Electrical load identification method, apparatus and device, and storage medium
By constructing an initial model and optimizing parameters, and utilizing Gaussian mixture distribution and hidden Markov model, the problem of low load identification results for industrial production lines under complex operating conditions was solved, achieving accurate identification of production line operating status and accurate analysis of load power consumption.
Patent Information
- Application Number
- CN202410925420.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-11
AI Technical Summary
Existing technologies have low load identification results in industrial production lines under complex operating conditions, making it difficult to achieve accurate load forecasting and demand response.
An initial model is constructed, and the model parameters are optimized by using a mixture of Gaussian distribution model and hidden Markov model. The target model is trained based on historical load data, and the target model is used to identify the current load status and energy distribution, so as to accurately identify the operating status of the production line.
It enables accurate identification of production line operating status under complex working conditions, improving the accuracy of load forecasting and demand response.
Smart Images

Figure CN118944051A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of power systems, and in particular to a method, device, equipment and storage medium for identifying power loads. Background Art
[0002] With the development of social economy, industrial electricity consumption has increased year by year, and the imbalance between power supply and demand has become increasingly severe. Although the widespread access of distributed photovoltaics provides backup support for the load gap, accurate and efficient orderly power consumption and demand response are still important means to ensure the safe and stable operation of the power grid, and the refined perception of industrial users' sub-item loads can support more accurate load forecasting and demand response plans.
[0003] Most of the related load identification technologies focus on the identification research of event detection and feature analysis in traditional signal analysis methods, such as industrial load identification methods based on random forests and steady-state waveforms; such as industrial event identification methods based on the mapping relationship between explanation space and feature space. The above-mentioned feature-based load identification relies on the correct detection of the "switch" events of equipment in the production line. It is suitable for industrial scenarios with stable production line equipment and a small total amount of equipment. It can realize event detection and type discrimination through feature increment thresholds. However, when identifying the load of production lines involving complex working conditions such as a large number of production line equipment and frequent equipment status switching, the load identification results are low due to the limitation of the identification dimension of the above-mentioned related load identification methods. Summary of the invention
[0004] The present invention provides a method, device, equipment and storage medium for identifying power loads, so as to at least solve the technical problem of low load identification results when identifying production line loads in complex working conditions in related technologies. The technical solution of the present invention is as follows:
[0005] According to a first aspect of an embodiment of the present invention, a method for identifying power load is provided, the method comprising: constructing initial models corresponding to a plurality of preset time periods respectively; the initial models characterize the correlation and correspondence between the total power of each production line and the operating states of each production line and the energy distribution of each corresponding production line in each preset time period; taking the maximum probability of each historical operating state of each production line under the historical total power of each production line as the goal, according to the historical load data of each production line in each preset time period, the model parameters of each initial model are optimized respectively, so as to obtain the target models corresponding to the plurality of preset time periods respectively; the historical load data comprises a historical active power sequence composed of a plurality of historical active powers of each production line in each preset time period and a historical operating state number, each historical active power under each historical active power ... The historical distribution probability of each historical operating state and the historical transition probability of operating state switching of each historical operating state of each production line in all production lines are obtained; the observed load data are obtained from the observed operation data of each production line in the preset time period before the current time, and the observed load data are input into the target model corresponding to the preset time period to which the preset time period before the current time belongs, so as to obtain each target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state; wherein, each target operating state is obtained with the goal of maximizing the probability of each operating state to which each production line belongs under the observed total power of each production line; the target energy distribution of each production line under each target operating state is obtained with the goal of maximizing the probability of active power to which each production line belongs under each target operating state of each production line.
[0006] The above-mentioned different multiple preset time periods constitute a time cycle.
[0007] Optionally, the multiple preset time periods are continuous in time, that is, the multiple continuous preset time periods constitute a time cycle.
[0008] Different multiple preset time periods can be obtained by dividing according to the correlation or similarity of the operating status of each production line within the time range being higher than the preset similarity, so that the correlation or similarity of the operating status of each production line in the same preset time period of different time periods is high, so as to ensure that the operating rules of the production lines represented by each target model are more accurate, thereby avoiding the influence of time factors.
[0009] In a possible implementation, the target model includes a mixed Gaussian distribution model and a hidden Markov model; the current load data is obtained from the current observed operating data of each production line at the current total power at the current time, including: calling the mixed Gaussian distribution model in the target model, clustering the current operating observation data of each production line respectively, and obtaining the number of multiple observed operating states included in each production line and the distribution probability of each production line being in each observed operating state under each observed active power; the multiple observed active powers of each production line constitute an observed active power sequence; using the hidden Markov model in the target model, based on the switching frequency of operating state switching between different observed operating states adjacent in time in each production line, the transition probability of operating state switching of each observed operating state of each production line in all production lines is determined.
[0010] The above clustering of the current operation observation data of each production line specifically includes extracting and clustering the current operation observation data belonging to each production line according to the equipment characteristics and / or power consumption characteristics and / or load identification of each production line.
[0011] Optionally, given that the hierarchical hidden Markov model can significantly improve the application effect and interpretability of the model in complex data analysis and prediction by introducing multi-level abstraction and representation capabilities, and can better handle multi-scale, multi-dimensional multi-production line operation observation data, the hidden Markov model is set to a hierarchical hidden Markov model to more accurately determine the probability of operating state switching from multiple dimensions of production state and switching state.
[0012] In the above implementation, given the high flexibility and powerful expressive power of the mixed Gaussian distribution model, it performs well in processing complex data distribution and various applications, and can more accurately identify the status of production lines in complex working conditions such as a large number of production line equipment and frequent equipment status switching.
[0013] In another possible implementation, the initial model includes a mixed Gaussian distribution model and a hidden Markov model; with the goal of maximizing the probability of each historical operating state to which each production line belongs under the historical total power of each production line, the model parameters of each initial model are optimized according to the historical load data of each production line in each preset time period, and before obtaining the target models corresponding to multiple preset time periods, the method also includes: obtaining the historical operating observation data of each production line in each preset time period; using the mixed Gaussian distribution model in the initial model, clustering the historical operating observation data of each production line respectively, to obtain the number of multiple historical operating states included in each production line and the distribution probability of each production line being in each historical operating state under each observed historical active power; multiple historical active powers of each production line constitute a historical active power sequence;
[0014] The hidden Markov model in the initial model is used to determine the transition probability of operating state switching in each historical operating state of all production lines based on the switching frequency of operating state switching between different historical operating states that are adjacent in time in each production line.
[0015] In another possible implementation, the mixed Gaussian distribution model specifically includes:
[0016]
[0017] Among them, B i Indicates any observation operation state x from 1 to m observation operation states i The probability distribution of different states corresponding to the observed active power; When the observed power of production line i at time t is When the running status is The probability of i,s and σ i,s The production line i is in the observation operation state. The mean power and standard deviation at .
[0018] In another possible implementation, the hidden Markov model specifically includes:
[0019]
[0020] in, represents the transition probability of the operating state switching of production line i in the hth hour; The production line i changes from the running state x from time t-1 to time t. t-1 i Transfer to running state The probability of i,h (s1,s2) is the observed active power sequence y of production line i i The running status of the time before and after is from x t-1 i Transfer to x t-1 i The number of N is the number of production line i in the observed power sequence y i The total number of observed operating states included above.
[0021] In another possible implementation, the initial model and the target model specifically further include:
[0022]
[0023] in, Indicates that the i-th production line starts from x in the h-th hourt-1 i Running status to The transition probability of the operating state, When production line i is in operation state at time t The total observed power is y t probability.
[0024] In another possible implementation, the initial model and the target model specifically further include:
[0025]
[0026] in, At time t, production line i is in the running state The observed active power is probability.
[0027] According to a second aspect of an embodiment of the present invention, there is provided an apparatus for identifying an electric load, the apparatus comprising: a construction unit for constructing initial models corresponding to a plurality of preset time periods; the initial models characterize the correlation and correspondence between the total power of each production line and the operating states of each production line and the energy distribution of each corresponding production line in each preset time period; a training unit for optimizing the model parameters of each initial model according to the historical load data of each production line in each preset time period with the goal of maximizing the probability of each historical operating state of each production line under the historical total power of each production line, so as to obtain target models corresponding to the plurality of preset time periods; the historical load data comprises a historical active power sequence composed of a plurality of historical active powers of each production line in each preset time period, and a historical active power sequence comprising the number of historical operating states, each historical active power under each historical active power. The historical distribution probability of each historical operating state under the power and the historical transition probability of operating state switching of each historical operating state of each production line in all production lines; an identification unit is used to obtain observed load data from the observed operation data of each production line in the preset time period before the current time, and input the observed load data into the target model corresponding to the preset time period to which the preset time period before the current time belongs, so as to obtain each target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state; wherein each target operating state is obtained with the goal of maximizing the probability of each operating state to which each production line belongs under the observed total power of each production line; and the target energy distribution of each production line under each target operating state is obtained with the goal of maximizing the probability of active power to which each production line belongs under each target operating state of each production line.
[0028] According to a third aspect of an embodiment of the present invention, there is provided a power load identification system configured to execute the power load identification method of the first aspect and any possible implementation thereof.
[0029] According to a fourth aspect of an embodiment of the present invention, an electric power device is provided, comprising: a processor and a memory for storing processor executable instructions; wherein the processor is configured to execute executable instructions to implement a method for identifying an electric load as in the first aspect and any possible implementation thereof.
[0030] According to a fifth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which instructions are stored. When the instructions in the computer-readable storage medium are executed by a processor of an electric power device, the electric power device is enabled to execute a method for identifying an electric load such as the first aspect and any possible implementation thereof.
