A method and device for managing power consumption

Abstract

The present invention discloses a method and device for managing power consumption, which belongs to the technical field of power consumption management and aims to optimize the energy consumption of power supply. The method comprises first collecting a normal power consumption sequence, an abnormal power consumption sequence and a power operation parameter adjustment strategy made by a staff member for anomalies corresponding to the abnormal power consumption sequence, and then learning the normal power consumption sequence and the abnormal power consumption sequence, so as to realize automatic detection rules for power consumption, and when abnormal power consumption is found, automatically executing the power operation parameter adjustment strategy corresponding to the abnormal power consumption sequence, so as to achieve the purpose of energy saving.

Description

Technical Field

[0001] The present invention belongs to the technical field of power supply energy consumption management, and in particular relates to a power supply energy consumption management method and device. Background Art

[0002] With the development of social economy, the demand for electricity is growing, and the problem of energy consumption is becoming more and more serious. Among all kinds of electrical equipment, the management of power consumption is particularly important. The traditional power consumption management method has the following shortcomings: the power consumption data cannot be monitored in real time, resulting in energy waste; the power distribution strategy is unreasonable and cannot achieve optimal energy utilization; there is a lack of effective energy consumption analysis methods, which cannot provide strong support for energy conservation. Summary of the invention

[0003] The present invention provides a power supply energy consumption management method and device, which are used to solve the problem that the prior art lacks effective energy consumption analysis means, cannot provide strong support for energy saving, and causes energy waste.

[0004] In one aspect, the present invention provides a method for managing power consumption, comprising:

[0005] Collecting historical data on power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the abnormal power consumption sequence;

[0006] A deep learning algorithm is used to learn the normal power consumption sequence and the abnormal power consumption sequence to determine the abnormal power consumption identification rules;

[0007] Collecting a real-time power consumption sequence about power consumption, and using the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequence to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal;

[0008] When the power consumption identification result is abnormal, the power consumption management strategy is determined according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormality corresponding to the power consumption abnormality sequence, and the power consumption management is completed.

[0009] Furthermore, the historical data of power consumption is data input by human-computer interaction or data pre-stored in a database.

[0010] Furthermore, a deep learning algorithm is used to learn the normal power consumption sequence and the abnormal power consumption sequence to determine the abnormal power consumption identification rules, including:

[0011] Initialize algorithm parameters corresponding to the deep learning algorithm to determine multiple algorithm parameter vectors for optimizing the deep learning algorithm; wherein each algorithm parameter vector includes all algorithm parameters to be optimized of the deep learning algorithm;

[0012] Construct an objective function for evaluating the performance of a deep learning algorithm, and use a normal power consumption sequence and an abnormal power consumption sequence as supporting data to evaluate the objective function value corresponding to each algorithm parameter vector when applied to the deep learning algorithm according to the objective function;

[0013] According to the objective function value corresponding to each algorithm parameter vector, determine the optimal parameter vector and the worst parameter vector;

[0014] Based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy carrying position influence to obtain an updated algorithm parameter vector;

[0015] Performing position fusion on the algorithm parameter vector after the first update to determine the reference vector, and using the reference vector as a basis, adopting a disturbance guidance strategy to update the algorithm parameter vector after the first update to obtain the algorithm parameter vector after the second update;

[0016] Based on the worst parameter vector, the algorithm parameter vector after the second update is updated by adopting a global guidance strategy to obtain the algorithm parameter vector after the third update;

[0017] Determine whether the iteration end condition is met. If so, determine the power consumption abnormality identification rule based on the algorithm parameter vector after the three updates, otherwise return to the step of determining the optimal parameter vector and the worst parameter vector.

[0018] Furthermore, the objective function for evaluating the performance of the deep learning algorithm is constructed as:

[0019]

[0020] in, f Represents the objective function, and the larger the objective function, the better. Indicates i After the power consumption sequence is input into the deep learning algorithm, the actual output obtained is j elements, Indicates i The expected output corresponding to the power consumption sequence j elements, I represents the total number of batch input power consumption sequences, J represents the total number of outputs of the deep learning algorithm, and the power consumption sequence is used to characterize a normal power consumption sequence or an abnormal power consumption sequence. Represents a preset constant term.

