Industrial equipment load identification method, device, storage medium and computer equipment
By acquiring total power load data and utilizing pre-trained load identification models and neural network models, load events of industrial equipment are identified, solving the problem of poor load identification accuracy among industrial users and achieving accurate identification of equipment type and event type.
Patent Information
- Application Number
- CN202311478500.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-08
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2043-11-08
AI Technical Summary
Existing non-intrusive load identification methods have poor accuracy in industrial users due to similar equipment energy consumption, complex noise, and the common phenomenon of segmented start-stop during equipment operation.
By acquiring total power load data, detecting load events, and based on a pre-trained load identification model, using a time series prediction model and a neural network model with an attention mechanism, active power, reactive power, and current data are extracted to identify the equipment type and event type corresponding to the load events.
It improves the accuracy of load identification for industrial equipment under steady-state power conditions and can effectively identify equipment start-up and shutdown events.
Smart Images

Figure CN117543592B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of load identification, and in particular to an industrial equipment load identification method, device, storage medium and computer equipment. Background Art
[0002] On the distribution side of the power grid, industrial users are an important part of electricity users, and their electricity load is showing a high proportion and diversified development trend, which has a significant impact on the upgrade, operation and maintenance of the power grid. Therefore, it is necessary to identify the load of industrial users' electrical equipment and understand the working status, power consumption and power consumption patterns of each equipment, so as to provide data support for environmental monitoring, energy efficiency services, abnormal source positioning and other needs.
[0003] There are two ways to identify the load of a device: intrusive and non-intrusive. Intrusive load identification usually requires the installation of power monitoring equipment on each device, which is costly and complex to install. Non-intrusive load identification calculates the power consumption of each device by processing the total load data of the entire system. It is easier to implement and integrate into the system, so non-intrusive load identification is more commonly used.
[0004] Existing non-intrusive load identification methods are relatively mature in applications for home users with simple equipment structures. However, industrial users usually have a wide variety of electrical equipment, and the energy consumption of the equipment is similar, the noise is complex, and the equipment is usually started and stopped in stages and slowly started and stopped during operation, which will interfere with load identification. In this case, the traditional load identification method based on the change in electrical data characteristics before and after the load event occurs has poor accuracy. Summary of the invention
[0005] In view of this, the present application provides an industrial equipment load identification method, device, storage medium and computer equipment, the main purpose of which is to solve the technical problem of poor load identification accuracy when the energy consumption of industrial equipment is similar.
[0006] According to a first aspect of the present invention, there is provided an industrial equipment load identification method, the method comprising:
[0007] Obtaining total power load data of the user to be tested, the total power load data including active power, reactive power and current data;
[0008] Performing load event detection on the total power load data, wherein the types of load events include equipment startup and equipment shutdown;
[0009] If a load event is detected in the total power load data, active power, reactive power and current data corresponding to the load event are extracted from the total power load data to obtain a sequence of events to be tested;
[0010] According to the event sequence to be tested, based on the pre-trained load identification model, a load identification result is obtained, wherein the load identification result includes the equipment type and the load event type corresponding to the load event.
[0011] Optionally, active power, reactive power and current data corresponding to the load event are extracted from the total power load data to obtain a sequence of events to be measured, specifically including:
[0012] The active power, reactive power and current data corresponding to the load event in the total load data are intercepted according to the preset sequence length to obtain the event sequence to be measured, wherein the event sequence to be measured includes a first steady-state sequence before the load event occurs, a second steady-state sequence after the load event occurs, and a load event sequence between the first steady-state sequence and the second steady-state sequence.
[0013] Optionally, according to the event sequence to be tested, based on a pre-trained load identification model, a load identification result is obtained, specifically including:
[0014] Calculate a first current average value corresponding to the first steady-state sequence and a second current average value corresponding to the second steady-state sequence;
[0015] If the first current average value is less than the second current average value, determining that the type of the load event is equipment startup;
[0016] If the first current average value is greater than the second current average value, it is determined that the type of the load event is equipment shutdown;
[0017] According to the type of load event, a data difference sequence corresponding to the data change amount caused by the load event is extracted from the event sequence to be tested;
[0018] Calculate the active power, reactive power, fifth harmonic current and seventh harmonic current corresponding to the data difference sequence to obtain characteristic data, and determine the equipment type corresponding to the load event using the load identification model based on the characteristic data;
[0019] The load identification results are obtained based on the load event type and equipment type.