[0031] According to a sixth aspect of an embodiment of the present application, a computer program product is provided, the computer program product comprising computer instructions, which, when executed on an electric power device, enables the electric power device to execute the electric power load identification method of the first aspect and any possible implementation thereof.
[0032] The technical solution provided by the embodiment of the present invention brings at least the following beneficial effects: first, the correlation and operation law between the total power of the production line and the various operating states of each production line and the corresponding energy distribution of each production line are constructed to obtain an initial model. Then, the probability of each historical operating state to which each production line belongs under the historical total power of each production line is maximized, and the model parameters of the initial model are optimized and trained based on the historical load data of multiple different preset time periods to obtain each target model matching different preset time periods. At the same time, the target model is obtained by optimizing the model parameters with the maximum probability of each historical operating state as the goal, so that the total power of the production line represented by the target model and the various operating states of each production line are the results of the maximum probability of the simultaneous occurrence and matching of the parameters of the two, thereby ensuring the accuracy of the target model in identifying the operating state. Using the target model matching the preset time period matching the current time (that is, the target model corresponding to the preset time period before the current time and close to the current time), the target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state are determined, so as to intelligently realize the identification of the operating state of each production line and the load power consumption status.
[0033] Through the above-mentioned power load identification method, according to the target model, based on the number of each operating state and the distribution probability of each operating state under each observed active power and the transition probability of operating state switching of each production line, the target operating state of each production line can be accurately obtained with the probability of maximizing the operating state of each production line under the observed total power of each production line as the goal, so as to ensure the accurate identification of the operating state of each production line under different complexity scenarios; at the same time, it is also possible to accurately obtain the target energy distribution of each production line under each target operating state with the probability of maximizing the active power of each production line under each target operating state of each production line as the goal, so as to ensure the accurate identification of the power load of each production line under different complexity scenarios.
[0034] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application, and do not constitute an improper limitation on the present application.
[0036] Figure 1 is a flow chart of a method for identifying an electric load according to an exemplary embodiment;
[0037] Figure 2 is a schematic block diagram of a power load identification device according to an exemplary embodiment;
[0038] Figure 3 The diagram is a schematic diagram of an electric power device according to an exemplary embodiment. DETAILED DESCRIPTION
[0039] In order to enable ordinary persons in the art to better understand the technical solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.
[0040] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims.
[0041] Before introducing in detail the power load identification method provided in the embodiment of the present application, a brief introduction to the application scenarios involved in the embodiment of the present application is first given.
[0042] With the development of social economy, industrial electricity consumption has increased year by year, and the imbalance between power supply and demand has become increasingly severe. Although the widespread access of distributed photovoltaics provides backup support for the load gap, accurate and efficient orderly power consumption and demand response are still important means to ensure the safe and stable operation of the power grid, and the refined perception of industrial users' sub-item loads can support more accurate load forecasting and demand response plans. Non-intrusive load identification technology can obtain the usage of various types of user equipment at a low cost, realize online monitoring and analysis of power loads, and is of great significance to supporting the development of applications such as load forecasting and demand response.
[0043] Load identification technology is applied to residential electricity analysis. Benefiting from the concentrated characteristics of residential electrical equipment, similar usage patterns and easy data collection, load decomposition strategies have gradually focused on representation learning methods under various deep neural network models. At the same time, in order to increase the generalization and ease of use of such supervised learning algorithms, emerging algorithms such as transfer learning and reinforcement learning have also become the main direction of residential load identification research. However, the industrial load identification application scenarios involve a wide range of user types, a wide range of equipment types, and complex equipment operation patterns. There are significant differences between them and residential load identification scenarios, and it is difficult to have a large-scale data-driven network learning method implementation foundation. Most of the related methods focus on the identification research of event detection and feature analysis in traditional signal analysis methods: for example, industrial load identification methods based on random forests and steady-state waveforms; for example, industrial event identification methods based on the mapping relationship between explanation space and feature space. The above-mentioned feature-based method relies on the correct detection of the "switch" events of the equipment. Generally speaking, for industrial scenarios with stable equipment operation and a small total amount of equipment, event detection and type discrimination can be achieved through feature increment thresholds. However, for production line type identification scenarios involving a large number of electrical appliances, frequent equipment switching, and a small load magnitude, the algorithm needs to be deeply customized to adapt to various complex working conditions, and there are certain limitations.
[0044] According to research, in residential load identification applications, the Hidden Markov Model (HMM) is the most widely used analysis method before the popularization of deep network identification algorithms. It uses the maximum expectation value algorithm to find the best hidden state sequence, thereby realizing the identification and decomposition of power loads. It has the ability to handle the simultaneous occurrence of complex switching events and the characteristics of relatively small data collection requirements. It is one of the preferred methods for complex industrial load identification. A single production line contains a complex number of equipment and a complex production process. If a single Gaussian distribution is used to describe it, there will be certain errors in the probability-driven state judgment in the overlapping scenario of multiple production lines. In addition, there is a certain impact on energy decomposition. Therefore, it is possible to consider clustering the load of a single production line based on a Gaussian mixture distribution, and then conduct learning and training to achieve the distinction between different operating states of a single production line.