[0021] Furthermore, based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy with position influence, and an updated algorithm parameter vector is obtained, including:

[0022]

[0023]

[0024] in, Indicates the t-1th optimization. k The algorithm parameter vector l Algorithm parameters, k =1,2,…,K, K represents the total number of algorithm parameter vectors, l =1,2,…,L, where L represents the total number of algorithm parameters in the algorithm parameter vector, Indicates the first k The algorithm parameter vector l Algorithm parameters, represents the first adaptive weight, represents the second adaptive weight, Indicates the tth optimization k The algorithm parameter vector l The speed corresponding to the algorithm parameters is Indicates the t-1th optimization. k The algorithm parameter vector l The speed corresponding to the algorithm parameters is represents the first learning factor, represents the second learning factor, Indicates k The development frequency corresponding to the algorithm parameter vector, represents the optimal parameter vector in the t-1th optimization. l Algorithm parameters, Indicates k Algorithm parameter vector The corresponding historical optimal value l algorithm parameters.

[0025] Furthermore, the first adaptive weight and the second adaptive weight for:

[0026]

[0027]

[0028] in, represents the minimum weight value, represents the maximum weight, and T represents the maximum number of optimizations;

[0029] The said k The development frequency corresponding to the algorithm parameter vector for:

[0030]

[0031] in, Indicates the initial value of the development frequency, Indicates the frequency correction factor.

[0032] Further, the algorithm parameter vector after the first update is subjected to position fusion, a reference vector is determined, and based on the reference vector, a disturbance guidance strategy is adopted to update the algorithm parameter vector after the first update, so as to obtain the algorithm parameter vector after the second update, including:

[0033] Perform position fusion on the algorithm parameter vector after an update and determine the reference vector as:

[0034]

[0035] in, represents the reference vector at the t-1th optimization time, Indicates m The algorithm parameter vector after one update, Represents the algorithm parameter vector The corresponding objective function value;

[0036] Based on the reference vector, the perturbation guidance strategy is used to update the algorithm parameter vector after one update as follows:

[0037]

[0038] in, Represents the algorithm parameter vector after the second update , Represents a random number between (0,1).

[0039] Furthermore, based on the worst parameter vector, the algorithm parameter vector after the second update is updated using a global guidance strategy, and the algorithm parameter vector after the third update is obtained as follows:

[0040]

[0041] in, Indicates the t-1th optimization. n The algorithm parameter vector after the second update l Algorithm parameters, Indicates n The algorithm parameter vector after three updates is l Algorithm parameters, Represents a random number between (0,1), Represents algorithm parameters The corresponding adjustment factor is represents the worst parameter vector l Algorithm parameters, Indicates n The algorithm parameter vector after the second update The corresponding objective function value is represents the worst parameter vector The corresponding objective function value.

[0042] Furthermore, the algorithm parameters The corresponding adjustment factor for:

[0043]

[0044]

[0045] Among them, e represents a natural constant, represents the intermediate parameter, Represents the objective function value corresponding to the optimal parameter vector.

[0046] On the other hand, the present invention provides a power consumption management device, including: a data acquisition module, a deep learning module, an identification and analysis module, and a power consumption management module;

[0047] The data collection module is used to collect historical data about power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the abnormal power consumption sequence;

[0048] The deep learning module is used to learn the normal power consumption sequence and the abnormal power consumption sequence using a deep learning algorithm to determine the abnormal power consumption recognition rules;

[0049] The identification and analysis module is used to collect real-time power consumption sequences about power consumption, and use the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequences to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal;

[0050] The power consumption management module is used to determine the power consumption management strategy and complete the power consumption management according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the power consumption abnormality sequence when the power consumption identification result is abnormal.

[0051] A power supply energy consumption management method and device provided by the present invention are intended to optimize the energy consumption of a power supply. The method includes first collecting a normal power supply energy consumption sequence, an abnormal power supply energy consumption sequence, and a power supply operation parameter adjustment strategy made by a staff member for anomalies corresponding to the abnormal power supply energy consumption sequence, and then learning the normal power supply energy consumption sequence and the abnormal power supply energy consumption sequence, so as to realize automatic detection rules for power supply energy consumption, and when an abnormal power supply energy consumption is found, automatically executing the power supply operation parameter adjustment strategy corresponding to the abnormal power supply energy consumption sequence, thereby achieving the purpose of saving energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0053] Figure 1 The present invention is a flowchart of a method for managing power consumption according to an embodiment of the present invention.

[0054] Figure 2 The present invention is a schematic diagram of the structure of a power consumption management device according to an embodiment of the present invention.

[0055] Among them, 201-data acquisition module, 202-deep learning module, 203-identification and analysis module, 204-power consumption management module.

[0056] The above drawings have shown clear embodiments of the present invention, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present invention in any way, but to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0057] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0058] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0059] like Figure 1 As shown, an embodiment of the present invention provides a method for managing power consumption, including:

[0060] S101, collecting historical data on power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by a staff member for anomalies corresponding to the abnormal power consumption sequence;

[0061] The power operation type corresponding to the normal power consumption sequence is normal. Similarly, the power operation type corresponding to the abnormal power consumption sequence is abnormal. When the power operation type is abnormal at a historical moment, staff will often adjust the power operation parameters, that is, the historical power operation parameter adjustment strategy is retained. Therefore, based on this, abnormal power consumption can be automatically identified and managed, thereby saving energy consumption.

[0062] S102, using a deep learning algorithm to learn a normal power consumption sequence and an abnormal power consumption sequence to determine a power consumption abnormality identification rule;

[0063] You can choose BP deep learning algorithm or LSTM deep learning algorithm to learn the normal power consumption sequence and abnormal power consumption sequence, so as to identify the abnormal power consumption rules. In the process of data learning, you can use intelligent optimization algorithms such as genetic algorithms to optimize the algorithm parameters of the deep learning algorithm, so as to achieve data recognition and analysis.

[0064] S103, collecting a real-time power consumption sequence about power consumption, and using the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequence to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal;

[0065] After determining the power consumption anomaly identification rules, the real-time power consumption sequence can be analyzed based on the power consumption anomaly identification rules to determine whether the power consumption is abnormal. If it is abnormal, it is convenient to make corresponding strategies to save energy.

[0066] S104. When the power consumption identification result is abnormal, determine the power consumption management strategy according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormality corresponding to the power consumption abnormality sequence, and complete the power consumption management.

[0067] According to the abnormal power consumption sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormal power consumption sequence, it can include: determining the cosine similarity between the real-time power consumption sequence and the abnormal power consumption sequence in the historical data, and running the power operation parameter adjustment strategy corresponding to the abnormal power consumption sequence with the largest cosine similarity, so as to achieve the purpose of energy saving.

[0068] The power supply operation parameter adjustment strategy may be to adjust the working mode or load of the power supply to reduce energy consumption.

[0069] A power supply energy consumption management method provided by the present invention is intended to optimize the energy consumption of a power supply. The method includes first collecting a normal power supply energy consumption sequence, an abnormal power supply energy consumption sequence, and a power supply operation parameter adjustment strategy made by a staff member for anomalies corresponding to the abnormal power supply energy consumption sequence, and then learning the normal power supply energy consumption sequence and the abnormal power supply energy consumption sequence, so as to realize automatic detection rules for power supply energy consumption, and when an abnormal power supply energy consumption is found, automatically executing the power supply operation parameter adjustment strategy corresponding to the abnormal power supply energy consumption sequence, thereby achieving the purpose of saving energy.

[0070] In the embodiment of the present invention, the historical data of power consumption is data input by human-computer interaction or data pre-stored in a database.

