A smart power control device and method based on load identification

By mapping one-dimensional steady-state current data to a multi-dimensional phase space for feature extraction and load identification, a power control strategy is generated, solving the problems of load confusion and missed identification of low-power devices in NILM technology, and realizing high-precision load identification and energy-saving control.

CN123074124APending Publication Date: 2026-11-10BEIJING ZHIDAKE INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611302874.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-26
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing NILM technology suffers from prominent issues of load confusion and under-identification of low-power devices in scenarios with multiple electrical appliances stacked together. It also has low identification accuracy and fails to make targeted adjustments based on load characteristics, making it difficult to meet user needs.

Method used

An intelligent power control device based on load identification is adopted, including modules for steady-state current acquisition, data processing, event segmentation, phase space reconstruction, load identification, and strategy generation. By mapping one-dimensional steady-state current data to a multi-dimensional phase space for feature extraction, and using the K-nearest neighbor algorithm or convolutional neural network for load identification, a power control strategy is generated to regulate the start-up, shutdown, and operating power of electrical appliances.

Benefits of technology

It improves load identification accuracy, enables accurate identification in multi-electrical appliance scenarios, and can generate reasonable power control strategies based on load characteristics, avoiding accidental shutdowns and achieving energy-saving effects by switching flexible loads during off-peak hours.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123074124A_ABST
    Figure CN123074124A_ABST
Patent Text Reader

Abstract

This invention relates to the field of intelligent power control technology, and particularly to an intelligent power control device and method based on load identification. The device includes a steady-state current acquisition module, a data processing module, an event segmentation module, a phase space reconstruction module, a load identification module, a strategy generation module, and an execution module connected in sequence. It maps one-dimensional steady-state current data of multiple electrical appliances to a multi-dimensional phase space and extracts features from the reconstructed multi-dimensional phase space sequence. The load identification module then identifies the load. The strategy generation module generates a power control strategy based on the load identification result and load characteristics. The execution module regulates the start-up and shutdown time and operating power of the load according to the power control strategy. The invention also discloses a method based on the above device. Using this intelligent power control device and method based on load identification improves load identification accuracy. While ensuring basic power consumption based on load characteristics, flexible loads can achieve peak-shaving switching based on time-of-use pricing, thus achieving energy-saving effects.
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent power control, and in particular to an intelligent power control device and method based on load identification. BACKGROUND

[0002] With the development of social economy, household load has gradually become the main part of power consumption, in addition, the energy-saving consciousness of residents is also gradually improving. Therefore, in order to realize reasonable energy management, how to monitor household load becomes an important problem for carrying out household energy management. At present, the traditional load monitoring method mostly uses invasive sensors to monitor single electrical appliances in real time. Although this method can directly obtain the running state of electrical appliances, it needs to install independent measurement devices on each electrical appliance, resulting in large hardware deployment, high installation cost and complex system maintenance, which is not suitable for large-scale promotion in ordinary families. The non-intrusive load monitoring (NILM) technology only needs to collect power signals at the total incoming line end of the family, and can indirectly identify the running state and energy consumption level of different electrical appliances through signal processing and pattern recognition algorithm, thereby significantly reducing the hardware investment and having good application prospect. However, the existing NILM technology still has obvious deficiencies in actual application: the load confusion in multi-appliance superposition scene and the missing identification of low-power devices are prominent, and the identification accuracy is low. And it is difficult to meet the characteristics of users without targeted regulation and control according to the load characteristics. SUMMARY

[0003] The purpose of the present application is to provide an intelligent power control device and method based on load identification, which solves the above technical problems.

[0004] To achieve the above purpose, the present application provides an intelligent power control device based on load identification, comprising: a steady-state current acquisition module for acquiring steady-state current data of a user; a data processing module for preprocessing the acquired steady-state current data; an event segmentation module for time segmentation of the preprocessed data; a phase space reconstruction module for mapping one-dimensional steady-state current data of multiple electrical appliances to a multi-dimensional phase space and performing feature extraction on the multi-dimensional phase space reconstruction sequence; a load identification module for load identification of the extracted features; a strategy generation module for generating a power control strategy according to the load identification result; an execution module for regulating and controlling the start-stop time and running power of the load according to the power control strategy; The steady-state current acquisition module, data processing module, event segmentation module, phase space reconstruction module, load identification module, strategy generation module, and execution module are connected in sequence.

