip as claimed in claim 4, wherein: step S32, adjusting the weight of the energy consumption prediction model according to the maximum adaptive weight and the minimum adaptive weight of the model, includes:
s321, according to the self-adaptive weight calculation method, adjusting the weight w of the energy consumption prediction model by the following formula:
w＝wmin+(wmax+wmin)×kμ； (3)
in the formula, wmaxIs a preset maximum adaptive weight, wminK is the adjustment ratio of the adaptive weight, and μ is the coefficient of relationship of decreasing weight.
7. The utility model provides a special chip energy consumption prediction device of electric power which characterized in that: the device comprises a data acquisition module, a model construction module and an energy consumption prediction module, wherein:
the data acquisition module is used for determining target energy consumption data of the special power chip;
the model construction module is used for constructing an energy consumption prediction model adaptive to the architecture of the electric power special chip on the basis of a whale algorithm introducing a nonlinear convergence factor alpha; wherein the expression of the nonlinear convergence factor α is:
α＝αint-(αint-αout)t2； (1)
in the formula (1), αintIs a preset initial value of alpha; alpha is alphaoutIs a preset final value of α; t is the predicted running time of the energy consumption prediction model;
and the energy consumption prediction module is used for processing the target energy consumption data according to the energy consumption prediction model to obtain a corresponding energy consumption prediction result.
8. The device according to claim 7, wherein the data obtaining module is further configured to obtain initial energy consumption data of the power-dedicated chip, and process the initial energy consumption data according to a preset preprocessing manner to obtain corresponding preprocessed data; the preprocessing mode comprises at least one of a vertical and horizontal contrast mode, a unification processing mode and a noise reduction processing mode; and determining target energy consumption data based on the obtained preprocessing data.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the steps of the method of any one of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210318806.XA CN114676632A (en) | 2022-03-29 | 2022-03-29 | Method and device for predicting energy consumption of chip special for electric power and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210318806.XA CN114676632A (en) | 2022-03-29 | 2022-03-29 | Method and device for predicting energy consumption of chip special for electric power and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114676632A true CN114676632A (en) | 2022-06-28 |
Family
ID=82076923
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210318806.XA Pending CN114676632A (en) | 2022-03-29 | 2022-03-29 | Method and device for predicting energy consumption of chip special for electric power and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114676632A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115616374A (en) * | 2022-09-20 | 2023-01-17 | 重庆鹰谷光电股份有限公司 | Machine learning-based semiconductor chip test system |
-
2022
- 2022-03-29 CN CN202210318806.XA patent/CN114676632A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115616374A (en) * | 2022-09-20 | 2023-01-17 | 重庆鹰谷光电股份有限公司 | Machine learning-based semiconductor chip test system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA3069185C (en) | Operation accelerator | |
| Nabavinejad et al. | Coordinated batching and DVFS for DNN inference on GPU accelerators | |
| JP2022501675A (en) | Data processing methods, devices, computer devices, and storage media | |
| KR20190091858A (en) | Heterogenous Processor Architecture to Integrate CNN and RNN Neural Networks on a Single Chip | |
| US9239734B2 (en) | Scheduling method and system, computing grid, and corresponding computer-program product | |
| CN104216783A (en) | Method for automatically managing and controlling virtual GPU (Graphics Processing Unit) resource in cloud gaming | |
| CN111105023A (en) | Data stream reconstruction method and reconfigurable data stream processor | |
| CN114676632A (en) | Method and device for predicting energy consumption of chip special for electric power and computer equipment | |
| Phan et al. | On understanding the energy impact of speculative execution in hadoop | |
| CN110850957A (en) | Scheduling method for reducing system power consumption through dormancy in edge computing scene | |
| CN116360921A (en) | Cloud platform resource optimal scheduling method and system for electric power Internet of things | |
| US11308349B1 (en) | Method to modify adaptive filter weights in a decentralized wireless sensor network | |
| CN110415162B (en) | Adaptive graph partitioning method facing heterogeneous fusion processor in big data | |
| Jiang et al. | An optimized resource scheduling strategy for Hadoop speculative execution based on non-cooperative game schemes | |
| CN110766133B (en) | Data processing method, device, equipment and storage medium in embedded equipment | |
| CN116185942A (en) | Data processing method, device, storage medium and electronic equipment | |
| CN104077398B (en) | Job assignment system and method based on environment of gathering together Hadoop more | |
| Berned et al. | Combining thread throttling and mapping to optimize the edp of parallel applications | |
| CN114069666B (en) | Control method and system for energy storage power instruction in wind storage system | |
| Gao et al. | Revisiting thread configuration of SpMV kernels on GPU: A machine learning based approach | |
| CN117971509B (en) | Heterogeneous computing power cluster operation performance optimization method, heterogeneous computing power cluster operation performance optimization device, heterogeneous computing power cluster operation performance optimization equipment and medium | |
| Ray et al. | Proposing ε-greedy Reinforcement Learning Technique to Self-Optimize Memory Controllers | |
| Liu et al. | Energy-aware automatic tuning on many-core platform via adaptive evolution | |
| Liu et al. | Enabling Energy-Efficient and Reliable Neural Network via Neuron-Level Voltage Scaling | |
| CN109753134B (en) | Global decoupling-based GPU internal energy consumption control system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |