A power load control method and system based on artificial intelligence

Abstract

The present invention discloses a power load control method and system based on artificial intelligence, and relates to the technical field of power load. A power load control system based on artificial intelligence includes: a power load data acquisition module, a power load prediction control module and a power load abnormality control module. The present invention monitors the status and parameters of the load device in real time, and the system can quickly respond to load changes to ensure that the power load is controlled in the best state, thereby improving the response speed and control accuracy of the system; the load device hierarchical control strategy library established by the optimized clustering algorithm can implement differentiated control strategies for different types of load devices, thereby improving the flexibility and effectiveness of power load management; the use of the power load prediction model can predict potential abnormal situations and deploy abnormal parameter control strategies in advance, thereby achieving prevention and rapid processing of abnormal situations.

Description

Technical Field

[0001] The present invention relates to the field of power load technology, and in particular to a power load control method and system based on artificial intelligence. Background Art

[0002] In the prior art, one of the main challenges facing power load control is how to maintain a stable output of the power load when an abnormal situation occurs in the load device. Although the prior art can perform targeted power load control when an abnormal situation occurs, in some cases, it may not be able to respond to rapidly changing load demands in a timely manner, resulting in voltage instability and affecting the normal operation of the power supply device. For example, when the load current undergoes rapid transient changes, if the change speed exceeds the response capability of the regulator, the regulator may not be able to maintain the stability of the output voltage within the specified tolerance range. Summary of the invention

[0003] The present invention aims to provide a power load control method and system based on artificial intelligence, which can predict abnormal situations and automatically adjust the control parameters of the power load to ensure the stable operation of the power load equipment.

[0004] A power load control method based on artificial intelligence, comprising:

[0005] S1. Obtain the real-time load information of the target power load, which includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ;

[0006] S2. For the current load device X n Perform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance;

[0007] S3. Record the abnormal judgment results R of all current load devices n The abnormality in the middle corresponds to the current load device X n , and the total number of abnormal load devices is obtained; special control is implemented on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function.

[0008] As a preferred technical solution of the present invention, the power load classification control model in step S1 includes a feature extraction layer, a feature matching layer and a result output layer;

[0009] The feature extraction layer is used to analyze the basic parameters C of the current load device. n Perform feature extraction to obtain the basic parameter feature C of the load device n ';

[0010] The feature matching layer is used to match the basic parameter characteristics of the load device C n 'Perform strategy matching to obtain load device hierarchical control strategy label Q n ; According to the load equipment classification control strategy label Q n Matching is performed in the load device hierarchical control strategy library to obtain the load device hierarchical control strategy L n ;

[0011] The result output layer is used to output the load device hierarchical control strategy L n ;

[0012] The specific steps for training the feature matching layer include:

[0013] Collecting several groups of load equipment parameter feature training samples, each group of load equipment parameter feature training samples includes a group of training load equipment basic parameter features with target values, and the target values ​​are hierarchical control strategy labels; combining several groups of load equipment parameter feature training samples to obtain a load equipment parameter feature training set;

[0014] The load device parameter feature training set is input into the feature matching layer for training with the target value as the target to obtain the initial feature matching layer; the initial feature matching layer is model evaluated to obtain the initial feature matching layer model evaluation result; if the initial feature matching layer model evaluation result is passed, the initial feature matching layer is used as the trained feature matching layer; otherwise, the load device parameter feature training set is used to continue model training.

[0015] As a preferred technical solution of the present invention, the specific steps of establishing a load device hierarchical control strategy library based on the optimized clustering algorithm in step S1 include:

[0016] Collect M load equipment parameter characteristics T m , m=1, 2, ..., M; M load equipment parameter characteristics T m Abstracted as a sample point set, we get the load equipment parameter feature sample point set A={T1', T2', ..., T M '}; Establish a weighted undirected graph for the load device parameter feature sample point set to obtain a classification graph of the load device to be divided; establish a weighted matrix of the load device to be divided according to the classification graph of the load device to be divided;

[0017] Calculate the sample point T m The local density B m and distance O m ; Using formula E m =B m *O m Calculate the sample point weight value E m ;

[0018] Use the fast search algorithm to calculate the sample point weight value E m Sample points T whose values ​​are greater than the preset dominant threshold m ', and the number of clusters is obtained; the value of the number of clusters is D;

[0019] Perform data clustering based on the number of clusters and the weight matrix of the load equipment to be divided, and obtain D load equipment cluster centers;

[0020] The load equipment cluster center is used as the classification label reference value, and a classification control strategy is established according to the classification label reference value;

[0021] Traverse all load equipment cluster centers to establish a load equipment hierarchical control strategy library.