[0005] One method involves using the aforementioned intelligent power control device based on load identification for power control, with the following specific steps: Step S1: Acquire steady-state current signal and preprocess it. Based on CUSUM event detection, segment the preprocessed steady-state current signal to obtain multi-electrical appliance steady-state current data. Step S2: Reconstruct the phase space of the steady-state current data of multiple electrical appliances, map the one-dimensional steady-state current data to the multi-dimensional phase space to obtain the multi-dimensional phase space reconstruction sequence, and extract features from the multi-dimensional phase space reconstruction sequence. Step S3: Input the extracted features into the load classification and recognition model for load recognition; Step S4: Generate an electricity control strategy by solving for the minimum operating cost based on the load identification results and load characteristics.

[0006] Preferably, in step S1, the preprocessing includes one or any combination of low-pass filtering, wavelet-based denoising, or adaptive smoothing.

[0007] Preferably, in step S2, a time delay method is used to reconstruct the multidimensional phase space, and the multidimensional phase space reconstruction sequence is... for: ; in, For the first The first appliance Reconstructing vectors in phase space, ; The expression is as follows: ; in, For the first The first electrical appliance One current sample value, For the first The first electrical appliance Delay of each current sampling value The current value after that, For the first The first electrical appliance Each current sample value is delayed by 2 seconds. The current value after that, For the first The first electrical appliance Delay of each current sampling value The current value after that, Let be the dimension of the multidimensional phase space.

[0008] Preferably, in step S2, when In this case, feature extraction is performed directly using two-dimensional Fourier series; when When the number is even, the elements in the multidimensional phase space reconstruction sequence are grouped into pairs to form the sequence. A two-dimensional contour, for Each two-dimensional contour is used for feature extraction using two-dimensional Fourier series; when When the number is odd, group adjacent elements in the multidimensional phase space reconstruction sequence into a set, forming... A two-dimensional contour, for The two-dimensional contours are used to extract features using two-dimensional Fourier series.

[0009] The preferred feature extraction process using two-dimensional Fourier series is as follows: make: , ; ; in, and These are the cosine and sine coefficients in the X-axis direction, respectively. and These are the cosine and sine coefficients in the Y-axis direction, respectively. The total number of points on the outline. The harmonic order is... For time; Solve for the cosine and sine coefficients of the first S terms along the X and Y axes, excluding... The cosine and sine coefficients in the X and Y directions are used to obtain the 4S-dimensional features as follows: ; The 4S-dimensional features are then normalized.

[0010] Preferably, the load classification and identification model adopts the K-nearest neighbor algorithm model or the convolutional neural network model. The input is the normalized features, and the output is the load label, the load active power, and the equipment start-up and shutdown time.

[0011] Preferably, the load characteristics include rigid loads and flexible loads. Rigid loads are loads that cannot be disconnected from power, while flexible loads allow start-stop and power regulation.

[0012] Preferably, in step S4, the function for minimizing operating costs is as follows: ; in, and These represent the active power of rigid loads and flexible loads, respectively. For real-time electricity prices, and The quantities of rigid loads and flexible loads are respectively. and These are indices for rigid loads and flexible loads, respectively. and These are the start and stop times, respectively.

[0013] Preferably, in step S4, the constraint function of the minimum running cost function is as follows: ; ; in, This represents the maximum total power consumption of the entire house. and The first The minimum and maximum power of a flexible load.

[0014] Therefore, the present invention employs the above-mentioned intelligent power control device and method based on load identification, which has the following beneficial effects: (1) By mapping the one-dimensional steady-state current data of multiple electrical appliances to a multi-dimensional phase space and extracting features from the reconstructed sequence of the multi-dimensional phase space, the implicit nonlinear dynamic geometric features of the current of household appliances such as motors and switching power supplies are fully explored, breaking through the limitations of the traditional time-frequency domain linear feature expression.

