A fusion computing method for host security quantification

Abstract

The present invention discloses a fusion calculation method and system for host security quantification. The method includes obtaining different power system host data for fusion, using a least squares support vector machine to automatically classify the algorithm coefficients and protection factors of all host security quantification models in the different power system host data to obtain a selected library of power system host data for fusion; sending the different levels of power system host standard compliance operations obtained from the power system host standard compliance operation to the algorithm coefficient and protection factor fusion model of the host security quantification model to obtain the solution of the fusion equation of the algorithm coefficient and protection factor of the host security quantification model; sending the solution of the fusion equation of the algorithm coefficient and protection factor of the host security quantification model to a deep belief network model to perform data training on the fused algorithm coefficients and protection factors of the host security quantification model. The present invention is used to accurately and efficiently fuse the algorithm coefficients and protection factors of the host security quantification model.

Description

Technical Field

[0001] The present invention relates to the field of power system host security supervision, and particularly to a fusion calculation method and system for host security quantification. Background Art

[0002] The power host system generally includes a power system host acquisition system, a power system host analysis system, and a power system host display system. The power system host acquisition system will import the data obtained from certain portal websites or certain types of topics that users are concerned about into the power system host analysis system through a data synchronization module. After being processed by the analysis system, the obtained data is classified and clustered to form topics and focus points, and is intuitively presented to users through the power system host display system. Therefore, if all the unfiltered acquired data is imported into the analysis system during data synchronization, it will lead to the import of a lot of data that is not related to the power system host. This part of the data not only increases the analysis data volume pressure of the power system host analysis system, but also reduces the accuracy of the solutions of the analysis equations in aspects such as topics and focus points due to the chaotic acquired data; at the same time, the power system host display system will display a lot of content that users are not concerned about, affecting the user experience. Summary of the Invention

[0003] In order to overcome the shortcomings and deficiencies of the prior art, the present invention provides a fusion calculation method and system for host security quantification.

[0004] The technical solution adopted by the present invention is that a fusion calculation method and system for host security quantification provided by the present invention includes the following steps:

[0005] Step L1, obtain different power system host data for fusion, and use the least squares support vector machine to automatically classify the algorithm coefficients and protection factors of all host security quantification models in the different power system host data to obtain a power system host data selection library for fusion;

[0006] Step L2, use the power system host data selection library to perform data linear regression on the algorithm coefficients and protection factors based on the host security quantification model by intelligent optimization using the least squares support vector machine;

[0007] Step L3, calculate the fusion algorithm parameters and the algorithm coefficients and protection factor calculation parameters of the host security quantification model by using the PID controller of human-computer interaction.

[0008] Step L4: Obtain the data to be managed for the standard-compliant operation of the power system host, and send the data to be managed for the standard-compliant operation of power system hosts at different levels obtained in the standard-compliant operation of the power system host into the algorithm coefficient and protection factor fusion model of the host security quantization model to obtain the solution of the algorithm coefficient and protection factor fusion equation of the host security quantization model;

[0009] Step L5: Send the solution of the algorithm coefficient and protection factor fusion equation of the host security quantization model into the deep belief network model to train the data of the algorithm coefficient and protection factor of the fused host security quantization model;

[0010] Step L6: Check whether the algorithm coefficients and protection factors of the host security quantization model after data training and fusion meet the set specifications, and determine whether there are any abnormal situations with the algorithm coefficients and protection factors of the host security quantization model.

[0011] The deep belief network model has the following expression:

[0012]

[0013] Where, represents the algorithm coefficient and protection factor matrix of the host security quantization model, σ represents the constant coefficient of the true value matrix, represents the true value matrix of the algorithm coefficient and protection factor of the host security quantization model, ξ represents the weight of the algorithm coefficient and protection factor of the host security quantization model, and B zs represents the gain matrix of the algorithm coefficient and protection factor of the host security quantization model;

[0014] The algorithm coefficient and protection factor of the host security quantization model have the following expression:

[0015] D x = L X - Y(E x W x - D x-1 )

[0016] Where, D x represents the set of algorithm coefficients and protection factors of the host security quantization model at the current moment, L x represents the set of algorithm coefficients and protection factors of the host security quantization model at the previous moment, W x represents the influence range of different levels of optimization, Y represents the type matrix of the algorithm coefficients and protection factors of the host security quantization model, and E x represents the weight of the influence of the algorithm coefficients and protection factors of different host security quantization models on the optimization level, and D x-1 represents the predicted value of the influence of the optimization level.

[0017] Preferably, after the algorithm coefficients and protection factors of the host security quantization model are intelligently optimized using least squares support vector machine for data linear regression, the parameters of the least squares support vector machine algorithm are iteratively calculated.

[0018] Preferably, optimizing whether the algorithm coefficients and protection factors of the host security quantization model meet the set specifications includes the following steps:

[0019] Step U1, obtain the scale, retrieval speed, and initial values of the algorithm coefficients and protection factor calculation parameters of a certain level of host security quantization model in the current power system host standard compliance operation to be managed.