[0045] Therefore, the present application provides a method for identifying power loads, first constructing the correlation and operating rules between the total power of the production line and the various operating states of each production line and the corresponding energy distribution of each production line to obtain an initial model. Then, with the maximum probability of each historical operating state of each production line under the historical total power of each production line as the goal, the model parameters of the initial model are optimized and trained based on the historical load data of multiple different preset time periods, and each target model matching different preset time periods is obtained. At the same time, the target model is obtained by optimizing the model parameters with the maximum probability of each historical operating state as the goal, so that the total power of the production line represented by the target model and the various operating states of each production line are the results of the maximum probability of the simultaneous occurrence and matching of the two parameters, thereby ensuring the accuracy of the target model in identifying the operating state. Using the target model that matches the preset time period matching the current time (that is, the target model corresponding to the preset time period before the current time and close to the current time), the target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state are determined, so as to intelligently realize the identification of the operating state of each production line and the load power consumption status.
[0046] The power load identification method provided in the embodiment of the present application can be applied to power equipment used for power load identification. For ease of understanding, the power load identification method provided in the present application is specifically introduced below in conjunction with the accompanying drawings.
[0047] Figure 1 is a flow chart of a method for identifying a power load according to an exemplary embodiment. Figure 1 As shown, the power load identification method includes the following steps.
[0048] S11, constructing initial models corresponding to multiple preset time periods respectively.
[0049] The initial model represents the correlation and correspondence between the total power of each production line and each operating state of each production line and the corresponding energy distribution of each production line in each preset time period.
[0050] The above-mentioned different multiple preset time periods constitute a time cycle.
[0051] Optionally, the multiple preset time periods are continuous in time, that is, the multiple continuous preset time periods constitute a time cycle.
[0052] Different multiple preset time periods can be obtained by dividing according to the correlation or similarity of the operating status of each production line within the time range being higher than the preset similarity, so that the correlation or similarity of the operating status of each production line in the same preset time period of different time periods is high, so as to ensure that the operating rules of the production lines represented by each target model are more accurate, thereby avoiding the influence of time factors.
[0053] The operating status of the above production line can be understood as the start-up status and shutdown status of each device on the production line.
[0054] S12, with the goal of maximizing the probability of each historical operating state of each production line under the historical total power of each production line, optimize the model parameters of each initial model according to the historical load data of each production line in each preset time period, and obtain the target models corresponding to multiple preset time periods.
[0055] The historical load data includes a historical active power sequence composed of multiple historical active powers belonging to each production line in each preset time period, as well as the number of historical operating states under each historical active power, the historical distribution probability of each historical operating state under each historical active power, and the historical transfer probability of operating state switching for each historical operating state of each production line in all production lines.
[0056] In one embodiment, the initial model includes a Gaussian mixture distribution model (also called a Gaussian mixture model (GMM model)) and a hidden Markov model (Hidden Markov Model, HMM model); with the goal of maximizing the probability of each historical operating state to which each production line belongs under the historical total power of each production line, according to the historical load data of each production line in each preset time period, respectively optimizing the model parameters of each initial model, before obtaining the target models corresponding to multiple preset time periods, the method also includes: obtaining the historical operation observation data of each production line in each preset time period; using the mixed Gaussian distribution model in the initial model, respectively clustering the historical operation observation data of each production line, to obtain the number of multiple historical operating states included in each production line and the distribution probability of each production line being in each historical operating state under each observed historical active power; multiple historical active powers of each production line constitute a historical active power sequence; using the hidden Markov model in the initial model, based on the switching frequency of operating state switching between different historical operating states adjacent in time in each production line, determining the transition probability of operating state switching of each historical operating state of each production line in all production lines.
[0057] S13, obtaining observed load data from the observed operation data of each production line in the preset time period before the current time, and inputting the observed load data into the target model corresponding to the preset time period to which the preset time period before the current time belongs, and obtaining each target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state.
[0058] The target energy distribution of each of the above production lines is the target power corresponding to each production line.
[0059] Among them, each target operating state is obtained by maximizing the probability of each operating state of each production line under the observed total power of each production line; the target energy distribution of each production line under each target operating state is obtained by maximizing the probability of active power of each production line under each target operating state of each production line.
[0060] In one embodiment, the target model includes a mixed Gaussian distribution model and a hidden Markov model; the current load data is obtained from the current observed operation data of each production line at the current total power at the current time, including: calling the mixed Gaussian distribution model in the target model, clustering the current operation observation data of each production line respectively, and obtaining the number of multiple observed operation states included in each production line and the distribution probability of each production line being in each observed operation state under each observed active power; the multiple observed active powers of each production line constitute an observed active power sequence; using the hidden Markov model in the target model, based on the switching frequency of operation state switching between different observed operation states adjacent in time in each production line, the transition probability of operation state switching of each observed operation state of each production line in all production lines is determined.