[0071] In the prior art, although genetic algorithms have strong randomness, they have poor convergence and are difficult to find a global optimal solution. Therefore, an embodiment of the present invention provides an algorithm parameter optimization method for a deep learning algorithm to solve the technical problems existing in the prior art.

[0072] In an embodiment of the present invention, a deep learning algorithm is used to learn a normal power consumption sequence and an abnormal power consumption sequence to determine an abnormal power consumption recognition rule, including:

[0073] Initialize algorithm parameters corresponding to the deep learning algorithm to determine multiple algorithm parameter vectors for optimizing the deep learning algorithm; wherein each algorithm parameter vector includes all algorithm parameters to be optimized of the deep learning algorithm;

[0074] For example, assuming that the deep learning algorithm is a BP algorithm, it can be randomly initialized between the upper and lower limits of its weights to obtain an algorithm parameter vector. After repeated initialization, multiple algorithm parameters to be optimized can be obtained.

[0075] Construct an objective function for evaluating the performance of a deep learning algorithm, and use a normal power consumption sequence and an abnormal power consumption sequence as supporting data to evaluate the objective function value corresponding to each algorithm parameter vector when applied to the deep learning algorithm according to the objective function;

[0076] According to the objective function value corresponding to each algorithm parameter vector, determine the optimal parameter vector and the worst parameter vector;

[0077] Based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy carrying position influence to obtain an updated algorithm parameter vector;

[0078] Performing position fusion on the algorithm parameter vector after the first update to determine the reference vector, and using the reference vector as a basis, adopting a disturbance guidance strategy to update the algorithm parameter vector after the first update to obtain the algorithm parameter vector after the second update;

[0079] Based on the worst parameter vector, the algorithm parameter vector after the second update is updated by adopting a global guidance strategy to obtain the algorithm parameter vector after the third update;

[0080] Determine whether the iteration end condition is met. If so, determine the power consumption abnormality identification rule based on the algorithm parameter vector after the three updates, otherwise return to the step of determining the optimal parameter vector and the worst parameter vector.

[0081] In an embodiment of the present invention, the objective function for evaluating the performance of a deep learning algorithm is constructed as follows:

[0082]

[0083] in, f Represents the objective function, and the larger the objective function, the better. Indicates i After the power consumption sequence is input into the deep learning algorithm, the actual output obtained is j elements, Indicates i The expected output corresponding to the power consumption sequence j elements, I represents the total number of batch input power consumption sequences, J represents the total number of outputs of the deep learning algorithm, and the power consumption sequence is used to characterize a normal power consumption sequence or an abnormal power consumption sequence. Represents a preset constant term.

[0084] For example, the categories generally include normal categories and multiple different abnormal categories, and the number of outputs of the deep learning algorithm should be the same as the number of output categories.

[0085] In the embodiment of the present invention, based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy carrying position influence to obtain an updated algorithm parameter vector, including:

[0086]

[0087]

[0088] in, Indicates the t-1th optimization. k The algorithm parameter vector l Algorithm parameters, k =1,2,…,K, K represents the total number of algorithm parameter vectors, l =1,2,…,L, where L represents the total number of algorithm parameters in the algorithm parameter vector, Indicates the first k The algorithm parameter vector l Algorithm parameters, represents the first adaptive weight, represents the second adaptive weight, Indicates the tth optimization k The algorithm parameter vector l The speed corresponding to the algorithm parameters is Indicates the t-1th optimization. k The algorithm parameter vector l The speed corresponding to the algorithm parameters is represents the first learning factor, represents the second learning factor, Indicates k The development frequency corresponding to the algorithm parameter vector, represents the optimal parameter vector in the t-1th optimization. l Algorithm parameters, Indicates k Algorithm parameter vector The corresponding historical optimal value l algorithm parameters.

[0089] The fast search strategy with position influence provided in the embodiment of the present invention can adaptively adjust the update speed based on the current optimization times and the position of each algorithm parameter vector itself, and simultaneously learn the information in the optimal parameter vector and the historical optimal value, thereby realizing fast search and effectively improving the search speed of the algorithm.