[0020] Step U2, when both the scale and retrieval speed of the algorithm coefficients and protection factor calculation parameters of the host security quantization model are within the algorithm coefficients and protection factor calculation parameters of the host security quantization model, there is an unmerged situation in optimizing the algorithm coefficients and protection factors of the host security quantization model.

[0021] Step U3, if only some of the algorithm coefficients and protection factors of the algorithm coefficients and protection factor calculation parameters of the host security quantization model are within the algorithm coefficients and protection factor calculation parameters of the host security quantization model, then go to Step U4.

[0022] Step U4, optimize whether the algorithm coefficients and protection factor setting specifications of the host security quantization model are within the algorithm coefficients and protection factor calculation parameters of the host security quantization model. If the algorithm coefficients and protection factor setting specifications of the host security quantization model are within the algorithm coefficients and protection factor calculation parameters of the host security quantization model, then there is an unmerged situation in optimizing the algorithm coefficients and protection factors of the host security quantization model; otherwise, proceed to the next step.

[0023] Step U5, continue to obtain the algorithm coefficients and protection factors of other integrated host security quantization models in the current power system host standard compliance operation to be managed and perform the operation of optimizing whether there is an unoptimized situation for the algorithm coefficients and protection factors.

[0024] This application also includes a method for optimizing whether to discard the algorithm coefficients and protection factors when the algorithm coefficients and protection factors of the host security quantization model are within the algorithm coefficients and protection factor calculation parameters of the host security quantization model, including the following steps:

[0025] Step R1, calculate the algorithm coefficients and protection factor setting specifications of the host security quantification model for the previous power system host standard-compliant operation to be managed, and the optimization range of the algorithm coefficients and protection factor setting specifications of the host security quantification model for the current power system host standard-compliant operation to be managed. If the optimization range exceeds the preset interval, it is determined that the calculation of the algorithm coefficients and protection factors of the host security quantification model is abnormal;

[0026] Step R2, if the algorithm coefficients and protection factors for optimizing the host security quantification model are in an abnormal calculation state, then clear the duration of the unmerged situation in the algorithm coefficients and protection factors of the host security quantification model and process the algorithm coefficients and protection factors of other host security quantification models for the power system host standard-compliant operation to be managed;

[0027] Step R3, if the algorithm coefficients and protection factors of the host security quantification model are not updated, then obtain the current time and calculate the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model. Compare the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model with the preset interval of the duration of the unmerged situation. If the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model exceeds the preset interval of the duration of the unmerged situation, then optimize the algorithm coefficients and protection factors of the host security quantification model to have an unmerged situation. If the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model does not exceed the duration of the unmerged situation, then the processing of the algorithm coefficients and protection factors of the current host security quantification model ends, and continue to process other power system host standard-compliant operations to be managed for fusion.

[0028] Preferably, if in the data training data, the algorithm coefficients and protection factors of a certain host security quantification model are trained in the previous iteration of the power system host standard-compliant operation data training, and the algorithm coefficients and protection factors of the host security quantification model are not trained in the current data, then the system sets a maximum standard threshold. Before the maximum standard threshold is reached, the algorithm coefficients and protection factors of the host security quantification model are not optimized incorrectly. Then, use the unscented deep belief network in the deep belief network model to predict the current algorithm coefficients and protection factor area of the host security quantification model based on the algorithm coefficients and protection factors area of the previous iteration. The solution of the prediction equation is used as the algorithm coefficients and protection factor area of the current host security quantification model;

[0029] If the characteristic power system host standard of the next host security quantification model conforms to the operation to be managed and is fused into the algorithm coefficients and protection factor area position of the host security quantification model, and matches the algorithm coefficients and protection factor area of the current host security quantification model, then it is optimized that the algorithm coefficients and protection factors of the host security quantification model caused by the fusion algorithm error disappear;

[0030] If the maximum standard threshold is reached, it is directly regarded that the algorithm coefficients and protection factors of the host security quantification model disappear, and the system deletes the algorithm coefficients and protection factor data training data of this host security quantification model;

[0031] If the algorithm coefficients and protection factors of the host security quantification model reappear during the period when the maximum standard threshold is not reached, then it is optimized that the algorithm coefficients and protection factors of the host security quantification model are not fused for a short time.

[0032] Preferably, the time for obtaining the operation to be managed that conforms to the power system host standard is set according to the operation demand that conforms to the standard, and different numbers of operations to be managed that conform to the power system host standard are obtained per second.

[0033] The system of the present application includes an algorithm coefficient and protection factor fusion component of the host security quantification model, a power system host monitoring component, and a feature non-fusion optimization component of the host security quantification model, where:

[0034] The algorithm coefficient and protection factor fusion component of the host security quantification model is used to fuse the algorithm coefficients and protection factors of the host security quantification model for the operation to be managed data of the power system host obtained by the power system host monitoring component, and obtain the calculation parameters and calculation parameter data of the algorithm coefficients and protection factors of the host security quantification model;

[0035] The power system host monitoring component is used to obtain the operation to be managed data of the power system host that conforms to the standard, set the supervision area position, and after matching the specific fusion code for the algorithm coefficient and protection factor calculation parameters and calculation parameter data of the host security quantification model obtained by the algorithm coefficient and protection factor fusion component of the host security quantification model, transmit them to the feature non-fusion optimization component of the host security quantification model;

[0036] After receiving the algorithm coefficients, protection factor calculation parameters, and calculation parameter data of the host security quantification model with the specific fusion code matched, the feature un-fusion optimization component of the host security quantification model outputs data trainer data, searches for the data trainer data at different levels, optimizes whether the algorithm coefficients and protection factors of this host security quantification model are within the supervision area according to the supervision area position, and updates the data trainer data again. Then, it optimizes whether the algorithm coefficients and protection factors of the host security quantification model should issue an alarm according to the updated data trainer data.

[0037] Preferably, the data trainer data includes: the algorithm coefficient and protection factor fusion code of the host security quantification model, the algorithm coefficient and protection factor calculation parameters of the host security quantification model, whether the algorithm coefficient and protection factor of the host security quantification model enter the algorithm coefficient and protection factor calculation parameters of the host security quantification model, the time when the algorithm coefficient and protection factor of the host security quantification model enter the algorithm coefficient and protection factor calculation parameters of the host security quantification model, and whether the algorithm coefficient and protection factor of the host security quantification model have issued an alarm.

[0038] Preferably, the power system host data selection library uses a human-computer interaction algorithm to perform timed supervision on the data to be optimized of the power system host.

[0039] The present invention applies the algorithm coefficient and protection factor fusion technology of the least squares support vector machine for the host security quantification model to the fusion of the algorithm coefficients and protection factors of the host security quantification model. This method can accurately fuse the algorithm coefficients and protection factors of the host security quantification model from the operation and management of the power system host's standard compliance, perform data training on the algorithm coefficients and protection factors of the host security quantification model, and then through a series of logical optimizations for the situation where the algorithm coefficients and protection factors of the host security quantification model are not fused, accurately and efficiently fuse the algorithm coefficients and protection factors of the host security quantification model and generate an alarm, thereby realizing the intelligent supervision of the situation where the algorithm coefficients and protection factors of the host security quantification model are not fused. This greatly improves the efficiency of the staff and also saves a large amount of manpower and material resources.

[0040] Algorithm coefficient and protection factor fusion component of the host security quantification model: The present invention uses the power system host data selection library to perform data linear regression on the intelligent optimization of the algorithm coefficients and protection factors based on the host security quantification model using the least squares support vector machine to ensure the fusion accuracy of the algorithm coefficients and protection factors of the host security quantification model. The present invention also performs iterative calculations on the parameters of the least squares support vector machine algorithm to ensure the performance optimization under a limited budget.

[0041] Power system host monitoring component: The present invention conducts data training on the algorithm coefficients and protection factors of the host security quantification model integrated therein based on a data training algorithm, and maintains its individual structure for optimizing the states of the algorithm coefficients and protection factors of the host security quantification model, and sets a preset interval for data training failure to avoid false alarms caused by non-integration.

[0042] Feature non-integration optimization component of the host security quantification model: The present invention first optimizes whether the algorithm coefficients and protection factors of the host security quantification model obtained through data training are within the supervision area, and proposes various optimization methods for discarding the algorithm coefficients and protection factors of the algorithm coefficients and protection factors of the host security quantification model. By discarding the algorithm coefficients and protection factors for optimizing the algorithm coefficients and protection factors of the host security quantification model, the problem of false alarms of the algorithm coefficients and protection factors of the host security quantification model in the traditional method is avoided.

[0043] The present invention can conveniently set the fusion algorithm for non-integration situations from the cloud service page, and the alarm messages can also be displayed in real time on the cloud service page for staff to view, and together with the core fusion algorithm, it constitutes an intelligent fusion system for the algorithm coefficients and protection factors of the host security quantification model with non-integration situations. Brief Description of the Drawings

[0044] Figure 1 It is the first flowchart of the method of the present invention;

[0045] Figure 2 It is the second flowchart of the method of the present invention;

[0046] Figure 3 It is the third flowchart of the method of the present invention. Detailed Description of the Embodiment

[0047] It should be noted that, without conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The following further describes the present application in detail with reference to the drawings and specific embodiments.

[0048] As Figure 1 shown, a fusion calculation method for host security quantification includes the following steps:

[0049] Step L1, obtain different power system host data for fusion, and automatically classify all the algorithm coefficients and protection factors of the host security quantification model in the different power system host data using the least squares support vector machine to obtain a selected library of power system host data for fusion;

[0050] Step L2, using the host data selection library of the power system, perform data linear regression on the algorithm coefficients and protection factors based on the host security quantification model by using the least squares support vector machine for intelligent optimization;

[0051] Step L3, set the fusion algorithm parameters, the algorithm coefficients of the host security quantification model, and the calculation parameters of the protection factors by using the PID controller of the human-computer interaction;

[0052] Step L4, obtain the power system host standard compliance operation data to be managed, and send the different levels of power system host standard compliance operation data obtained from the power system host standard compliance operation into the algorithm coefficient and protection factor fusion model of the host security quantification model to obtain the solution of the algorithm coefficient and protection factor fusion equation of the host security quantification model;

[0053] Step L5, send the solution of the algorithm coefficient and protection factor fusion equation of the host security quantification model into the deep belief network model to perform data training on the algorithm coefficients and protection factors fused into the host security quantification model;

[0054] The deep belief network model, in English is Deep Belief Network, abbreviated as DBN. It is a hybrid generative model composed of a restricted Boltzmann machine (RBM) and a sigmoid belief network (SBN). Each node of the model follows a Bernoulli distribution, which is consistent with the assumptions of RBM and SBN. The bottom layer is the observed variable layer, and then the first layer, the second layer... are arranged in sequence upwards. Each layer is represented by weights, including the bias term. In the deep belief network model of the present invention, the expression is:

[0055]

[0056] Among them, represents the algorithm coefficient and protection factor matrix of the host security quantification model, σ represents the constant coefficient of the true value matrix, represents the true value matrix of the algorithm coefficient and protection factor of the host security quantification model, ξ represents the weight of the algorithm coefficient and protection factor of the host security quantification model, B zs represents the gain matrix of the algorithm coefficient and protection factor of the host security quantification model;

[0057] Step L6, fuse and optimize whether the algorithm coefficients and protection factors of the data-trained host security quantification model meet the set specifications, and judge whether there are any abnormal situations for the algorithm coefficients and protection factors of the host security quantification model.

[0058] After performing data linear regression on the algorithm coefficients and protection factors of the host security quantification model using the least squares support vector machine for intelligent optimization, the parameters of the least squares support vector machine algorithm are iteratively calculated.

[0059] The algorithm coefficients and protection factors of the host security quantification model are expressed as:

[0060] D x = L x - Y(E x W x - D x-1 )

[0061] Where D x represents the set of algorithm coefficients and protection factors of the host security quantification model at the current moment, L x represents the set of algorithm coefficients and protection factors of the host security quantification model at the previous moment, W x represents the influence range of different levels of optimization, Y represents the type matrix of the algorithm coefficients and protection factors of the host security quantification model, and E x represents the weight of the algorithm coefficients and protection factors of different host security quantification models on the influence of the optimization level, and D x-1 represents the predicted value of the influence of the optimization level.

[0062] As Figure 2 shown, whether the algorithm coefficients and protection factors of the optimized host security quantification model meet the set specifications includes the following steps:

[0063] Step U1, obtain the scale and retrieval speed of the algorithm coefficients and protection factor calculation parameters of a certain level of the host security quantification model in the current power system host standard compliance operation to be managed, and the initial values of the algorithm coefficients and protection factor calculation parameters of the host security quantification model;

[0064] Step U2, both the scale and retrieval speed of the algorithm coefficients and protection factor calculation parameters are within the algorithm coefficients and protection factor calculation parameters of the host security quantification model, and there is an unmerged situation in the optimized algorithm coefficients and protection factors of the host security quantification model;

[0065] Step U3, only some of the algorithm coefficients and protection factors of the algorithm coefficients and protection factor calculation parameters of the host security quantification model are within the algorithm coefficients and protection factor calculation parameters of the host security quantification model, then go to Step U4;

[0066] Step U4, optimize the algorithm coefficients of the host security quantification model and the setting specifications of the protection factors. Check whether the algorithm coefficients and protection factors are within the calculation parameters of the algorithm coefficients and protection factors of the host security quantification model. If the algorithm coefficients and protection factors of the host security quantification model are within the calculation parameters of the algorithm coefficients and protection factors, then check if there is an unmerged situation for the optimized algorithm coefficients and protection factors of the host security quantification model. Otherwise, proceed to the next step;

[0067] Step U5, continue to obtain other algorithm coefficients and protection factors of the host security quantification model that are being managed for compliance operation of the current power system host standard, and perform an optimization operation on whether there is an unoptimized situation for the algorithm coefficients and protection factors of the host security quantification model.

[0068] As Figure 3 shown, this application also includes a method for determining whether to discard the algorithm coefficients and protection factors when the algorithm coefficients and protection factors of the host security quantification model are within the calculation parameters of the algorithm coefficients and protection factors of the host security quantification model, including the following steps:

[0069] Step R1, calculate the optimization range between the setting specifications of the algorithm coefficients and protection factors of the host security quantification model in the previous power system host standard compliance operation to be managed and the setting specifications of the algorithm coefficients and protection factors of the host security quantification model in the current power system host standard compliance operation to be managed. If the optimization range exceeds the preset interval, it is determined that the calculation of the algorithm coefficients and protection factors of the host security quantification model is abnormal;

[0070] Step R2, if the optimization of the algorithm coefficients and protection factors of the host security quantification model is in an abnormal calculation state, then clear the duration of the unmerged situation in the algorithm coefficients and protection factors of the host security quantification model and process the algorithm coefficients and protection factors of other host security quantification models for the power system host standard compliance operation to be managed;

[0071] Step R3, if the algorithm coefficients and protection factors of the host security quantification model have not been updated, then obtain the current time and calculate the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model. Compare the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model with the preset interval of the duration of the unmerged situation. If the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model exceeds the preset interval of the duration of the unmerged situation, then it is optimized that there is an unmerged situation for the algorithm coefficients and protection factors of the host security quantification model. If the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model does not exceed the duration of the unmerged situation, then the processing of the algorithm coefficients and protection factors of the current host security quantification model ends, and continue to process other power system host standard compliance operations to be managed for integration.