[0015] (2) Based on the load identification results and load characteristics, power control strategies are generated. Rigid loads are set up with operation protection to avoid accidental shutdowns affecting users' basic power consumption. Flexible loads rely on time-of-use pricing to achieve peak-shaving switching and achieve energy-saving effects.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an intelligent power control device based on load identification according to the present invention. Figure 2 This is a flowchart of the method of the present invention; Figure 3 For the fan delay of this invention After and Delay 2 The two-dimensional spatial image trajectory diagram, (a) is the time delay The trajectory diagram of the two-dimensional spatial image after delay (b) is shown. Post-two-dimensional spatial image trajectory diagram. Detailed Implementation

[0018] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product is in use. They are used only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," and "connect" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] like Figure 1 As shown, a smart power control device based on load identification includes: The steady-state current acquisition module is used to acquire the user's steady-state current data. It uses a high-precision current transformer with an ADC module for current acquisition, with a sampling frequency set to 100kHz. The high-precision current transformer is installed on the main incoming line to achieve non-destructive installation.

[0021] The data processing module is used to preprocess the collected steady-state current data, using existing data processing devices.

[0022] The event segmentation module has a built-in bilateral CUSUM cumulative sum and change point detection algorithm to perform time segmentation on the preprocessed data.

[0023] The phase space reconstruction module maps the one-dimensional steady-state current data of multiple electrical appliances to a multi-dimensional phase space and extracts features from the multi-dimensional phase space reconstruction sequence. The load identification module is used to identify the load of the extracted features. For lightweight scenarios, the K-nearest neighbor (KNN) model is used, with the optimal value of K set to 2, and Manhattan distance used as the similarity metric. For scenarios with sufficient computing power, a lightweight CNN convolutional neural network is used.

[0024] The strategy generation module generates power control strategies based on load identification results. The execution module regulates the start-stop time and operating power of the load according to the power control strategy. It consists of a multi-channel intelligent relay drive circuit and a communication unit for the main control intelligent circuit breaker. It supports two control methods: branch circuit power disconnection and flexible load power voltage regulation. When high-power dangerous resistive loads such as electric kettles and electric stoves are detected, an emergency disconnect command is immediately output; when long-term standby equipment (standby current < 0.05A, idle for more than 30 minutes) is detected, the branch circuit is automatically disconnected after a delay; during peak hours, the operating power of the air conditioner is reduced and the start of the washing machine is delayed.

[0025] The steady-state current acquisition module, data processing module, event segmentation module, phase space reconstruction module, load identification module, strategy generation module, and execution module are connected sequentially. The entire unit is installed inside the residential inlet distribution box. All modules adopt an onboard integrated design, eliminating the need to modify the individual branch sockets in the room. Sampling wiring is only completed at the live and neutral wires, achieving non-intrusive whole-house load monitoring and automatic control.

[0026] like Figure 2 As shown, one method uses the aforementioned intelligent power control device based on load identification for power control. The specific steps are as follows: Step S1: Acquire steady-state current signal and perform preprocessing. Based on CUSUM event detection, segment the preprocessed steady-state current signal to obtain multi-electrical appliance steady-state current data. Preprocessing includes low-pass filtering, wavelet-based denoising, or adaptive smoothing.

[0027] Step S2: Reconstruct the phase space of the steady-state current data of multiple electrical appliances, map the one-dimensional steady-state current data to the multi-dimensional phase space to obtain the multi-dimensional phase space reconstruction sequence, and extract features from the multi-dimensional phase space reconstruction sequence.

[0028] This embodiment uses a time-delay method for multidimensional phase space reconstruction, and the multidimensional phase space reconstruction sequence... for: ; in, For the first The first appliance Reconstructing vectors in phase space, ; The expression is as follows: ; in, For the first The first electrical appliance One current sample value, For the first The first electrical appliance Delay of each current sampling value ( The current value after a delay of 800 sampling points. For the first The first electrical appliance Each current sample value is delayed by 2 seconds. The current value after that, For the first The first electrical appliance Delay of each current sampling value The current value after that, Let be the dimension of the multidimensional phase space.

[0029] This embodiment Feature extraction is performed directly using a two-dimensional Fourier series. The feature extraction process using a two-dimensional Fourier series is as follows: make: , ; in, and These are the cosine and sine coefficients in the X-axis direction, respectively. and These are the cosine and sine coefficients in the Y-axis direction, respectively. The total number of points on the outline. The harmonic order is... For time; Find the cosine and sine coefficients in the X and Y directions of the first S (3-5) terms, excluding... The cosine and sine coefficients in the X and Y directions are used to obtain the 4S-dimensional features as follows: ; The 4S-dimensional features are then normalized. For example... Figure 3 As shown, the electric fan has a delay. and 2 The two-dimensional spatial trajectory diagram is generated, with the coordinates representing current and the vertical axis representing time. Both are normalized. The trajectory diagrams after the delay are different, and load characteristic information is extracted from them.

[0030] It can also perform higher-dimensional spatial reconstruction to obtain more information, but the computational requirements are higher. When the number is even, the elements in the multidimensional phase space reconstruction sequence are grouped into pairs to form the sequence. A two-dimensional contour, for The two-dimensional contours are used for feature extraction using a two-dimensional Fourier series. When the number is odd, group adjacent elements in the multidimensional phase space reconstruction sequence into a set, forming... A two-dimensional contour, for The two-dimensional contours are used to extract features using two-dimensional Fourier series.

[0031] For example, At that time, the two sets of data were respectively , .

[0032] Step S3: Input the extracted features into the load classification and recognition model for load identification; the load classification and recognition model adopts the K nearest neighbor algorithm model or the convolutional neural network model, the input is the normalized features, and the output is the load label, the load active power and the equipment start-up and shutdown time.

[0033] Step S4: Generate an electricity control strategy by solving for the minimum operating cost based on the load identification results and load characteristics. Load characteristics include rigid loads and flexible loads. Rigid loads are loads that cannot be powered off (refrigerators, lighting, or those limited according to user customization), while flexible loads allow start / stop and power adjustment (air conditioners, water heaters, washing machines, etc.).

[0034] The function for minimizing operating costs is as follows: ; in, and These represent the active power of rigid loads and flexible loads, respectively. The electricity price is set to real-time. The time-of-use electricity price parameters are preset locally: peak period 8:00-22:00, the price is 0.65 yuan / kWh, and off-peak period 22:00-8:00 the next day, the price is 0.22 yuan. and The quantities of rigid loads and flexible loads are respectively. and These are indices for rigid loads and flexible loads, respectively. and These are the start-up and shutdown times, respectively. Priority is given to ensuring uninterrupted operation of rigid loads, and then the washing machine and water heater are all scheduled to start during off-peak hours. During peak hours, the air conditioner's operating power is reduced by 20% to reduce the peak load of the whole house.

[0035] The constraints of the minimum operating cost function include ensuring that the active power of the flexible load is within a safe range and that the total power is less than or equal to the maximum total power of the entire building, as follows: ; in, This represents the maximum total power consumption of the entire house. and The first The minimum and maximum power of a flexible load.

[0036] To verify the effectiveness of this embodiment, a control test was set up. The control group performed load identification directly after time segmentation, while this embodiment performed load identification after spatial reconstruction. The average recognition accuracy was calculated using 10-fold cross-validation. The recognition accuracy of the control group was 87%, while the recognition accuracy of this embodiment was 99%. The technical solution of this application improves the accuracy of load identification.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart power control device based on load identification, characterized in that, include: The steady-state current acquisition module is used to acquire the user's steady-state current data; The data processing module is used to preprocess the collected steady-state current data; The event segmentation module performs time segmentation on the preprocessed data; The phase space reconstruction module is used to map the one-dimensional steady-state current data of multiple electrical appliances to a multi-dimensional phase space and extract features from the multi-dimensional phase space reconstruction sequence. The load identification module is used to identify the load of the extracted features; The strategy generation module generates power control strategies based on load identification results and load characteristics. The execution module regulates the start-up and stop times and operating power of the load according to the power control strategy; The steady-state current acquisition module, data processing module, event segmentation module, phase space reconstruction module, load identification module, strategy generation module, and execution module are connected in sequence.