[0022] As a preferred technical solution of the present invention, the power load prediction model in step 2 includes a parameter feature extraction layer, a prediction parameter analysis layer and a prediction result output layer;

[0023] The parameter feature extraction layer is used to extract the load device parameters F to be analyzed. nPerform feature extraction to obtain the load equipment parameter feature F to be analyzed n ';

[0024] The prediction parameter analysis layer is used to predict the load device parameter characteristics F n 'Perform forecast analysis and obtain the forecast load equipment parameter Y n ;

[0025] The prediction result output layer is used to predict the load equipment parameters Y n and load equipment hierarchical control strategy L n For the current load device X n Make a judgment and get the load equipment abnormality judgment result R n ;

[0026] The specific steps for training the prediction parameter analysis layer include:

[0027] Collecting a plurality of sets of prediction parameter training sample pairs; each set of prediction parameter training sample pairs includes an input value and an output value; combining the plurality of sets of prediction parameter training sample pairs to obtain a prediction parameter training set;

[0028] The prediction parameter training set is input into the prediction parameter analysis layer for training with the output value as the target to obtain the initial prediction parameter analysis layer; the model of the initial prediction parameter analysis layer is evaluated to obtain the model evaluation result of the initial prediction parameter analysis layer; if the model evaluation result of the initial prediction parameter analysis layer is passed, the initial prediction parameter analysis layer is used as the trained prediction parameter analysis layer; otherwise, the model training is continued using the prediction parameter training set.

[0029] As a preferred technical solution of the present invention, according to the predicted load device parameter Y n and load equipment hierarchical control strategy L n For the current load device X n The specific steps to make a judgment include:

[0030] According to the load equipment classification control strategy L n The target parameter threshold range V n Make a judgment; if the load equipment parameter Y is predicted n Exceeds the target parameter threshold range V n , then the load device abnormality judgment result R n is abnormal; otherwise, the load device abnormality judgment result R n Is non-anomalous.

[0031] As a preferred technical solution of the present invention, the specific steps of step S3 include:

[0032] S31. Compare the total number of abnormal load devices with the target power load abnormal threshold. If the total number of abnormal load devices exceeds the target power load abnormal threshold, start the target power load special control strategy and release all abnormal parameter control strategies in the current target power load; otherwise, continue to control the power load according to the abnormal parameter control strategy;

[0033] S32. Use the target power load identification function to calculate and predict the probability of occurrence of special abnormalities. If the predicted probability of occurrence of special abnormalities exceeds the preset abnormality probability threshold, the target power load abnormality threshold is updated using the optimized genetic algorithm; otherwise, the power load control continues according to the target power load abnormality threshold.

[0034] As a preferred technical solution of the present invention, the specific steps of setting the target power supply load abnormality threshold by the optimized genetic algorithm include:

[0035] Construct K load anomaly threshold individuals G k , k=1, 2, ..., K; each load abnormality threshold individual G k Randomly generate in the solution space; K load abnormality threshold individuals G k Combine to obtain the initial load anomaly threshold iteration population; set the maximum number of iterations H max , record the current number of iterations as h;

[0036] According to the target power load, a digital twin model is constructed to obtain the simulated target power load; based on the initial load abnormality threshold, the load abnormality threshold individual G in the iterative population is k In the simulation target power load, several random faults are generated to obtain the switching abnormal state time S k ; Will switch to abnormal state time S k As the load abnormality threshold individual G k The fitness P k ;

[0037] According to the fitness P k Set the optimal individual Z1 and the suboptimal individual Z2; use the formula The population search factor J is calculated, and W1 and W2 are adjustment coefficients; when the population search factor J is greater than the search threshold, the initial load anomaly threshold iteration population is globally updated; otherwise, the initial load anomaly threshold iteration population is locally updated; the local individual update is based on the optimal individual Z1 and the suboptimal individual Z2.

[0038] When the maximum number of iterations is reached, the load abnormality threshold individual G corresponding to the maximum output fitness is k, which is the optimal load abnormality threshold individual; the load abnormality threshold in the optimal load abnormality threshold individual is used as the target power supply load abnormality threshold.