[0072] If the algorithm coefficients and protection factors of a certain host security quantization model are trained from the power system host standard compliance operation to be managed data in the previous iteration of the data training data, and the algorithm coefficients and protection factors of the host security quantization model are not trained in the current data, then the system sets a maximum standard threshold. Before the maximum standard threshold is reached, the algorithm coefficients and protection factors of the host security quantization model are not optimized incorrectly. After that, the unscented deep belief network in the deep belief network model is used to predict the algorithm coefficients and protection factor area of the current host security quantization model based on the algorithm coefficients and protection factor area of the host security quantization model in the previous iteration. The solution of the predicted equation is used as the algorithm coefficients and protection factor area of the current host security quantization model;

[0073] If the characteristics of the next host security quantization model, the power system host standard compliance operation to be managed, are fused into the algorithm coefficient and protection factor area position of the host security quantization model and match the algorithm coefficient and protection factor area of the current host security quantization model, then the algorithm coefficients and protection factors of the host security quantization model caused by the fusion algorithm error are optimized to disappear;

[0074] If the maximum standard threshold is reached, it is directly regarded as the disappearance of the algorithm coefficients and protection factors of the host security quantization model, and the system deletes the algorithm coefficient and protection factor data training data of this host security quantization model;

[0075] If the algorithm coefficients and protection factors of the host security quantization model reappear during the period when the maximum standard threshold is not reached, then the algorithm coefficients and protection factors of the host security quantization model are optimized to be not fused for a short time.

[0076] The time to obtain the power system host standard compliance operation to be managed is set according to the standard compliance operation requirements for different lengths of time, and different numbers of power system host standard compliance operations to be managed are obtained per second.

[0077] A fusion calculation method for host security quantization in this application includes:

[0078] The algorithm coefficient and protection factor fusion component of the host security quantization model first obtains the data for fusion based on the actual scenario. After obtaining the data, all the algorithm coefficients and protection factors of the host security quantization model in different power system host data are automatically classified using the least squares support vector machine to obtain the selected library of power system host data for fusion.

[0079] To better ensure the fusion effect, the present invention can greatly increase the amount of fusion data by means of data augmentation such as translation, flipping, and scaling of the selected library of fused power system host data.

[0080] Then, through the fusion power system host data selection library, the algorithm coefficients and protection factors based on the host security quantization model are intelligently optimized, and the least squares support vector machine is used for data linear regression to ensure the accuracy of the model.

[0081] After obtaining the fused model, the traditional method generally directly deploys and utilizes the fused original model in combination with the business program. This method not only has a slow inference speed but also extremely occupies hardware resources.

[0082] To ensure the real-time performance of the fusion of the algorithm coefficients and protection factors of the host security quantization model in the business, in the present invention, the original model uses the particle swarm algorithm to perform operations such as speed update, position update, and weight quantization on the model to optimize the inference throughput of the model, and perform forward inference to accelerate the inference.

[0083] Power system host monitoring component. First, the intelligent fusion system of the present invention sets the fusion algorithm parameters by using the PID controller of human-computer interaction. The cloud service page displays the picture of the power system host standard compliance operation to be set in real time. The user can draw the algorithm coefficients and protection factor calculation parameters of the host security quantization model on the picture with the mouse. After the user finishes drawing, the setting of the algorithm will be sent to the algorithm server through the network.

[0084] After the algorithm receives the setting, it will pull the real-time data of the power system host standard compliance operation to be set according to the specified video stream address, and send the obtained power system host standard compliance operation at different levels to be managed into the fusion model of the algorithm coefficients and protection factors of the host security quantization model, and obtain the solution of the fusion equation of the algorithm coefficients and protection factors of the host security quantization model.

[0085] Then, the solution of the fusion equation of the algorithm coefficients and protection factors of the host security quantization model is sent into the deep belief network model to perform data training on the algorithm coefficients and protection factors of the fused host security quantization model.

[0086] The advantage of data training is to match the algorithm coefficients and protection factors of the same host security quantization model in the time series and assign the same fusion code, avoiding the problem of continuous alarms of the features continuously fused into the same host security quantization model.

[0087] The feature non-fusion optimization component of the host security quantization model traverses all the algorithm coefficients and protection factors of the host security quantization model trained by the data trainer data of the power system host monitoring component. If they are the algorithm coefficients and protection factors of the host security quantization model trained with new data, the data thereof will be initialized;

[0088] If the system has previously trained the algorithm coefficients and protection factors of this host security quantization model with data, and currently trains the algorithm coefficients and protection factors of this host security quantization model with data again, then first optimize whether it is within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, and use the scale and retrieval speed of the algorithm coefficient and protection factor calculation parameters of the host security quantization model and the initial values of the algorithm coefficient and protection factor calculation parameters of the host security quantization model to optimize whether the algorithm coefficients and protection factors of the host security quantization model meet the set specifications.