2. A method for controlling electricity consumption using the intelligent electricity control device based on load identification as described in claim 1, characterized in that, The specific steps are as follows: Step S1: Acquire steady-state current signal and preprocess it. Based on CUSUM event detection, segment the preprocessed steady-state current signal to obtain multi-electrical appliance steady-state current data. Step S2: Reconstruct the phase space of the steady-state current data of multiple electrical appliances, map the one-dimensional steady-state current data to the multi-dimensional phase space to obtain the multi-dimensional phase space reconstruction sequence, and extract features from the multi-dimensional phase space reconstruction sequence. Step S3: Input the extracted features into the load classification and recognition model for load recognition; Step S4: Generate an electricity control strategy by solving for the minimum operating cost based on the load identification results and load characteristics.

3. The method according to claim 2, characterized in that: In step S1, the preprocessing includes one or any combination of low-pass filtering, wavelet-based denoising, or adaptive smoothing.

4. The method according to claim 3, characterized in that: In step S2, a time delay method is used to reconstruct the multidimensional phase space, resulting in a multidimensional phase space reconstruction sequence. for: ; in, For the first The first appliance Reconstructing vectors in phase space, ; The expression is as follows: ; in, For the first The first electrical appliance One current sample value, For the first The first electrical appliance Delay of each current sampling value The current value after that, For the first The first electrical appliance Each current sample value is delayed by 2 seconds. The current value after that, For the first The first electrical appliance Delay of each current sampling value The current value after that, Let be the dimension of the multidimensional phase space.

5. A method according to claim 4, characterized in that: In step S2, when In this case, feature extraction is performed directly using two-dimensional Fourier series; when When the number is even, the elements in the multidimensional phase space reconstruction sequence are grouped into pairs to form the sequence. A two-dimensional contour, for Each two-dimensional contour is used for feature extraction using two-dimensional Fourier series; when When the number is odd, group adjacent elements in the multidimensional phase space reconstruction sequence into a set, forming... A two-dimensional contour, for The two-dimensional contours are used to extract features using two-dimensional Fourier series.

6. A method according to claim 5, characterized in that: The feature extraction process using two-dimensional Fourier series is as follows: make: , ; ; in, and These are the cosine and sine coefficients in the X-axis direction, respectively. and These are the cosine and sine coefficients in the Y-axis direction, respectively. The total number of points on the outline. The harmonic order is... For time; Solve for the cosine and sine coefficients of the first S terms along the X and Y axes, excluding... The cosine and sine coefficients in the X and Y directions are used to obtain the 4S-dimensional features as follows: ; The 4S-dimensional features are then normalized.

7. A method according to claim 6, characterized in that: The load classification and identification model adopts the K-nearest neighbor algorithm or the convolutional neural network model. The input is the normalized features, and the output is the load label, the load active power, and the equipment start-up and shutdown time.

8. A method according to claim 7, characterized in that: Load characteristics include rigid loads and flexible loads. Rigid loads are loads that cannot be disconnected from power, while flexible loads allow start-stop and power regulation.

9. A method according to claim 8, characterized in that: In step S4, the function for minimizing operating costs is as follows: ; in, and These represent the active power of rigid loads and flexible loads, respectively. For real-time electricity prices, and The quantities of rigid loads and flexible loads are respectively. and These are indices for rigid loads and flexible loads, respectively. and These are the start and stop times, respectively.

10. A method according to claim 9, characterized in that: In step S4, the constraint function for the minimum running cost function is as follows: ; ; in, This represents the maximum total power consumption of the entire house. and The first The minimum and maximum power of a flexible load.