[0061] The above clustering of the current operation observation data of each production line specifically includes extracting and clustering the current operation observation data belonging to each production line according to the equipment characteristics and / or power consumption characteristics and / or load identification of each production line.
[0062] Optionally, given that the hierarchical hidden Markov model can significantly improve the application effect and interpretability of the model in complex data analysis and prediction by introducing multi-level abstraction and representation capabilities, and can better handle multi-scale, multi-dimensional multi-production line operation observation data, the hidden Markov model is set to a hierarchical hidden Markov model (Factorial Hidden Markov Model, FHMM model) to more accurately determine the probability of operating state switching from multiple dimensions of production state and switching state.
[0063] In the above implementation, given that the mixed Gaussian distribution model is highly flexible and has powerful expressive power, it performs well in processing complex data distribution and various applications, and can more accurately identify the status of production lines in complex working conditions such as a large number of production line equipment and frequent equipment status switching.
[0064] Optionally, for continuously changing active power, Gaussian distribution is used to describe the probability distribution (ie, distribution probability) of observations in the running state and the stopped state. The mixed Gaussian distribution model in the above target model and the initial model specifically includes a model represented by the following formula 1.
[0065]
[0066] Among them, B i Indicates any observation operation state x from 1 to m observation operation states i The probability distribution of different states corresponding to the observed active power; When the observed power of production line i at time t is When the running status is The probability of i,s and σ i,s The production line i is in the observation operation state. The mean power and standard deviation at .
[0067] Optionally, the hidden Markov model in the above target model and initial model specifically includes a model represented by the following formula 2.
[0068]
[0069] in, represents the transition probability of the operating state switching of production line i in the hth hour; The production line i changes from the running state x from time t-1 to time t. t-1 i Transfer to running state The probability of i,h (s1,s2) is the observed active power sequence y of production line i i The running status of the time before and after is from x t-1 i Transfer to x t-1 i The number of N is the number of production line i in the observed power sequence y i The total number of observed operating states included above.
[0070] Optionally, based on the determination of the initial model and related model parameters in the above embodiment, the hidden operation state of the production line corresponding to the total power curve (hereinafter referred to as the operation state) can be further analyzed. Since the current state in the Markov chain model is only related to the previous state, the problem of determining the production line operation state can be described as follows: In the scenario where the equipment operation state is known at time t-1, the total power at time t is y t The optimal state sequence of each electrical appliance at time t is solved, and based on the probability of each production line in various operating states at time t, the operating state corresponding to the maximum probability is taken as the current state (i.e., the target state). Therefore, the above target model and initial model specifically also include the model represented by the following (Formula 3).
[0071]
[0072] in, Indicates that the i-th production line starts from x in the h-th hour t-1 i Running status to The transition probability of the operating state, When production line i is in operation state at time t The total observed power is y t probability.
[0073] Optionally, the above initial model and target model specifically further include a model represented by the following formula 4:
[0074]
[0075] in, At time t, production line i is in the running state The observed active power is probability.
[0076] Optionally, due to the complexity of the solution, in practical applications, the mean value of the power distribution of each production line in the current operating state (i.e., the target operating state) is generally used as the target power of the production line at the current moment in the target operating state. Further, the mean value of the Gaussian distribution of the active power of each production line in the current operating state (i.e., the target operating state) can be selected as the target power of the production line at the current moment in the target operating state.
[0077] In some embodiments, clustering is performed by taking active power and reactive power as examples among the observed operating parameters, and the probability density function of the Gaussian distribution is shown in the following formula (5).
[0078]
[0079] Among them, K is the number of operating states of a single production line, that is, the number of clusters; α k is the prior probability that the data point belongs to the kth Gaussian. The initial value only needs to satisfy all α k The sum is 1; p(x,y|μ 1,k ,σ 1,k ,μ 2,k ,σ 2,k ) is the probability density of the kth Gaussian distribution; μ 1,k , σ 1,k is the mean and variance of the active power of the single production line under the kth working state; μ 2,k , σ 2,k is the mean and variance of reactive power under the kth working state of a single production line. In cluster analysis, K is a given value, and other parameters are estimated by the expectation-maximization (EM) algorithm. Since the actual production line usually cannot give a specific number of operating states, a clustering evaluation algorithm is used to regard the sum of squared errors between clusters as a function of the number of clusters K, and locate the optimal K value at the "inflection point" of the curve to complete clustering based on the mixed Gaussian distribution model.