[0020] Optionally, according to the load event type, extracting a data difference sequence corresponding to the data change amount caused by the load event in the event sequence to be tested, specifically including:
[0021] If the load event type is load start, then according to the first steady-state sequence, a pre-trained time series prediction model is used to obtain a first extended sequence corresponding to the first steady-state sequence, and the difference between the event sequence to be tested and the first extended sequence is calculated to obtain a data difference sequence;
[0022] If the load event type is load stop, the second extended sequence corresponding to the second steady-state sequence is obtained according to the second steady-state sequence using the pre-trained time series prediction model, and the difference between the event sequence to be tested and the second extended sequence is calculated to obtain a data difference sequence.
[0023] Optional training methods for time series forecasting models, including:
[0024] Acquire total active power, total reactive power and total current data corresponding to the simultaneous steady-state operation of multiple devices to obtain multiple steady-state data samples, wherein each of the steady-state data samples includes an observation sequence and a target sequence, and the last data of the observation sequence and the first data of the target sequence are adjacent data;
[0025] Taking the observation sequence as input and the target sequence as output, the preset neural network model is trained to obtain a time series prediction model.
[0026] Optional training methods for load identification models include:
[0027] Acquire multiple power load samples of multiple devices, wherein each power load sample includes active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the power load sample when a load event occurs;
[0028] A neural network model based on the attention mechanism is constructed, and the power load samples are used to train the neural network model to obtain a pre-trained model. The parameters of the pre-trained model are optimized through the cross entropy loss function to obtain the load recognition model.
[0029] Optional training methods for load identification models include:
[0030] Obtaining a total power load sample of multiple devices, and multiple sub-power load samples of multiple devices corresponding to the total power load sample, wherein the total power load sample includes the total active power, total reactive power, total fifth harmonic current, and total seventh harmonic current corresponding to the multiple devices corresponding to the total power load sample when power load events occur simultaneously, and each sub-power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current, and device type corresponding to the device corresponding to the sub-power load sample when a load event occurs;
[0031] Perform data alignment processing on the total power load sample and the sub-power load sample based on the least squares fitting algorithm to synchronize the data of the total power load sample and each sub-power load sample;
[0032] A neural network model based on the attention mechanism is constructed, and the total power load samples and sub-power load samples are used to construct joint features to train the neural network model to obtain a pre-trained model. The pre-trained model is optimized by the cross entropy loss function to obtain the load recognition model.
[0033] According to a second aspect of the present invention, there is provided an industrial equipment load identification device, the device comprising:
[0034] A data acquisition module is used to acquire the total power load data of the user to be tested, and the total power load data includes active power, reactive power and current data;
[0035] A load event detection module, used to perform load event detection on the total power load data, wherein the types of load events include equipment startup and equipment shutdown;
[0036] A sequence extraction module is used to extract active power, reactive power and current data corresponding to the load event from the total power load data if a load event is detected in the total power load data, so as to obtain a sequence of events to be tested;
[0037] The load identification module is used to obtain a load identification result according to the event sequence to be tested and based on a pre-trained load identification model, wherein the load identification result includes the equipment type and the load event type corresponding to the load event.
[0038] According to a third aspect of the present invention, there is provided a storage medium having a computer program stored thereon, wherein the program implements the above-mentioned industrial equipment load identification method when executed by a processor.
[0039] According to a fourth aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned industrial equipment load identification method when executing the program.
[0040] The present invention provides an industrial equipment load identification method, device, storage medium and computer equipment, firstly obtaining the total power load data of the user to be tested, wherein the total power load data includes active power, reactive power and current data; then performing load event detection on the total power load data, wherein the types of load events include equipment startup and equipment stop; if a load event is detected in the total power load data, then extracting the active power, reactive power and current data corresponding to the load event from the total power load data to obtain a sequence of events to be tested; finally, according to the sequence of events to be tested, based on a pre-trained load identification model, obtaining a load identification result, wherein the load identification result includes the equipment type and load event type corresponding to the load event. The technical solution of the present application is based on the power data and current data when a load event occurs, and uses the load identification model to learn the fluctuation data generated by the load event, thereby identifying the equipment type where the load event occurs and the corresponding load event type. This method can improve the recognition accuracy of industrial equipment loads close to steady-state power.
[0041] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0043] Figure 1 A schematic diagram of a flow chart of an industrial equipment load identification method provided by an embodiment of the present invention is shown;
[0044] Figure 2 A schematic diagram of a neural network structure provided by an embodiment of the present invention is shown;
[0045] Figure 3 A schematic diagram of the structure of a basic residual block provided by an embodiment of the present invention is shown;
[0046] Figure 4 A schematic diagram of the structure of an industrial equipment load identification device provided by an embodiment of the present invention is shown;
[0047] Figure 5 A schematic diagram of the structure of a storage medium and a computer device provided in an embodiment of the present invention is shown. DETAILED DESCRIPTION
[0048] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.