[0039] A power load control system based on artificial intelligence, comprising:

[0040] The power load data acquisition module is used to deploy the power load hierarchical control model and obtain the real-time load information of the target power load. The real-time load information includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ;

[0041] The power load prediction control module is used to deploy the power load prediction model and n Perform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n ; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance;

[0042] The power load abnormality control module is used to record the abnormality judgment results of all current load devices R n The abnormality in the middle corresponds to the current load device X n, and the total number of abnormal load devices is obtained; special control is implemented on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function.

[0043] The present invention has the following advantages:

[0044] By real-time monitoring of the status and parameters of load devices, the present invention enables the system to quickly respond to load changes, ensure that the power load is controlled in the optimal state, and improve the response speed and control accuracy of the system; the load device hierarchical control strategy library established by the optimized clustering algorithm can implement differentiated control strategies for different types of load devices, thereby improving the flexibility and effectiveness of power load management; the use of the power load prediction model can predict potential abnormal situations and deploy abnormal parameter control strategies in advance, thereby achieving prevention and rapid processing of abnormal situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the structure of an artificial intelligence-based power load control system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0047] Embodiment 1, a power load control method based on artificial intelligence, comprising:

[0048] S1. Obtain the real-time load information of the target power load, which includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ;

[0049] The target power load is a power system that needs to be automatically controlled. The target power load may include multiple load devices, and the parameter value of each load device is different;

[0050] The power load classification control model in step S1 includes a feature extraction layer, a feature matching layer and a result output layer;

[0051] The feature extraction layer is used to analyze the basic parameters C of the current load device. n Perform feature extraction to obtain the basic parameter feature C of the load device n ';

[0052] The feature matching layer is used to match the basic parameter characteristics of the load device C n 'Perform strategy matching to obtain load device hierarchical control strategy label Q n ; According to the load equipment classification control strategy label Q n Matching is performed in the load device hierarchical control strategy library to obtain the load device hierarchical control strategy L n ;

[0053] The result output layer is used to output the load device hierarchical control strategy L n ;

[0054] The specific steps for training the feature matching layer include:

[0055] Collecting several groups of load equipment parameter feature training samples, each group of load equipment parameter feature training samples includes a group of training load equipment basic parameter features with target values, and the target values ​​are hierarchical control strategy labels; combining several groups of load equipment parameter feature training samples to obtain a load equipment parameter feature training set;

[0056] Input the load device parameter feature training set into the feature matching layer and train it with the target value as the target to obtain the initial feature matching layer; perform model evaluation on the initial feature matching layer to obtain the initial feature matching layer model evaluation result; if the initial feature matching layer model evaluation result is passed, the initial feature matching layer is used as the trained feature matching layer; otherwise, continue model training using the load device parameter feature training set;

[0057] The BP neural network model can be used as the basic model of the power load hierarchical control model. The feature extraction layer focuses on extracting key features from the basic parameters of the load device, which helps to more accurately understand and analyze the behavior and needs of the load device; the feature matching layer quickly locates the appropriate hierarchical control strategy by matching with the extracted features, thereby improving the efficiency of the control strategy selection; by collecting training samples and continuously iterating training, the feature matching layer can be continuously optimized to improve matching accuracy and response speed; through precise feature extraction and matching, more precise control strategies can be provided for different load devices;

[0058] The specific steps of establishing a load device hierarchical control strategy library based on the optimized clustering algorithm in step S1 include:

[0059] Collect M load equipment parameter characteristics T m , m=1, 2, ..., M; M load equipment parameter characteristics T m Abstracted as a sample point set, we get the load equipment parameter feature sample point set A={T1', T2', ..., T M '}; Establish a weighted undirected graph for the load device parameter feature sample point set to obtain a classification graph of the load device to be divided; establish a weighted matrix of the load device to be divided according to the classification graph of the load device to be divided;

[0060] Calculate the sample point T m The local density B m and distance O m ; Using formula E m =B m *O m Calculate the sample point weight value E m ;

[0061] Use the fast search algorithm to calculate the sample point weight value E m Sample points T whose values ​​are greater than the preset dominant threshold m ', and the number of clusters is obtained; the value of the number of clusters is D;

[0062] Perform data clustering based on the number of clusters and the weight matrix of the load equipment to be divided, and obtain D load equipment cluster centers;