[0080] The three steps of the FHMM model construction process based on GMM clustering are basically the same. The probability distribution of the observation quantity of the production line equipment in the start and stop state in formula (1) becomes the probability distribution of the observation quantity in multiple different states. When performing load decomposition, the states contained in a single device in formula (2) and formula (3) are added synchronously to realize the FHMM load decomposition of a single production line in multiple states. Finally, by adding the pre-statistical prior information of the production line operation law to the FHMM model, different state transition probabilities are set in different time sections to improve the accuracy of actual load decomposition.
[0081] Through the above implementation, the correlation and operation law between the total power of the production line and each operating state of each production line and the corresponding energy distribution of each production line are first constructed to obtain an initial model. Then, the target is to maximize the probability of each historical operating state of each production line under the historical total power of each production line, and the model parameters of the initial model are optimized and trained based on the historical load data of multiple different preset time periods to obtain each target model matching different preset time periods. At the same time, the target model is obtained by optimizing the model parameters with the maximum probability of each historical operating state as the target, so that the total power of the production line represented by the target model and the operating state of each production line are the results of the maximum probability of the two parameters occurring simultaneously and matching, thereby ensuring the accuracy of the target model in identifying the operating state. Using the target model that matches the preset time period matching the current time (that is, the target model corresponding to the preset time period before the current time and close to the current time), the target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state are determined to intelligently realize the identification of the operating state of each production line and the load power consumption status.
[0082] Therefore, according to the target model, based on the number of each operating state and the distribution probability of each operating state under each observed active power and the transition probability of operating state switching of each production line, the target operating state of each production line can be accurately obtained with the probability of maximizing the operating state of each production line under the observed total power of each production line as the goal, so as to ensure the accurate identification of the operating state of each production line under different complexity scenarios; at the same time, the target energy distribution of each production line under each target operating state can also be accurately obtained with the probability of maximizing the active power of each production line under each target operating state of each production line as the goal, so as to ensure the accurate identification of the power load of each production line under different complexity scenarios.
[0083] In order to realize the above functions, the power load identification device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0084] The present application also provides a method Figure 2 The power load identification device shown includes: a construction unit 201, a training unit 202 and an identification unit 203.
[0085] The construction unit 201 is used to construct initial models corresponding to multiple preset time periods respectively; the initial models represent the correlation and correspondence between the total power of each production line and each operating state of each production line and the corresponding energy distribution of each production line in each preset time period.
[0086] The training unit 202 is used to optimize the model parameters of each initial model according to the historical load data of each production line in each preset time period with the goal of maximizing the probability of each historical operating state of each production line under the historical total power of each production line, so as to obtain the target models corresponding to multiple preset time periods; the historical load data includes a historical active power sequence composed of multiple historical active powers of each production line in each preset time period and the number of historical operating states under each historical active power, the historical distribution probability of each historical operating state under each historical active power, and the historical transition probability of operating state switching of each historical operating state of each production line in all production lines.
[0087] The identification unit 203 is used to obtain observed load data from the observed operation data of each production line within a preset time period before the current time, and input the observed load data into the target model corresponding to the preset time period to which the preset time period before the current time belongs, so as to obtain each target operation state of each production line under the current total power and the target energy distribution of each production line under each target operation state; wherein each target operation state is obtained by maximizing the probability of each operation state to which each production line belongs under the observed total power of each production line; and the target energy distribution of each production line under each target operation state is obtained by maximizing the probability of active power to which each production line belongs under each target operation state of each production line.
[0088] In one possible implementation, the target model includes a mixed Gaussian distribution model and a hidden Markov model; the identification unit 203 is specifically used to: call the mixed Gaussian distribution model in the target model, cluster the current operation observation data of each production line respectively, and obtain the number of multiple observed operating states included in each production line and the distribution probability of each production line being in each observed operating state under each observed active power; the multiple observed active powers of each production line constitute an observed active power sequence; use the hidden Markov model in the target model to determine the transition probability of operating state switching for each observed operating state of each production line in all production lines based on the switching frequency of operating state switching between different observed operating states that are adjacent in time in each production line.
[0089] In another possible implementation, the initial model includes a mixed Gaussian distribution model and a hidden Markov model; the training unit 202 is also used to: obtain historical operation observation data of each production line in each preset time period; use the mixed Gaussian distribution model in the initial model to cluster the historical operation observation data of each production line respectively, and obtain the number of multiple historical operating states included in each production line and the distribution probability of each production line being in each historical operating state under each observed historical active power; the multiple historical active powers of each production line constitute a historical active power sequence; use the hidden Markov model in the initial model to determine the transition probability of operating state switching for each historical operating state of each production line in all production lines based on the switching frequency of operating state switching between different historical operating states that are adjacent in time in each production line.
[0090] In another possible implementation, the mixed Gaussian distribution model specifically includes:
[0091]
[0092] Among them, B i Indicates any observation operation state x from 1 to m observation operation states i The probability distribution of different states corresponding to the observed active power; When the observed power of production line i at time t is When the running status is The probability of i,s and σ i,s The production line i is in the observation operation state. The mean power and standard deviation at .