[0049] Load identification technology collects users' electricity usage data as a basis to understand users' electricity usage habits, power consumption and operation status, and then guide users' use of electrical equipment. Load identification technology can be divided into two types: intrusive and non-intrusive according to the data collection method. Intrusive load identification technology requires the installation of monitoring devices on each electrical device. This method can accurately obtain the electricity consumption data information of each electrical device, but the installation and maintenance of the monitoring device will consume a lot of money and time, resulting in reduced system reliability, and will involve issues such as infringement of user privacy during installation and maintenance, and the monitoring device will generate energy consumption, affecting the power system. Non-intrusive load identification technology only requires the installation of a monitoring device at the main power supply port of the power user. Compared with intrusive load identification, it reduces installation and maintenance costs and is easier to implement and integrate into the system. Therefore, non-intrusive load identification is more commonly used.
[0050] Most of the existing non-invasive load identification methods are relatively mature in applications for home users. The reason is that the equipment structure of home users is relatively simple, and the power consumption or electrical characteristics of each device vary greatly. Based on the characteristic differences of each device, it is easy to distinguish the equipment category corresponding to the load data. For industrial users, taking ceramic factories as an example, there are many types of electrical equipment, and the energy consumption of the equipment is similar, the noise is complex, and the segmented start and stop and slow start and stop phenomena are common when the equipment is running, which will interfere with load identification. The traditional load identification based on the change in steady-state power before and after the equipment is switched on is not accurate enough. Therefore, the embodiment of this application proposes a method for load identification of industrial equipment to improve the accuracy of load identification when the energy consumption of industrial equipment is similar.
[0051] In one embodiment, Figure 1 As shown, a method for identifying load of industrial equipment is provided, comprising the following steps:
[0052] 101. Obtain total power load data of the user to be tested, where the total power load data includes active power, reactive power, and current data.
[0053] 102. Perform load event detection on the total power load data, wherein the types of load events include equipment startup and equipment shutdown.
[0054] 103. If a load event is detected in the total power load data, active power, reactive power and current data corresponding to the load event are extracted from the total power load data to obtain a sequence of events to be tested.
[0055] 104. According to the event sequence to be tested, based on the pre-trained load identification model, a load identification result is obtained, wherein the load identification result includes the equipment type and the load event type corresponding to the load event.
[0056] In this embodiment, active power, reactive power and current data at the user bus to be tested are obtained based on a preset time interval to obtain total power load data. Load event detection is performed on the acquired total power load data. When a load event is detected in the total power load data, a data sequence corresponding to the load event is intercepted from the total power load data to obtain a sequence of events to be tested, so as to perform subsequent analysis based on the sequence of events to be tested. The event detection of the total power load data can be performed using a sliding window algorithm, a time series analysis method, etc., which is not limited in the embodiment of the present application.
[0057] In this embodiment, the load identification model can be obtained by training a convolutional neural network model based on the attention mechanism based on the power load data samples corresponding to each device when the load event occurs. The load identification model identifies the data to be tested by learning the characteristics of the power load data corresponding to each electrical device when the load event occurs, and obtains the equipment type corresponding to the load event in the data to be tested. Among them, taking the ceramic factory as an example, the equipment types may include fans, cranes, ball mills, mixers, and electric kilns. The event sequence to be tested may include data corresponding to the entire process of the load event and a section of steady-state data corresponding to each before and after the load event. By comparing the steady-state data before and after the load event, it is determined whether the load event type that occurred is equipment startup or equipment stop. It is also possible to add a label of the corresponding load event type to the power load data sample of the training load identification model, so that the model learns the data characteristics of different load event types, and simultaneously identifies the equipment type and load event type through the load identification model.
[0058] The industrial equipment load identification method provided in this embodiment is based on the power data and current data when the load event occurs, and uses the load identification model to learn the fluctuation data generated by the load event, and then identifies the equipment type where the load event occurs and the corresponding load event type. This method can improve the recognition accuracy of industrial equipment loads with close steady-state power.
[0059] In an optional embodiment, in step 103, the active power, reactive power and current data corresponding to the load event are extracted from the total power load data to obtain the event sequence to be measured. The method can be specifically implemented by the following steps: the active power, reactive power and current data corresponding to the load event in the total load data are intercepted according to a preset sequence length to obtain the event sequence to be measured, wherein the event sequence to be measured includes a first steady-state sequence before the load event occurs, a second steady-state sequence after the load event occurs, and a load event sequence between the first steady-state sequence and the second steady-state sequence.
[0060] In the above embodiment, the length of the sequence obtained by intercepting the total power load data should not be less than a certain range, and the intercepted event sequence to be measured must include the steady-state sequence before and after the load event and the load event sequence when the load event occurs. For example, the total power load data can be intercepted according to a continuous time step of 100. The value of the preset sequence length can be set based on actual application experience, and the embodiment of the present application does not limit the specific value.