[0090] In this embodiment of the present invention, the first adaptive weight and the second adaptive weight for:

[0091]

[0092]

[0093] in, represents the minimum weight value, represents the maximum weight, and T represents the maximum number of optimizations;

[0094] The said k The development frequency corresponding to the algorithm parameter vector for:

[0095]

[0096] in, Indicates the initial value of the development frequency, Indicates the frequency correction factor.

[0097] The embodiment of the present invention sets a first adaptive weight and a second adaptive weight. In the early stage of the algorithm optimization process, the second adaptive weight is greater than the first adaptive weight, so that the algorithm parameter vector can retain more of its own historical optimal position information, which is beneficial to the global search. In the later stage of the algorithm optimization process, the second adaptive weight is less than the first adaptive weight, and the position between the algorithm parameter vector and the optimal parameter vector is relatively close, which is beneficial to local search, balances local search and global search, effectively avoids the algorithm premature phenomenon, and enables the algorithm to have stronger search capabilities and convergence speed. The setting of the adaptive development frequency can automatically adjust the step size of the early and late stages, which can achieve fast search while ensuring search accuracy.

[0098] In an embodiment of the present invention, the algorithm parameter vector after the first update is subjected to position fusion, a reference vector is determined, and based on the reference vector, a disturbance guidance strategy is adopted to update the algorithm parameter vector after the first update to obtain the algorithm parameter vector after the second update, including:

[0099] Perform position fusion on the algorithm parameter vector after an update and determine the reference vector as:

[0100]

[0101] in, represents the reference vector at the t-1th optimization time, Indicates m The algorithm parameter vector after one update, Represents the algorithm parameter vector The corresponding objective function value;

[0102] Based on the reference vector, the perturbation guidance strategy is used to update the algorithm parameter vector after one update as follows:

[0103]

[0104] in, Represents the algorithm parameter vector after the second update , Represents a random number between (0,1).

[0105] The disturbance guidance strategy provided in the embodiment of the present invention can realize the fusion of population position information and, with the assistance of disturbance, realize the search of more unfamiliar areas and can effectively improve the search efficiency of a better solution.

[0106] In the embodiment of the present invention, based on the worst parameter vector, the algorithm parameter vector after the second update is updated using a global guidance strategy, and the algorithm parameter vector after the third update is obtained as follows:

[0107]

[0108] in, Indicates the t-1th optimization. n The algorithm parameter vector after the second update l Algorithm parameters, Indicates n The algorithm parameter vector after three updates is l Algorithm parameters, Represents a random number between (0,1), Represents algorithm parameters The corresponding adjustment factor is represents the worst parameter vector l Algorithm parameters, Indicates n The algorithm parameter vector after the second update The corresponding objective function value is represents the worst parameter vector The corresponding objective function value.

[0109] In the embodiment of the present invention, the algorithm parameters The corresponding adjustment factor for:

[0110]

[0111]

[0112] Among them, e represents a natural constant, represents the intermediate parameter, Represents the objective function value corresponding to the optimal parameter vector.

[0113] The global guidance strategy provided by the embodiment of the present invention can search in a non-optimal direction based on the position of the worst parameter vector, so that the global search capability is available in the entire algorithm optimization process, thereby improving the search performance.

[0114] Optionally, in order to ensure the search efficiency of the algorithm, in the process of executing the global guidance strategy, the greedy principle can also be used to retain the better solution, that is, after the objective function value of the algorithm parameter vector after three updates increases, the search is retained, otherwise the original algorithm parameter vector after the second update is retained. It is worth noting that when the parameter is out of bounds, the parameter can be processed to ensure the validity of the parameter.