[0063] The load equipment cluster center is used as the classification label benchmark value, and a classification control strategy is established based on the classification label benchmark value; the classification control strategy is set by professional technicians based on the actual classification label benchmark value;

[0064] Traverse all load equipment cluster centers to establish a load equipment hierarchical control strategy library;

[0065] Through detailed analysis and clustering of load device parameter characteristics, the characteristics of different load devices can be more accurately identified, thereby achieving more sophisticated hierarchical control; the established hierarchical control strategy library can quickly adjust the control strategy according to the load changes monitored in real time, improve adaptability and response speed; optimize the clustering algorithm to help identify load devices of different priorities, achieve priority protection for critical loads and reasonable restrictions on non-critical loads, and optimize overall resource allocation; through reasonable hierarchical control, reduce the fluctuation of power loads and improve the stability and reliability of target power loads;

[0066] S2. For the current load device X nPerform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n ; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance:

[0067] The power load prediction model in step 2 includes a parameter feature extraction layer, a prediction parameter analysis layer and a prediction result output layer;

[0068] The parameter feature extraction layer is used to extract the load device parameters F to be analyzed. n Perform feature extraction to obtain the load equipment parameter feature F to be analyzed n ';

[0069] The prediction parameter analysis layer is used to predict the load device parameter characteristics F n 'Perform forecast analysis and obtain the forecast load equipment parameter Y n ;

[0070] The prediction result output layer is used to predict the load equipment parameters Y n and load equipment hierarchical control strategy L n For the current load device X n Make a judgment and get the load equipment abnormality judgment result R n ;

[0071] According to the predicted load equipment parameter Y n and load equipment hierarchical control strategy L n For the current load device X n The specific steps to make a judgment include:

[0072] According to the load equipment classification control strategy L n The target parameter threshold range V n Make a judgment; if the load equipment parameter Y is predicted n Exceeds the target parameter threshold range V n , then the load device abnormality judgment result R n is abnormal; otherwise, the load device abnormality judgment result R n is non-anomalous;

[0073] The specific steps for training the prediction parameter analysis layer include:

[0074] Collecting a plurality of sets of prediction parameter training sample pairs; each set of prediction parameter training sample pairs includes an input value and an output value; combining the plurality of sets of prediction parameter training sample pairs to obtain a prediction parameter training set;

[0075] Input the prediction parameter training set into the prediction parameter analysis layer for training with the output value as the target, and obtain the initial prediction parameter analysis layer; perform model evaluation on the initial prediction parameter analysis layer, and obtain the model evaluation result of the initial prediction parameter analysis layer; if the model evaluation result of the initial prediction parameter analysis layer is passed, the initial prediction parameter analysis layer is used as the trained prediction parameter analysis layer; otherwise, continue model training using the prediction parameter training set;

[0076] The prediction model based on artificial intelligence machine learning is used as the basic model of the power load prediction model. Through the specially designed parameter feature extraction layer, key features can be accurately extracted from the load device parameters to be analyzed, providing a solid foundation for prediction analysis; the prediction parameter analysis layer can conduct in-depth analysis based on the extracted features, predict potential load changes, and effectively identify abnormal situations in combination with hierarchical control strategies; the prediction result output layer automatically determines the load device status based on the prediction parameters and control strategies, reduces human intervention, and improves response speed and operation consistency; through accurate prediction and timely response, power resources can be allocated more reasonably, giving priority to critical loads, while making appropriate adjustments to non-critical loads; protection can be done in advance before abnormal situations occur to prevent unexpected situations from happening to the target power load;

[0077] S3. Record the abnormal judgment results R of all current load devices n The abnormality in the middle corresponds to the current load device X n , and obtain the total number of abnormal load devices; implement special control on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function;

[0078] The specific steps of step S3 include:

[0079] S31. Compare the total number of abnormal load devices with the target power load abnormal threshold. If the total number of abnormal load devices exceeds the target power load abnormal threshold, start the target power load special control strategy and release all abnormal parameter control strategies in the current target power load; otherwise, continue to control the power load according to the abnormal parameter control strategy;

[0080] S32. The target power load identification function is used to calculate the predicted probability of occurrence of a special abnormality. If the predicted probability of occurrence of a special abnormality exceeds the preset abnormal probability threshold, the target power load abnormality threshold is updated using the optimized genetic algorithm; otherwise, the power load control is continued according to the target power load abnormality threshold;