[0089] The specific method is: first obtain the scale and retrieval speed of the calculation parameters and the initial values of the algorithm coefficient and protection factor calculation parameters of the host security quantization model;

[0090] If both the scale and the retrieval speed are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, optimize whether there is an unmerged situation of the algorithm coefficients and protection factors of the host security quantization model;

[0091] If only some of the algorithm coefficients and protection factors are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, then optimize whether the algorithm coefficients and protection factors set specifications of the host security quantization model are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model. If so, optimize whether there is an unmerged situation of the algorithm coefficients and protection factors of the host security quantization model; the rest are all regarded as having no unmerged situation.

[0092] If the algorithm coefficients and protection factors of the host security quantization model are not within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, then the processing of the algorithm coefficients and protection factors of the current host security quantization model ends, and continue to process other fusion power system host standards to meet the operation to be managed. When all the algorithm coefficients and protection factors of the host security quantization model are traversed, then the current processing ends, and continue to obtain the characteristics of the next host security quantization model.

[0093] The algorithm coefficients and protection factors of the host security quantization model are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model. The present invention adopts a method to optimize whether to discard the algorithm coefficients and protection factors of the host security quantization model to avoid the problem that the algorithm coefficients and protection factors of the host security quantization model are misjudged as having an unmerged situation and alarming when they are always within the algorithm coefficient and protection factor calculation parameters of the host security quantization model during the slow driving process due to traffic jams or other reasons in a large area of the algorithm coefficient and protection factor calculation parameters of the host security quantization model.

[0094] The specific method is:

[0095] (1) Calculate the algorithm coefficients and the setting specifications of the protection factors of the host security quantification model for the previous power system host in the standard-compliant operation to be managed. The algorithm coefficients and protection factors are calculated and optimized within the range of the algorithm coefficients and protection factor setting specifications of the current host security quantification model. If this distance exceeds a certain preset interval, it is determined that the calculation of the algorithm coefficients and protection factors of the host security quantification model is abnormal;

[0096] (2) Since the distances of the algorithm coefficients and protection factors of the host security quantification model from the standard-compliant operation of the power system host are different, there will be significant differences in the sizes of the databases fused from the algorithm coefficients and protection factors of the host security quantification models with different distances from the standard-compliant operation of the power system host in the standard-compliant operation of the power system host to be managed.

[0097] Therefore, if the method in step (1) uses a single preset interval as the judgment condition, it will result in different optimization discard criteria for the algorithm coefficients and protection factors of the host security quantification models with different distances.

[0098] The present invention proposes to use form to optimize whether to discard the algorithm coefficients and protection factors of the host security quantification model;

[0099] In the formula, J is the scale of the calculation parameters, and Y represents the preset interval proportionality factor;

[0100] During the optimization process, Y represents a fixed value that can be adjusted by the algorithm personnel.

[0101] The effect of setting a dynamic preset interval is that when the calculation parameters are large, the corresponding distance preset interval is larger, and when the calculation parameters are small, the corresponding distance preset interval is smaller, so as to make the calculation and optimization of the algorithm coefficients and protection factors of the host security quantification models with different distances from the standard-compliant operation of the power system host more accurate.

[0102] If the calculation of the algorithm coefficients and protection factors of the host security quantification model is abnormal, the time length of the un-fused situation in the algorithm coefficients and protection factors of the host security quantification model is cleared, and other fusions of the standard-compliant operation of the power system host to be managed are continued;

[0103] If the algorithm coefficients and protection factors of the host security quantification model are not calculated abnormally, obtain the current time and calculate the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model. Compare the stop cumulative time of the algorithm coefficients and protection factors of the host security quantification model with the preset interval of the duration of the unmerged situation. If it exceeds the preset interval of the duration of the unmerged situation, optimize it to the situation where the algorithm coefficients and protection factors of the host security quantification model have unmerged situations, and send the algorithm coefficients and protection factors of the host security quantification model to the cloud server through the network. After receiving the alarm message, the cloud server will display the alarm data on the page; if it does not exceed the preset interval of the duration of the unmerged situation, the processing of the current algorithm coefficients and protection factors of the host security quantification model ends, and continue to process other fusions of the power system host standard compliance operation to be managed.

[0104] If the features of a certain host security quantification model were trained in the previous iteration of the data training data but not in the current data training, the following three situations will occur:

[0105] 1. The fusion algorithm error fails to detect the algorithm coefficients and protection factors of the host security quantification model;

[0106] 2. The algorithm coefficients and protection factors of the host security quantification model exceed the library range;

[0107] 3. There is unmerged, and other objects unmerge the algorithm coefficients and protection factors of the host security quantification model, resulting in the fusion algorithm being unable to fuse with the algorithm coefficients and protection factors of the host security quantification model.

[0108] For the above situations, the specific method for the present invention to handle is: combining the above three possible situations, set a maximum standard threshold. Before the maximum standard threshold is reached, do not optimize the error of the algorithm coefficients and protection factors of this host security quantification model. Use the unscented deep belief network in the deep belief network model to predict the current algorithm coefficients and protection factor area of the host security quantification model based on the position of the algorithm coefficients and protection factors of the host security quantification model in the previous iteration. The solution of the predicted equation is used as the algorithm coefficients and protection factor area of the current host security quantification model, but the data training state is still not trained to.