[0061] In an optional embodiment, in step 104, according to the event sequence to be tested, based on a pre-trained load identification model, a method for obtaining a load identification result can be specifically implemented by the following steps: calculating a first current average value corresponding to the first steady-state sequence and a second current average value corresponding to the second steady-state sequence; if the first current average value is less than the second current average value, determining that the type of the load event is equipment startup; if the first current average value is greater than the second current average value, determining that the type of the load event is equipment stop; according to the type of the load event, extracting a data difference sequence corresponding to the data change amount caused by the occurrence of the load event in the event sequence to be tested; calculating the active power, reactive power, fifth harmonic current and seventh harmonic current corresponding to the data difference sequence to obtain characteristic data, and determining the type of equipment corresponding to the load event according to the characteristic data using the load identification model; and obtaining a load identification result according to the load event type and the equipment type.
[0062] In the above embodiment, the current and power of the total power load data will change before and after the load event occurs. The type of the load event can be determined by calculating the difference in current or power before and after the load event occurs. If the current or power before the load event occurs is greater than the current or power after the load event occurs, the load event type is equipment stop. If the current or power before the load event occurs is less than the current or power after the load event occurs, the load event type is equipment start.
[0063] In the above embodiment, because the load identification model is trained based on the power load data samples of each device when the load event occurs, when the load identification is actually performed on the user, there may be a device running in a steady state before the load event occurs. Taking the type of load event as device startup as an example, if the first steady-state sequence of the event sequence to be measured obtained under ideal conditions should be zero, the second steady-state sequence is the steady-state data corresponding to the device where the load event occurs, and the load event sequence between the first steady-state sequence and the second steady-state sequence is the transient data corresponding to the device startup process. In fact, there is a situation where the first steady-state sequence is not zero, that is, the load event sequence and the second steady-state sequence will contain the components of the first steady-state sequence, the load event sequence is the superposition of the steady-state data corresponding to the device that was already running before the load event occurred and the transient data when the load event occurred, and the second steady-state sequence is the superposition of the steady-state data corresponding to the device that was already running before the load event occurred and the steady-state data of the device corresponding to the load event, and the simultaneous operation of multiple devices will interfere with device identification.
[0064] In order to improve the accuracy of identification, it is necessary to eliminate the steady-state data components corresponding to the equipment that has been running before the load event occurs in the event sequence to be tested, and obtain the data difference sequence corresponding to the data change amount generated in the total power load data due to the load event. The data difference sequence includes active power, reactive power and current data. The fifth harmonic current and the seventh harmonic current are calculated based on the current data. Then, based on the active power, reactive power, fifth harmonic current and seventh harmonic current of the data difference sequence, the load identification model is used to determine the equipment type corresponding to the load event. The load identification result is obtained according to the load event type and the equipment type.
[0065] In an optional embodiment, according to the load event type, the method for extracting the data difference sequence corresponding to the data change caused by the load event in the event sequence to be tested can be specifically implemented by the following steps: if the load event type is load startup, then according to the first steady-state sequence, using the pre-trained time series prediction model, a first extended sequence corresponding to the first steady-state sequence is obtained, and the difference between the event sequence to be tested and the first extended sequence is calculated to obtain the data difference sequence; if the load event type is load stop, then according to the second steady-state sequence, using the pre-trained time series prediction model, a second extended sequence corresponding to the second steady-state sequence is obtained, and the difference between the event sequence to be tested and the second extended sequence is calculated to obtain the data difference sequence.
[0066] Among them, in an optional embodiment, the training method of the time series prediction model can be specifically implemented through the following steps: obtaining the total active power, total reactive power and total current data corresponding to the simultaneous steady-state operation of multiple devices, and obtaining multiple steady-state data samples, wherein each of the steady-state data samples includes an observation sequence and a target sequence, and the last data of the observation sequence and the first data of the target sequence are adjacent data; using the observation sequence as input and the target sequence as output, training a preset neural network model to obtain a time series prediction model.
[0067] In the above embodiment, if the load event type is equipment startup, then the steady-state data corresponding to the equipment that was already running before the load event occurred in the event sequence to be tested is the first steady-state sequence; if the load event type is equipment stop, then the steady-state data corresponding to the equipment that was already running before the load event occurred in the event sequence to be tested is the second steady-state sequence, wherein the equipment that was already running before the load event occurred refers to other steady-state running equipment in the event sequence to be tested except the equipment where the load event occurred. The sequence of multiple devices running steadily at the same time is not a stable value, but a sequence that may change periodically, so it is necessary to use a time series prediction model to expand the first steady-state sequence or the second steady-state sequence to obtain an expanded sequence corresponding to the steady-state data of other devices in the entire event sequence to be tested except the device where the load event occurred.