[0081] By recording and comparing the number of abnormal load devices with the preset abnormal threshold, the system can monitor the power load status in real time and respond quickly when the threshold is exceeded; once it is detected that the total number of abnormal load devices exceeds the threshold, a special control strategy will be activated to release all current abnormal parameter control strategies to protect the power system from damage; using the target power load identification function, the system can predict the probability of occurrence of special abnormalities, thereby assessing potential risks and taking preventive measures; dynamically updating the target power load abnormal threshold through the optimized genetic algorithm can adapt to the changing power load conditions and improve the effectiveness of the control strategy; through special control strategies and dynamically updated abnormal thresholds, the stability of the power load can be better maintained and the occurrence of abnormal situations can be reduced; the target power load special control strategy is set by professional and technical personnel in a targeted manner according to actual conditions;

[0082] The specific steps of setting the target power supply load abnormality threshold by the optimized genetic algorithm include:

[0083] Construct K load anomaly threshold individuals G k , k=1, 2, ..., K; each load abnormality threshold individual G k Randomly generate in the solution space; K load abnormality threshold individuals G k Combine to obtain the initial load anomaly threshold iteration population; set the maximum number of iterations H max , record the current iteration number as h; the solution space is set by professional technicians according to the actual situation to ensure the normal working requirements of the target power load;

[0084] According to the target power load, a digital twin model is constructed to obtain the simulated target power load; based on the initial load abnormality threshold, the load abnormality threshold individual G in the iterative population is k In the simulation target power load, several random faults are generated to obtain the switching abnormal state time S k ; Will switch to abnormal state time S k As the load abnormality threshold individual G k The fitness P k; The simulated target power load defaults to full load. When several random faults are generated, the number of faulty devices generated each time is the number corresponding to the load abnormality threshold. In the process of multiple fault generation, the simulated target power load may fail to control the strategy. The maximum number of faulty devices that can be tolerated is the load abnormality threshold. The switching abnormal state time indicates the maximum safe control time that the simulated target power load can tolerate.

[0085] According to the fitness P k Set the optimal individual Z1 and the suboptimal individual Z2; use the formula The population search factor J is calculated, W1 and W2 are adjustment coefficients, and W1 and W2 are set by professional technicians according to actual conditions; when the population search factor J is greater than the search threshold, the initial load anomaly threshold iterative population is globally updated; otherwise, the initial load anomaly threshold iterative population is locally updated; the local individual update is based on the optimal individual Z1 and the suboptimal individual Z2; the population search factor decreases slowly in the early stage of the iteration to obtain a larger exploration step, which helps the population to search extensively in the solution space and expand the generation range of the optimal solution; in the later stage of the iteration, the population search factor decays rapidly to a smaller value, so that the population performs a fine local search, which is conducive to accelerating the population convergence;

[0086] When the maximum number of iterations is reached, the load abnormality threshold individual G corresponding to the maximum output fitness is k , which is the optimal load abnormality threshold individual; the load abnormality threshold in the optimal load abnormality threshold individual is used as the target power supply load abnormality threshold;

[0087] By constructing multiple load anomaly threshold individuals and randomly generating them in the solution space, the diversity and comprehensiveness of the thresholds can be ensured and the accuracy of threshold setting can be improved; by combining multiple load anomaly threshold individuals into an initial population, the diversity of the population is increased, which helps the extensiveness and effectiveness of the genetic algorithm in searching for the global optimal solution; by setting the maximum number of iterations and evaluating and updating the individuals according to the fitness in each iteration, the threshold can be gradually optimized to improve the convergence of the algorithm and the quality of the solution; by simulating the target power load with a digital twin model, the effectiveness of the threshold can be tested and verified without affecting the actual system, thereby enhancing the practicality and reliability of the model; by generating random faults in the simulated target power load and evaluating the switching abnormal state time in real time as the fitness, the threshold evaluation is made more in line with the actual operating conditions.

[0088] Example 2, a power load control system based on artificial intelligence, see Figure 1 As shown, including:

[0089] The power load data acquisition module is used to deploy the power load hierarchical control model and obtain the real-time load information of the target power load. The real-time load information includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ;

[0090] The power load prediction control module is used to deploy the power load prediction model and n Perform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n ; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance;

[0091] The power load abnormality control module is used to record the abnormality judgment results of all current load devices R n The abnormality in the middle corresponds to the current load device X n , and the total number of abnormal load devices is obtained; special control is implemented on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function.