[0115] like Figure 2 As shown, an embodiment of the present invention provides a power consumption management device, including: a data acquisition module 201, a deep learning module 202, an identification and analysis module 203 and a power consumption management module 204;

[0116] The data collection module 201 is used to collect historical data about power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the abnormal power consumption sequence;

[0117] The deep learning module 202 is used to learn the normal power consumption sequence and the abnormal power consumption sequence using a deep learning algorithm to determine the abnormal power consumption recognition rule;

[0118] The identification and analysis module 203 is used to collect real-time power consumption sequences about power consumption, and use the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequences to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal;

[0119] The power consumption management module 204 is used to determine the power consumption management strategy and complete the power consumption management according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormality corresponding to the power consumption abnormality sequence when the power consumption identification result is abnormal.

[0120] The embodiment of the present invention provides a power consumption management device that can execute the above method and technical solution. Its principles and beneficial effects are similar and will not be described in detail here.

[0121] It should be understood by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0123] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0125] A person of ordinary skill in the art can understand that all or part of the steps in realizing the above-mentioned facts and methods can be completed by instructing the relevant hardware through a program, and the program involved or the program described can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: At this time, the corresponding method steps are derived, and the storage medium can be ROM / RAM, a disk, an optical disk, etc.

[0126] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for managing power consumption, characterized in that: include: Collecting historical data on power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the abnormal power consumption sequence; A deep learning algorithm is used to learn the normal power consumption sequence and the abnormal power consumption sequence to determine the abnormal power consumption identification rules; Collecting a real-time power consumption sequence about power consumption, and using the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequence to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal; When the power consumption recognition result is abnormal, the power consumption management strategy is determined according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormality corresponding to the power consumption abnormality sequence, so as to complete the power consumption management; A deep learning algorithm is used to learn the normal power consumption sequence and the abnormal power consumption sequence to determine the abnormal power consumption identification rules, including: Initialize algorithm parameters corresponding to the deep learning algorithm to determine multiple algorithm parameter vectors for optimizing the deep learning algorithm; wherein each algorithm parameter vector includes all algorithm parameters to be optimized of the deep learning algorithm; Construct an objective function for evaluating the performance of a deep learning algorithm, and use a normal power consumption sequence and an abnormal power consumption sequence as supporting data to evaluate the objective function value corresponding to each algorithm parameter vector when applied to the deep learning algorithm according to the objective function; According to the objective function value corresponding to each algorithm parameter vector, determine the optimal parameter vector and the worst parameter vector; Based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy carrying position influence to obtain an updated algorithm parameter vector; Performing position fusion on the algorithm parameter vector after the first update to determine the reference vector, and using the reference vector as a basis, adopting a disturbance guidance strategy to update the algorithm parameter vector after the first update to obtain the algorithm parameter vector after the second update; Based on the worst parameter vector, the algorithm parameter vector after the second update is updated by adopting a global guidance strategy to obtain the algorithm parameter vector after the third update; Determine whether the iteration end condition is met. If so, determine the power consumption abnormality identification rule based on the algorithm parameter vector after the three updates, otherwise return to the step of determining the optimal parameter vector and the worst parameter vector.

2. The method for managing power consumption according to claim 1, characterized in that: The historical data of power consumption is data input by human-computer interaction or data pre-stored in a database.

3. The method for managing power consumption according to claim 1, characterized in that: The objective function for evaluating the performance of deep learning algorithms is constructed as: Among them, f represents the objective function, and the larger the objective function, the better, y ij It represents the jth element in the actual output obtained after the i-th power consumption sequence is input into the deep learning algorithm. It represents the jth element in the expected output corresponding to the i-th power consumption sequence, I represents the total number of batch input power consumption sequences, J represents the total number of outputs of the deep learning algorithm, the power consumption sequence is used to characterize a normal power consumption sequence or an abnormal power consumption sequence, and ε represents a preset constant term.