[0068] Among them, the time series prediction model can be obtained by training the corresponding steady-state data samples when multiple devices are running simultaneously. When multiple devices are running simultaneously, the total active power, total reactive power and total current data are obtained to obtain a group of steady-state data samples corresponding to the devices, and the steady-state data samples corresponding to multiple groups of devices are obtained to obtain multiple steady-state data samples. For each steady-state data sample, it is divided into two groups of sequences according to the order of time, and the front group of sequences is set as the observation sequence as the input of the neural network model, and the back group of sequences is set as the target sequence as the output of the neural network model. It can be understood that the observation sequence and the target sequence are adjacent sequences, that is, the last data in the observation sequence is adjacent to the first data in the target sequence.
[0069] After obtaining the extended sequence, the difference between the event sequence to be tested and the extended sequence is calculated to obtain a data difference sequence. Ideally, the data difference sequence corresponding to the load event when the device is started should include: a zero value sequence before the device is started, a transient sequence during the device startup process, and a steady-state sequence of the device running stably after the device is started; the data difference sequence corresponding to the load event when the device is stopped should include: a steady-state sequence before the device stops, a transient sequence during the device stops, and a zero value sequence after the device stops. The description of the ideal situation in the embodiment of the present application is intended to help understand the content of the embodiment, and is not a limitation on the embodiment of the present application.
[0070] In an optional embodiment, the training method of the load identification model can be specifically implemented through the following steps: obtaining multiple power load samples of multiple devices, wherein each power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the power load sample when a load event occurs; constructing a neural network model based on an attention mechanism, using the power load samples to train the neural network model to obtain a pre-trained model, and optimizing the parameters of the pre-trained model through a cross-entropy loss function to obtain a load identification model.
[0071] In the above embodiment, the load identification model is trained by multiple power load samples corresponding to multiple devices. Based on the above embodiment, it can be understood that the sequence length and sequence structure of each power load sample are consistent with the data difference sequence. Before training the model, each power load sample is normalized to normalize the sample data between (-1, 1). Suppose the power load sample is X = {x1, x2, x3…x n}, the normalized calculation formula is as follows:
[0072]
[0073] Among them, X′ is the normalized sample data, X max and X min is the maximum and minimum value in the sample data, R max and R min are the right and left endpoints of the normalized interval of the sample data.
[0074] The normalized power load sample data is divided into a training set and a test set, and a convolutional neural network model based on the attention mechanism is constructed. The training set is used to train the neural network model, and the test set and the cross entropy loss function are used to optimize the model parameters to obtain the load recognition model. The expression of the cross entropy loss function is as follows:
[0075]
[0076] Among them, X is the recognition result of the model, Y is the actual result of the sample data, and N is the number of samples.
[0077] Among them, Figure 2 As shown in Figure 1, the neural network model includes basic residual blocks 1 to 5, a channel attention module, a conversion module, and a fully connected layer module. The structure of the basic residual block is shown in Figure 1. Figure 3 As shown in the figure, it includes convolutional layers and layer normalization modules. The size of the convolution kernel of the convolutional layer is set to 3, and the number of output channels of the convolutional layer is set to 64. The layer normalization module is used to calculate the mean and variance of each sample in the feature dimension and normalize it so that the output of each sample has the same or similar scale, which helps to reduce the variation of different samples in the feature dimension and improve the robustness and generalization ability of the model. The activation function of the basic residual block is the Leaky ReLU function. The channel attention module after the five basic residual blocks is used to weight different channel features. The conversion module is used to perform tensor flattening (Faltten) operations to expand the multidimensional array into one dimension without changing the order of the elements. The fully connected layer is used to linearly transform the input data and generate output through the activation function. The number of output nodes of the fully connected layer is the number of device types.
[0078] In an optional embodiment, the training method of the load identification model can also be specifically implemented by the following steps: obtaining a total power load sample of multiple devices, and multiple sub-power load samples of multiple devices corresponding to the total power load sample, wherein the total power load sample includes the total active power, total reactive power, total fifth harmonic current and total seventh harmonic current corresponding to the multiple devices corresponding to the total power load sample when a power load event occurs simultaneously, and each sub-power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the sub-power load sample when the load event occurs; performing data alignment processing on the total power load sample and the sub-power load sample based on the least squares fitting algorithm to synchronize the data of the total power load sample and each sub-power load sample; constructing a neural network model based on the attention mechanism, using the total power load sample and the sub-power load sample to construct a joint feature to train the neural network model to obtain a pre-trained model, and optimizing the pre-trained model through a cross-entropy loss function to obtain a load identification model.