[0092] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A power load control method based on artificial intelligence, characterized in that: include: S1. Obtain the real-time load information of the target power load, which includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ; S2. For the current load device X n Perform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n ; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance; S3. Record the abnormal judgment results R of all current load devices n The abnormality in the middle corresponds to the current load device X n , and obtain the total number of abnormal load devices; implement special control on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function; The power load classification control model in step S1 includes a feature extraction layer, a feature matching layer and a result output layer; The feature extraction layer is used to analyze the basic parameters C of the current load device. n Perform feature extraction to obtain the basic parameter feature C of the load device n '; The feature matching layer is used to match the basic parameter characteristics of the load device C n 'Perform strategy matching to obtain load device hierarchical control strategy label Q n ; According to the load equipment classification control strategy label Q n Matching is performed in the load device hierarchical control strategy library to obtain the load device hierarchical control strategy L n ; The result output layer is used to output the load device hierarchical control strategy L n ; The specific steps for training the feature matching layer include: Collecting several groups of load equipment parameter feature training samples, each group of load equipment parameter feature training samples includes a group of training load equipment basic parameter features with target values, and the target values ​​are hierarchical control strategy labels; combining several groups of load equipment parameter feature training samples to obtain a load equipment parameter feature training set; The load device parameter feature training set is input into the feature matching layer for training with the target value as the target to obtain the initial feature matching layer; the initial feature matching layer is model evaluated to obtain the initial feature matching layer model evaluation result; if the initial feature matching layer model evaluation result is passed, the initial feature matching layer is used as the trained feature matching layer; otherwise, the load device parameter feature training set is used to continue model training.

2. The power load control method based on artificial intelligence according to claim 1 is characterized in that: The specific steps of establishing a load device hierarchical control strategy library based on the optimized clustering algorithm in step S1 include: Collect M load equipment parameter characteristics T m , m=1, 2, ..., M; M load equipment parameter characteristics T m Abstracted as a sample point set, we get the load equipment parameter feature sample point set A={T1', T2', ..., T M '}; Establish a weighted undirected graph for the load device parameter feature sample point set to obtain a classification graph of the load device to be divided; establish a weighted matrix of the load device to be divided according to the classification graph of the load device to be divided; Calculate the sample point T m The local density B m and distance O m ; Using formula E m =B m *O m Calculate the sample point weight value E m ; Use the fast search algorithm to calculate the sample point weight value E m Sample points T whose values ​​are greater than the preset dominant threshold m ', and the number of clusters is obtained; the value of the number of clusters is D; Perform data clustering based on the number of clusters and the weight matrix of the load equipment to be divided, and obtain D load equipment cluster centers; The load equipment cluster center is used as the classification label reference value, and a classification control strategy is established according to the classification label reference value; Traverse all load equipment cluster centers to establish a load equipment hierarchical control strategy library.

3. The power load control method based on artificial intelligence according to claim 2 is characterized in that: The power load prediction model in step 2 includes a parameter feature extraction layer, a prediction parameter analysis layer and a prediction result output layer; The parameter feature extraction layer is used to extract the load device parameters F to be analyzed. n Perform feature extraction to obtain the load equipment parameter feature F to be analyzed n '; The prediction parameter analysis layer is used to predict the load device parameter characteristics F n 'Perform forecast analysis and obtain the forecast load equipment parameter Y n ; The prediction result output layer is used to predict the load equipment parameters Y n and load equipment hierarchical control strategy L n For the current load device X n Make a judgment and get the load equipment abnormality judgment result R n ; The specific steps for training the prediction parameter analysis layer include: Collecting a number of sets of prediction parameter training sample pairs; each set of prediction parameter training sample pairs includes an input value and an output value; Combining several groups of prediction parameter training sample pairs to obtain a prediction parameter training set; The prediction parameter training set is input into the prediction parameter analysis layer for training with the output value as the target to obtain the initial prediction parameter analysis layer; the model of the initial prediction parameter analysis layer is evaluated to obtain the model evaluation result of the initial prediction parameter analysis layer; if the model evaluation result of the initial prediction parameter analysis layer is passed, the initial prediction parameter analysis layer is used as the trained prediction parameter analysis layer; otherwise, the model training is continued using the prediction parameter training set.