[0093] In another possible implementation, the hidden Markov model specifically includes:
[0094]
[0095] in, represents the transition probability of the operating state switching of production line i in the hth hour; The production line i changes from the running state x from time t-1 to time t. t-1 i Transfer to running state The probability of i,h (s1,s2) is the observed active power sequence y of production line i i The running status of the time before and after is from x t-1 i Transfer to x t-1 i The number of N is the number of production line i in the observed power sequence y i The total number of observed operating states included above.
[0096] In another possible implementation manner, the initial model and the target model specifically further include:
[0097]
[0098] in, Indicates that the i-th production line starts from x in the h-th hour t-1 i Running state to x t i The transition probability of the operating state, When production line i is in operation state at time t The total observed power is y t probability.
[0099] In another possible implementation manner, the initial model and the target model specifically further include:
[0100]
[0101] in, At time t, production line i is in the running state The observed active power is probability.
[0102] Regarding the device in the above embodiment, the specific manner in which each unit module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.
[0103] Figure 3 is a schematic diagram of an electric power device provided by this application. Figure 3The power device 60 may include at least one processor 601 and a memory 603 for storing processor executable instructions. The processor 601 is configured to execute instructions in the memory 603 to implement the power load identification method in the following embodiments.
[0104] In addition, the power device 60 may further include a communication bus 602 , at least one communication interface 604 , an input device 606 , and an output device 605 .
[0105] The processor 601 may be a central processing unit (CPU), a microprocessing unit, an ASIC, or one or more integrated circuits for controlling the execution of the program of the present application.
[0106] Communication bus 602 may include a pathway for transmitting information between the above-mentioned components.
[0107] The communication interface 604 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.
[0108] The input device 606 is used to receive input signals and the output device 605 is used to output signals.
[0109] The memory 603 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory may exist independently and be connected to the processing unit through a bus. The memory may also be integrated with the processing unit.
[0110] The memory 603 is used to store instructions for executing the solution of the present application, and the execution is controlled by the processor 601. The processor 601 is used to execute the instructions stored in the memory 603, so as to realize the functions in the method of the present application.
[0111] In a specific implementation, as an embodiment, the processor 601 may include one or more CPUs, such as Figure 3 CPU0 and CPU1 in.
[0112] In a specific implementation, as an embodiment, the power device 60 may include multiple processors, such as Figure 3 601 and processor 607 in FIG. Each of these processors may be a single-CPU processor or a multi-CPU processor. A processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).
[0113] The electrical equipment Figure 3 The system includes: a processor 601 and a memory 603 for storing executable instructions of the processor 601; wherein the processor 601 is configured to execute the executable instructions to implement the power load identification method of any possible implementation method described above. And the same technical effect can be achieved, so it will not be repeated here to avoid repetition.
[0114] The embodiment of the present application also provides a computer-readable storage medium, when the instructions in the computer-readable storage medium are executed by the processor of the control device or control equipment, the control device or control equipment can execute the power load identification method as any of the possible implementations described above. And the same technical effect can be achieved, so it will not be repeated here to avoid repetition.
[0115] The embodiment of the present application also provides a computer program product, including a computer program or instruction, which is executed by a processor to perform a method for identifying a power load in any of the above possible implementations. The same technical effect can be achieved, and to avoid repetition, it will not be described here.
[0116] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.
[0117] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.
Claims
1. A method for identifying power load, characterized in that: The method comprises: Constructing initial models corresponding to a plurality of preset time periods respectively; the initial models represent the correlation and correspondence between the total power of each production line and each operating state of each production line and the energy distribution of each production line in each preset time period; Taking the maximum probability of each historical operating state of each production line under the historical total power of each production line as the goal, according to the historical load data of each production line in each preset time period, the model parameters of each initial model are optimized respectively, so as to obtain the target models corresponding to the multiple preset time periods respectively; the historical load data include a historical active power sequence composed of multiple historical active powers of each production line in each preset time period and the number of historical operating states under each historical active power, the historical distribution probability of each historical operating state under each historical active power, and the historical transition probability of operating state switching of each historical operating state of each production line in all production lines; Observed load data is obtained from observed operation data of each production line within a preset time period before the current time and with the minimum time interval between the current time, and the observed load data is input into the target model corresponding to the preset time period before the current time and with the minimum time interval between the current time, so as to obtain each target operating state of each production line under the current total power and the target energy distribution of each production line under each target operating state; wherein each target operating state is obtained with the goal of maximizing the probability of each operating state to which each production line belongs under the observed total power of each production line; and the target energy distribution of each production line under each target operating state is obtained with the goal of maximizing the probability of active power to which each production line belongs under each target operating state of each production line.