[0079] In the above embodiment, considering that in practice there may be a situation where multiple devices are turned on or stopped at the same time, when load events occur simultaneously in multiple devices, it will interfere with load identification. In this case, the total power load sample when load events occur simultaneously in multiple devices and the sub-power load sample corresponding to each device can be obtained. The joint features of the total power load sample and the sub-power load sample are established by training the neural network model to obtain the load identification model.
[0080] In this embodiment, since the total power load sample and each sub-power load sample are collected through different channels, there may be a starting delay time. Therefore, the least squares fitting algorithm can be used to perform data alignment processing on the total power load sample and the sub-power load sample to reduce the recognition error caused by data asynchrony.
[0081] The industrial equipment load identification method provided in the embodiment of the present application is based on the power data and current data when the load event occurs, and uses the load identification model to learn the data during the load event occurrence, so as to identify the equipment type where the load event occurs and the corresponding load event type. The embodiment provided in the present application constructs a load identification model obtained by joint feature training for the total power load samples and the power load samples corresponding to a single device, and can also identify the situation where load events occur simultaneously in multiple devices. This method can improve the accuracy of load identification for industrial equipment with similar steady-state power.
[0082] Further, as Figure 1 The specific implementation of the method shown in the embodiment provides an industrial equipment load identification device, such as Figure 4 As shown, the device includes: a data acquisition module 31, a load event detection module 32, a sequence extraction module 33, and a load identification module 34.
[0083] The data acquisition module 31 can be used to obtain the total power load data of the user to be tested, and the total power load data includes active power, reactive power and current data;
[0084] A load event detection module 32, which can be used to perform load event detection on the total power load data, wherein the types of load events include equipment startup and equipment shutdown;
[0085] The sequence extraction module 33 can be used to extract the active power, reactive power and current data corresponding to the load event from the total power load data if a load event is detected in the total power load data, so as to obtain a sequence of events to be tested;
[0086] The load identification module 34 can be used to obtain a load identification result according to the event sequence to be tested based on a pre-trained load identification model, wherein the load identification result includes the equipment type and the load event type corresponding to the load event.
[0087] In a specific application scenario, the sequence extraction module 33 can be specifically used to intercept the active power, reactive power and current data corresponding to the load event in the total load data according to a preset sequence length to obtain a sequence of events to be tested, wherein the sequence of events to be tested includes a first steady-state sequence before the load event occurs, a second steady-state sequence after the load event occurs, and a load event sequence between the first steady-state sequence and the second steady-state sequence.
[0088] In a specific application scenario, the load identification module 34 can be specifically used to calculate a first current average value corresponding to a first steady-state sequence and a second current average value corresponding to a second steady-state sequence; if the first current average value is less than the second current average value, the type of the load event is determined to be equipment startup; if the first current average value is greater than the second current average value, the type of the load event is determined to be equipment stop; according to the type of load event, a data difference sequence corresponding to the data change caused by the load event is extracted from the event sequence to be tested; the active power, reactive power, fifth harmonic current and seventh harmonic current corresponding to the data difference sequence are calculated to obtain characteristic data, and according to the characteristic data, the load identification model is used to determine the type of equipment corresponding to the load event; the load identification result is obtained according to the load event type and the equipment type.
[0089] In a specific application scenario, the sequence extraction module 33 can also be used to obtain a first extended sequence corresponding to the first steady-state sequence based on the first steady-state sequence using a pre-trained time series prediction model if the load event type is load start-up, and calculate the difference between the event sequence to be tested and the first extended sequence to obtain a data difference sequence; if the load event type is load stop, obtain a second extended sequence corresponding to the second steady-state sequence based on the second steady-state sequence using a pre-trained time series prediction model, and calculate the difference between the event sequence to be tested and the second extended sequence to obtain a data difference sequence.
[0090] In a specific application scenario, the sequence extraction module 33 can also be used to obtain the total active power, total reactive power and total current data corresponding to multiple devices when they are operating in a steady state at the same time, and obtain multiple steady-state data samples, wherein each of the steady-state data samples includes an observation sequence and a target sequence, and the last data of the observation sequence and the first data of the target sequence are adjacent data; with the observation sequence as input and the target sequence as output, the preset neural network model is trained to obtain a time series prediction model.
[0091] In a specific application scenario, the load identification module 34 can also be used to obtain multiple power load samples of multiple devices, wherein each power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the power load sample when a load event occurs; construct a neural network model based on the attention mechanism, use the power load samples to train the neural network model to obtain a pre-trained model, and optimize the parameters of the pre-trained model through the cross-entropy loss function to obtain a load identification model. In a specific application scenario, the load identification module 34 can also be used to obtain total power load samples of multiple devices, and multiple sub-power load samples of multiple devices corresponding to the total power load samples, wherein the total power load sample includes the total active power, total reactive power, total fifth harmonic current and total seventh harmonic current corresponding to the multiple devices corresponding to the total power load sample when power load events occur simultaneously, and each sub-power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the sub-power load sample when the load event occurs; the total power load sample and the sub-power load sample are data aligned based on the least squares fitting algorithm to synchronize the data of the total power load sample and each sub-power load sample; a neural network model based on an attention mechanism is constructed, and the total power load sample and the sub-power load sample are used to construct joint features to train the neural network model to obtain a pre-trained model, and the pre-trained model is optimized by a cross-entropy loss function to obtain a load identification model.