[0109] For the first situation, if the fusion algorithm error fails to detect the algorithm coefficients and protection factors of the host security quantification model, when the features of the next host security quantification model are fused into the algorithm coefficients and protection factor area of the host security quantification model, the algorithm coefficients and protection factor area obtained by the data training algorithm will match the current algorithm coefficients and protection factor area of the host security quantification model, avoiding the problem of repeated alarms caused by unstable fusion algorithms;

[0110] For the second case, when the maximum standard threshold is reached, the algorithm considers that the algorithm coefficients and protection factors of the current host security quantization model disappear, and deletes the algorithm coefficient and protection factor data training data of this host security quantization model;

[0111] For the third case, if there is no fusion for a short time, after the algorithm coefficients and protection factors of the host security quantization model reappear, the data training algorithm can still match the algorithm coefficients and protection factor areas of the previous host security quantization model with the algorithm coefficients and protection factor areas of the host security quantization model after non-fusion, and fuse them into the algorithm coefficients and protection factors of the same host security quantization model, avoiding the problem of repeated alarms.

[0112] The maximum standard threshold is recommended to be set according to the time required for the power system host standard to meet the operation to be managed, and the length of time is set according to the standard compliance operation requirements. Different numbers of power system host standards to meet the operation to be managed are obtained per second

[0113] Specifically, the algorithm coefficient and protection factor fusion component of the host security quantization model, the power system host monitoring component, and the feature non-fusion optimization component of the host security quantization model communicate through the following content:

[0114] The algorithm coefficient and protection factor fusion component of the host security quantization model generates a model file after fusion and acceleration;

[0115] The power system host monitoring component first obtains the cloud service settings through network communication; the cloud service settings specifically include: the power system host standard compliance operation flow address, the supervision area location;

[0116] After the settings are completed, the power system host monitoring component loads the model file of the algorithm coefficient and protection factor fusion component of the host security quantization model to fuse the algorithm coefficients and protection factors of the power system host standard compliance operation to be managed data and obtain the calculation parameters and calculation parameter data of the algorithm coefficients and protection factors of the host security quantization model;

[0117] Then, the calculation parameters are sent into the deep belief network model. After matching a specific fusion code for each calculation parameter, they are passed into the feature non-fusion optimization component of the host security quantization model to obtain the data trainer data. Among them, the data trainer data includes the algorithm coefficient and protection factor fusion code of the host security quantization model, the algorithm coefficient and protection factor calculation parameters of the host security quantization model, whether the algorithm coefficients and protection factors of the host security quantization model enter the algorithm coefficient and protection factor calculation parameters of the host security quantization model, the time when the algorithm coefficients and protection factors of the host security quantization model enter the algorithm coefficient and protection factor calculation parameters of the host security quantization model, and whether the algorithm coefficients and protection factors of the host security quantization model have issued an alarm;

[0118] The feature non-fusion optimization component of the host security quantization model traverses the data of trainers at different levels, optimizes whether the algorithm coefficients and protection factors of this host security quantization model are located within the supervision area according to the location of the supervision area, updates the data of the trainer, and optimizes whether the algorithm coefficients and protection factors of the host security quantization model should issue an alarm according to the data of the trainer.

[0119] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected", "connected", "fixed" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood through specific situations.

[0120] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various equivalent changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalent scope.

Claims

1. A fusion computing method for host security quantification, characterized in that: The method includes: Step L1: Obtain different power system host data for fusion, and use the least squares support vector machine to automatically classify the algorithm coefficients and protection factors of all host security quantization models in the different power system host data to obtain a power system host data selection library for fusion; Step L2: Use the power system host data selection library to perform data linear regression on the algorithm coefficients and protection factors of the host security quantization model by intelligent optimization using the least squares support vector machine; Step L3: Calculate the fusion algorithm parameters and the algorithm coefficients and protection factor calculation parameters of the host security quantization model by using a PID controller for human-computer interaction; Step L4: Obtain the power system host standard compliance operation data to be managed, and send the different-level power system host standard compliance operations obtained from the power system host standard compliance operation into the algorithm coefficient and protection factor fusion model of the host security quantization model to obtain the solution of the algorithm coefficient and protection factor fusion equation of the host security quantization model; Step L5: Send the solution of the algorithm coefficient and protection factor fusion equation of the host security quantization model into the deep belief network model to train the algorithm coefficients and protection factors of the fused host security quantization model; Step L6: Fusion and optimize whether the algorithm coefficients and protection factors of the data training meet the set specifications, and judge whether there are any unprocessed situations for the algorithm coefficients and protection factors of the host security quantization model; The deep belief network model, the expression is: Among them, represents the algorithm coefficients and protection factor matrix of the host security quantization model, σ represents the constant coefficient of the true value matrix, represents the true value matrix of the algorithm coefficients and protection factors of the host security quantization model, ξ represents the weights of the algorithm coefficients and protection factors of the host security quantization model, B zs represents the gain matrix of the algorithm coefficients and protection factors of the host security quantization model; The algorithm coefficients and protection factors of the host security quantization model, the expression is: D x = L x - Y(E x W x - D x-1 ) Among them, D x represents the set of algorithm coefficients and protection factors of the host security quantification model at the current moment, L x represents the set of algorithm coefficients and protection factors of the host security quantification model at the previous moment, W x represents the influence range of different levels of optimization, Y represents the type matrix of algorithm coefficients and protection factors of the host security quantification model, E x represents the weight of the influence of algorithm coefficients and protection factors of different host security quantification models on the optimization degree, D x-1 represents the predicted value of the influence of the optimization degree.