2. The method for identifying power load according to claim 1, characterized in that: The target model includes a mixed Gaussian distribution model and a hidden Markov model; the current load data is obtained from the current observed operation data of each production line at the current total power at the current time, including: The mixed Gaussian distribution model in the target model is called to cluster the current operation observation data of each production line respectively, to obtain the number of multiple observed operation states included in each production line and the distribution probability of each production line being in each observed operation state under each observed active power; the multiple observed active powers of each production line constitute an observed active power sequence; The hidden Markov model in the target model is used to determine the transition probability of operating state switching of each observed operating state of each production line in all production lines based on the switching frequency of operating state switching between different observed operating states adjacent in time in each production line.
3. The method for identifying power load according to claim 1, characterized in that: The initial model includes a mixed Gaussian distribution model and a hidden Markov model; before the target is to maximize the probability of each historical operating state of each production line under the historical total power of each production line, and optimize the model parameters of each initial model according to the historical load data of each production line in each preset time period, and obtain the target models corresponding to the multiple preset time periods, the method also includes: Obtaining historical operation observation data of each of the production lines in each of the preset time periods; The mixed Gaussian distribution model in the initial model is used to cluster the historical operation observation data of each production line respectively, to obtain the number of the multiple historical operation states respectively included in each production line and the distribution probability of each production line being in each historical operation state under the observed historical active power; the multiple historical active powers of each production line constitute a historical active power sequence; The hidden Markov model in the initial model is used to determine the transition probability of operating state switching of each historical operating state of each production line in all production lines based on the switching frequency of operating state switching between different historical operating states adjacent in time in each production line.
4. The method for identifying power load according to claim 2 or 3, characterized in that: The mixed Gaussian distribution model specifically includes: Among them, B i Indicates any observation operation state x from 1 to m observation operation states i The probability distribution of different states corresponding to the observed active power; P(y t i |x t i ) is the observed power of production line i at time t is y t i When the running status is x t i The probability of i,s and σ i,s The production line i is in the observed operation state x t i The mean power and standard deviation at .
5. The method for identifying power load according to claim 2 or 3, characterized in that: The hidden Markov model specifically includes: Among them, A i h represents the transition probability of production line i switching its operating state within the hth hour; P(x t i |x t-1 i ) is the time from when production line i changes from the running state x to when t-1. t-1 i Transfer to running state x t i The probability of i,h (s1,s2) is the observed active power sequence y of production line i i The running status of the time before and after is from x t-1 i Transfer to x t-1 i The number of N is the number of production line i in the observed power sequence y i The total number of observed operating states included above.
6. The method for identifying power load according to claim 2 or 3, characterized in that: The target model specifically includes: Among them, P(x t i |x t-1 i ) means that the i-th production line starts from x in the h-th hour. t-1 i Running state to x t i The transition probability of the operating state, P(y t |x t i ) is when production line i is in operating state x at time t t i The total observed power is y t probability.
7. The method for identifying power load according to claim 6, characterized in that: The target model specifically includes: Among them, P(y t i |x t i ) is the production line i at time t when the running state is x t i The observed active power is y t i probability.
8. An electrical load identification device, characterized in that: The device comprises: A construction unit, used to construct initial models corresponding to a plurality of preset time periods; the initial models represent the correlation and correspondence between the total power of each production line and each operating state of each production line and the energy distribution of each corresponding production line in each preset time period; A training unit is used to optimize the model parameters of each initial model according to the historical load data of each production line in each preset time period with the goal of maximizing the probability of each historical operating state of each production line under the historical total power of each production line, so as to obtain the target models corresponding to the multiple preset time periods; the historical load data includes a historical active power sequence composed of multiple historical active powers of each production line in each preset time period and the number of historical operating states under each historical active power, the historical distribution probability of each historical operating state under each historical active power, and the historical transition probability of operating state switching of each historical operating state of each production line in all production lines; An identification unit is used to obtain observed load data from the observed operation data of each production line within a preset time period before the current time and with the minimum time interval between the current time, and input the observed load data into the target model corresponding to the preset time period before the current time and with the minimum time interval between the current time, so as to obtain the target operating states of each production line under the current total power and the target energy distribution of each production line under the target operating states; wherein the target operating states are obtained by maximizing the probability of each operating state to which each production line belongs under the observed total power of each production line; and the target energy distribution of each production line under the target operating states is obtained by maximizing the probability of active power to which each production line belongs under the target operating states of each production line.
9. An electric power device, comprising computer instructions, which, when executed on the electric power device, enable the electric power device to execute the electric load identification method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor, the method for identifying power load according to any one of claims 1 to 7 can be executed.
Citation Information
Patent Citations
Non-intrusive power load decomposition and identification method and device, electronic equipment and storage medium
CN114330636A
Industrial equipment load identification method and device, storage medium and computer equipment
CN117543592A
Hybrid energy storage configuration method and system based on micro-grid wind and light storage system
CN117543706A
Enterprise direct carbon emission estimation method based on non-intrusive load decomposition
CN118114882A
Evaluating the Electrical Load Capacity of an Electrical Power Distribution Network
DE102016218352A1