4. The method for managing power consumption according to claim 3, characterized in that: Based on the optimal parameter vector, each algorithm parameter vector is updated by adopting a fast search strategy with position influence, and an updated algorithm parameter vector is obtained, including: in, represents the lth algorithm parameter in the kth algorithm parameter vector at the t-1th optimization, k = 1, 2, ..., K, K represents the total number of algorithm parameter vectors, l = 1, 2, ..., L, L represents the total number of algorithm parameters in the algorithm parameter vector, represents the lth algorithm parameter in the kth algorithm parameter vector after an update, ω1 represents the first adaptive weight, ω2 represents the second adaptive weight, represents the speed corresponding to the lth algorithm parameter in the kth algorithm parameter vector during the tth optimization, represents the speed corresponding to the lth algorithm parameter in the kth algorithm parameter vector during the t-1th optimization, c1 represents the first learning factor, c2 represents the second learning factor, represents the development frequency corresponding to the kth algorithm parameter vector, represents the lth algorithm parameter in the optimal parameter vector at the t-1th optimization, p k,l Represents the kth algorithm parameter vector The lth algorithm parameter in the corresponding historical optimal value.

5. The method for managing power consumption according to claim 4, characterized in that: The first adaptive weight ω1 and the second adaptive weight ω2 are: Among them, ω min represents the minimum weight value, ω max represents the maximum weight, and T represents the maximum number of optimizations; The development frequency corresponding to the kth algorithm parameter vector for: in, represents the initial value of the development frequency, and γ represents the frequency correction coefficient.

6. The method for managing power consumption according to claim 4, characterized in that: Performing position fusion on the algorithm parameter vector after the first update, determining the reference vector, and updating the algorithm parameter vector after the first update by adopting the disturbance guidance strategy based on the reference vector to obtain the algorithm parameter vector after the second update, including: Perform position fusion on the algorithm parameter vector after an update and determine the reference vector as: in, represents the reference vector at the t-1th optimization time, represents the algorithm parameter vector after the mth update, Represents the algorithm parameter vector The corresponding objective function value; Based on the reference vector, the perturbation guidance strategy is used to update the algorithm parameter vector after one update as follows: in, Represents the algorithm parameter vector after the second update r1 represents a random number between (0,1).

7. The method for managing power consumption according to claim 6, characterized in that: Based on the worst parameter vector, the algorithm parameter vector after the second update is updated using the global guidance strategy, and the algorithm parameter vector after the third update is obtained as follows: in, represents the lth algorithm parameter in the algorithm parameter vector after the nth secondary update during the t-1th optimization, represents the lth algorithm parameter in the algorithm parameter vector after the nth triple update, r2 represents a random number between (0,1), Represents algorithm parameters The corresponding adjustment factor is represents the lth algorithm parameter in the worst parameter vector, Represents the algorithm parameter vector after the nth secondary update The corresponding objective function value is represents the worst parameter vector The corresponding objective function value.

8. The method for managing power consumption according to claim 7, characterized in that: The algorithm parameters The corresponding adjustment factor for: Among them, e represents a natural constant, τ represents an intermediate parameter, Represents the objective function value corresponding to the optimal parameter vector.

9. A power supply energy consumption management device, the power supply energy consumption management device is used to execute the power supply energy consumption management method according to any one of claims 1 to 8, characterized in that: include: Data acquisition module, deep learning module, identification and analysis module, and power consumption management module; The data collection module is used to collect historical data about power consumption; wherein the historical data includes a normal power consumption sequence, an abnormal power consumption sequence, and a power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the abnormal power consumption sequence; The deep learning module is used to learn the normal power consumption sequence and the abnormal power consumption sequence using a deep learning algorithm to determine the abnormal power consumption recognition rules; The identification and analysis module is used to collect real-time power consumption sequences about power consumption, and use the power consumption abnormality identification rule to identify and analyze the real-time power consumption sequences to obtain a power consumption identification result; the power consumption identification result includes normal or abnormal; The power consumption management module is used to determine the power consumption management strategy and complete the power consumption management according to the power consumption abnormality sequence in the historical data and the power operation parameter adjustment strategy made by the staff for the abnormalities corresponding to the power consumption abnormality sequence when the power consumption identification result is abnormal.

Images