4. The power load control method based on artificial intelligence according to claim 3 is characterized in that: According to the predicted load equipment parameter Y n and load equipment hierarchical control strategy L n For the current load device X n The specific steps to make a judgment include: According to the load equipment classification control strategy L n The target parameter threshold range V n Make a judgment; if the load equipment parameter Y is predicted n Exceeds the target parameter threshold range V n , then the load device abnormality judgment result R n is abnormal; otherwise, the load device abnormality judgment result R n Is non-abnormal.

5. The power load control method based on artificial intelligence according to claim 4 is characterized in that: The specific steps of step S3 include: S31. Compare the total number of abnormal load devices with the target power load abnormal threshold. If the total number of abnormal load devices exceeds the target power load abnormal threshold, start the target power load special control strategy and release all abnormal parameter control strategies in the current target power load; otherwise, continue to control the power load according to the abnormal parameter control strategy; S32. Use the target power load identification function to calculate and predict the probability of occurrence of special abnormalities. If the predicted probability of occurrence of special abnormalities exceeds the preset abnormality probability threshold, the target power load abnormality threshold is updated using the optimized genetic algorithm; otherwise, the power load control continues according to the target power load abnormality threshold.

6. The power load control method based on artificial intelligence according to claim 5, characterized in that: The specific steps of setting the target power supply load abnormality threshold by the optimized genetic algorithm include: Construct K load anomaly threshold individuals G k , k=1, 2, ..., K; each load abnormality threshold individual G k Randomly generate in the solution space; K load abnormality threshold individuals G k Combine to obtain the initial load anomaly threshold iteration population; set the maximum number of iterations H max , record the current number of iterations as h; According to the target power load, a digital twin model is constructed to obtain the simulated target power load; based on the initial load abnormality threshold, the load abnormality threshold individual G in the iterative population is k In the simulation target power load, several random faults are generated to obtain the switching abnormal state time S k ; Will switch to abnormal state time S k As the load abnormality threshold individual G k The fitness P k ; According to the fitness P k Set the optimal individual Z1 and the suboptimal individual Z2; use the formula The population search factor J is calculated, and W1 and W2 are adjustment coefficients; when the population search factor J is greater than the search threshold, the initial load anomaly threshold iteration population is globally updated; otherwise, the initial load anomaly threshold iteration population is locally updated; the local individual update is based on the optimal individual Z1 and the suboptimal individual Z2. When the maximum number of iterations is reached, the load abnormality threshold individual G corresponding to the maximum output fitness is k , which is the optimal load abnormality threshold individual; the load abnormality threshold in the optimal load abnormality threshold individual is used as the target power supply load abnormality threshold.

7. A power load control system based on artificial intelligence, characterized in that: The system applies an artificial intelligence-based power load control method as described in any one of claims 1 to 6, including: The power load data acquisition module is used to deploy the power load hierarchical control model and obtain the real-time load information of the target power load. The real-time load information includes the current load device X n And the corresponding current load equipment basic parameters C n , n=1, 2, ..., N, N is the number of current load devices in the target power load; based on the real-time load information and the power load hierarchical control model, the current load device X is obtained. n Corresponding load equipment hierarchical control strategy L n ; The power load hierarchical control model includes a load device hierarchical control strategy library, which is established based on the optimized clustering algorithm and is used to perform hierarchical power load control on different load devices; Load device hierarchical control strategy L n The target parameter threshold range V is included n And the corresponding abnormal parameter control strategy U n ; The power load prediction control module is used to deploy the power load prediction model and n Perform continuous load equipment parameter monitoring to obtain the load equipment parameter F to be analyzed n ; Set the load device parameter F to be analyzed n and load equipment hierarchical control strategy L n Input into the power load prediction model for analysis to obtain the load equipment abnormality judgment result R n ; Load equipment abnormality judgment result R n Is abnormal or not abnormal; according to the abnormal judgment result of the load device R n For the current load device X n Automatically control the power load; the power load prediction model is used to predict the load device parameters of the current load device and deploy abnormal parameter control strategies in advance; The power load abnormality control module is used to record the abnormality judgment results of all current load devices R n The abnormality in the middle corresponds to the current load device X n , and the total number of abnormal load devices is obtained; special control is implemented on the power load based on the total number of abnormal load devices and the target power load abnormality threshold, the target power load abnormality threshold is set by the optimized genetic algorithm, and the target power load abnormality threshold is updated based on the target power load identification function.

Images