[0092] It should be noted that for other corresponding descriptions of the functional units involved in the industrial equipment load identification device provided in this embodiment, reference can be made to Figure 1 The corresponding description in will not be repeated here.
[0093] Based on the above Figure 1 The method shown in the embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned Figure 1 The industrial equipment load identification method shown.
[0094] Based on this understanding, the technical solution of the present application can be embodied in the form of a software product. The software product to be identified can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present application.
[0095] Based on the above Figure 1 The method shown, and Figure 4 In order to achieve the above-mentioned purpose, the embodiment of the industrial equipment load identification device shown in the figure is as follows: Figure 5 As shown, this embodiment also provides a computer device for industrial equipment load identification, which can be a personal computer, a server, a smart phone, a tablet computer, a smart watch, or other network devices, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program and an operating system; the processor is used to execute the computer program to implement the above-mentioned Figure 1 The method shown.
[0096] Optionally, the computer device may further include an internal memory, a communication interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, a display, an input device such as a keyboard, etc. Optionally, the communication interface may also include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.
[0097] Those skilled in the art will appreciate that the computer device structure for identifying an operation action provided in this embodiment does not constitute a limitation on the computer device, and may include more or fewer components, or a combination of certain components, or different component arrangements.
[0098] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the above-mentioned computer device hardware and the software resources to be identified, and supports the operation of the information processing program and other software and / or programs to be identified. The network communication module is used to realize the communication between the components inside the storage medium, and the communication with other hardware and software in the information processing computer device.
[0099] Through the description of the above implementation methods, the technicians in this field can clearly understand that the present application can be implemented by means of software plus the necessary general hardware platform, or by hardware. By applying the technical solution of the present application, the total power load data of the user to be tested is first obtained, and the total power load data includes active power, reactive power and current data; then the total power load data is subjected to load event detection, wherein the types of load events include equipment start and equipment stop; if a load event is detected in the total power load data, the active power, reactive power and current data corresponding to the load event are extracted from the total power load data to obtain the event sequence to be tested; finally, according to the event sequence to be tested, based on the pre-trained load identification model, a load identification result is obtained, wherein the load identification result includes the equipment type and load event type corresponding to the load event. Compared with the prior art, the technical solution of the present application is based on the power data and current data when the load event occurs, and the load identification model is used to learn the fluctuation data generated by the load event, thereby identifying the equipment type and the corresponding load event type of the load event, and this method can improve the recognition accuracy of the industrial equipment load with close steady-state power.
[0100] Those skilled in the art will appreciate that the accompanying drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily necessary for implementing the present application. Those skilled in the art will appreciate that the modules in the devices in the implementation scenario can be distributed in the devices of the implementation scenario according to the description of the implementation scenario, or can be changed accordingly and located in one or more devices different from the present implementation scenario. The modules of the above-mentioned implementation scenario can be combined into one module, or can be further split into multiple submodules.
[0101] The above serial numbers of this application are only for description and do not represent the advantages and disadvantages of the implementation scenarios. The above disclosure is only a few specific implementation scenarios of this application, but this application is not limited to them, and any changes that can be thought of by technicians in this field should fall within the scope of protection of this application.
Claims
1. A method for identifying load of industrial equipment, characterized in that: The method comprises: Obtaining total power load data of the user to be tested, wherein the total power load data includes active power, reactive power and current data; Performing load event detection on the total power load data, wherein the types of the load events include equipment startup and equipment shutdown; If a load event is detected in the total power load data, active power, reactive power and current data corresponding to the load event are extracted from the total power load data according to a preset sequence length to obtain a sequence of events to be tested, wherein the sequence of events to be tested includes a first steady-state sequence before the load event occurs, a second steady-state sequence after the load event occurs, and a load event sequence between the first steady-state sequence and the second steady-state sequence; according to the sequence of events to be tested, a load identification result is obtained based on a pre-trained load identification model, wherein the load identification result includes the equipment type and load event type corresponding to the load event; Among them, according to the event sequence to be tested, based on the pre-trained load identification model, the load identification result is obtained, including: determining the load event type according to the first steady-state sequence and the second steady-state sequence; if the load event type is load startup, then according to the first steady-state sequence, using the pre-trained time series prediction model, obtaining the first extended sequence corresponding to the first steady-state sequence, and calculating the difference between the event sequence to be tested and the first extended sequence to obtain a data difference sequence; if the load event type is load stop, then according to the second steady-state sequence, using the pre-trained time series prediction model, obtaining the second extended sequence corresponding to the second steady-state sequence, and calculating the difference between the event sequence to be tested and the second extended sequence to obtain a data difference sequence; calculating the active power, reactive power, fifth harmonic current and seventh harmonic current corresponding to the data difference sequence to obtain characteristic data, and according to the characteristic data, using the load identification model, determining the equipment type corresponding to the load event.
2. The method according to claim 1, characterized in that The determining of the load event type according to the first steady-state sequence and the second steady-state sequence specifically includes: Calculating a first current average value corresponding to the first steady-state sequence and a second current average value corresponding to the second steady-state sequence; If the first current average value is less than the second current average value, determining that the type of the load event is equipment startup; If the first current average value is greater than the second current average value, it is determined that the type of the load event is equipment shutdown.
3. The method according to claim 2, characterized in that The training method of the time series prediction model comprises: Acquire total active power, total reactive power and total current data corresponding to the simultaneous steady-state operation of multiple devices to obtain multiple steady-state data samples, wherein each of the steady-state data samples includes an observation sequence and a target sequence, and the last data of the observation sequence and the first data of the target sequence are adjacent data; The preset neural network model is trained with the observation sequence as input and the target sequence as output to obtain the time series prediction model.
4. The method according to claim 1, characterized in that: The training method of the load identification model comprises: Acquire multiple power load samples of multiple devices, wherein each of the power load samples includes active power, reactive power, fifth harmonic current, seventh harmonic current and device type corresponding to the device corresponding to the power load sample when a load event occurs; A neural network model based on an attention mechanism is constructed, the power load sample is used to train the neural network model to obtain a pre-trained model, and the parameters of the pre-trained model are optimized through a cross entropy loss function to obtain the load recognition model.
5. The method according to claim 1, characterized in that The training method of the load identification model comprises: Obtaining a total power load sample of multiple devices, and multiple sub-power load samples of multiple devices corresponding to the total power load sample, wherein the total power load sample includes the total active power, total reactive power, total fifth harmonic current, and total seventh harmonic current corresponding to the multiple devices corresponding to the total power load sample when power load events occur simultaneously, and each sub-power load sample includes the active power, reactive power, fifth harmonic current, seventh harmonic current, and device type corresponding to the device corresponding to the sub-power load sample when a load event occurs; Performing data alignment processing on the total power load sample and the sub-power load sample based on a least square fitting algorithm to synchronize the data of the total power load sample and each sub-power load sample; A neural network model based on an attention mechanism is constructed, and a joint feature is constructed using the total power load sample and the sub-power load sample to train the neural network model to obtain a pre-trained model, and the pre-trained model is optimized through a cross entropy loss function to obtain the load recognition model.
6. An industrial equipment load identification device, characterized in that: The device comprises: A data acquisition module is used to acquire total power load data of the user to be tested, wherein the total power load data includes active power, reactive power and current data; A load event detection module, used to perform load event detection on the total power load data, wherein the types of load events include equipment startup and equipment shutdown; A sequence extraction module, for extracting active power, reactive power and current data corresponding to the load event from the total power load data according to a preset sequence length if a load event is detected in the total power load data, to obtain a sequence of events to be tested, wherein the sequence of events to be tested includes a first steady-state sequence before the load event occurs, a second steady-state sequence after the load event occurs, and a load event sequence between the first steady-state sequence and the second steady-state sequence; A load identification module, configured to obtain a load identification result based on the pre-trained load identification model according to the event sequence to be tested, wherein the load identification result includes the equipment type and the load event type corresponding to the load event; Among them, according to the event sequence to be tested, based on the pre-trained load identification model, the load identification result is obtained, including: determining the load event type according to the first steady-state sequence and the second steady-state sequence; if the load event type is load startup, then according to the first steady-state sequence, using the pre-trained time series prediction model, obtaining the first extended sequence corresponding to the first steady-state sequence, and calculating the difference between the event sequence to be tested and the first extended sequence to obtain a data difference sequence; if the load event type is load stop, then according to the second steady-state sequence, using the pre-trained time series prediction model, obtaining the second extended sequence corresponding to the second steady-state sequence, and calculating the difference between the event sequence to be tested and the second extended sequence to obtain a data difference sequence; calculating the active power, reactive power, fifth harmonic current and seventh harmonic current corresponding to the data difference sequence to obtain characteristic data, and according to the characteristic data, using the load identification model, determining the equipment type corresponding to the load event.
7. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.
8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.