2. The fusion calculation method for host security quantification according to claim 1, characterized in that: After performing data linear regression on the algorithm coefficients and protection factors of the host security quantization model by intelligent optimization using the least squares support vector machine, the parameters of the least squares support vector machine algorithm are iteratively calculated at the same time.

3. A fusion calculation method for host security quantification according to claim 1, characterized in that: Optimizing whether the algorithm coefficients and protection factors of the host security quantization model meet the set specifications includes the following steps: Step U1: Obtain the scale, retrieval speed, and initial value of the algorithm coefficient and protection factor calculation parameters of a certain level of the host security quantization model in the current power system host standard compliance operation data to be managed; Step U2: Both the scale and retrieval speed of the algorithm coefficient and protection factor calculation parameters of the host security quantization model are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, and optimize whether there is an unfused situation for the algorithm coefficients and protection factors of the host security quantization model; Step U3: If only some of the algorithm coefficients and protection factors of the algorithm coefficient and protection factor calculation parameters of the host security quantization model are within the algorithm coefficient and protection factor calculation parameters of the host security quantization model, then go to Step U4; Step U4: Optimize the algorithm coefficients and protection factor setting specifications of the host security quantification model. Check whether the algorithm coefficients and protection factors are within the calculation parameters of the algorithm coefficients and protection factors of the host security quantification model. If the algorithm coefficients and protection factor setting specifications of the host security quantification model are within the calculation parameters of the algorithm coefficients and protection factors of the host security quantification model, then check if there are unmerged situations in the algorithm coefficients and protection factors of the host security quantification model. Otherwise, proceed to the next step. Step U5: Continue to obtain the algorithm coefficients and protection factors of other host security quantification models that are being managed for compliance operation in the current power system host, and perform operations to optimize the algorithm coefficients and protection factors of the host security quantification model if there are unoptimized situations.

4. The fusion calculation method for host security quantification according to claim 3, wherein: It also includes a method for determining whether to discard the algorithm coefficients and protection factors when the algorithm coefficients and protection factors of the host security quantification model are within the calculation parameters of the algorithm coefficients and protection factors of the host security quantification model, which includes the following steps: Step R1: Calculate the optimization range between the setting specifications of the algorithm coefficients and protection factors of the host security quantification model in the previous power system host being managed for compliance operation and the setting specifications of the algorithm coefficients and protection factors of the host security quantification model in the current power system host being managed for compliance operation. If the optimization range exceeds the preset interval, it is determined that the calculation of the algorithm coefficients and protection factors of the host security quantification model is abnormal. Step R2: If the optimization of the algorithm coefficients and protection factors of the host security quantification model is in an abnormal calculation state, then clear the duration of unmerged situations in the algorithm coefficients and protection factors of the host security quantification model and process the algorithm coefficients and protection factors of other host security quantification models for the power system host being managed for compliance operation. Step R3: If the algorithm coefficients and protection factors of the host security quantification model have not been updated, then obtain the current time and calculate the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model. Compare the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model with the preset interval of the duration of unmerged situations. If the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model exceeds the preset interval of the duration of unmerged situations, then optimize the algorithm coefficients and protection factors of the host security quantification model to have unmerged situations. If the stop accumulation time of the algorithm coefficients and protection factors of the host security quantification model does not exceed the duration of unmerged situations, then the processing of the algorithm coefficients and protection factors of the current host security quantification model is completed, and continue to process other power system hosts being managed for compliance operation with integrated models.

5. The fusion calculation method for host security quantification according to claim 4, characterized in that: If the algorithm coefficients and protection factors of a certain host security quantization model are trained from the power system host standard compliant operation data to be managed in the previous iteration of the training data, and the algorithm coefficients and protection factors of the host security quantization model are not trained in the current data, the system sets a maximum standard threshold. Before the maximum standard threshold is reached, the algorithm coefficients and protection factors of the host security quantization model are not optimized for errors. After that, the unscented deep belief network in the deep belief network model is used to predict the algorithm coefficients and protection factor area of the current host security quantization model based on the algorithm coefficients and protection factor area of the host security quantization model in the previous iteration. The solution of the prediction equation is used as the algorithm coefficients and protection factor area of the current host security quantization model; If the characteristics of the next host security quantization model, that is, the power system host standard compliant operation data to be managed, are fused into the algorithm coefficient and protection factor area position of the host security quantization model and match the algorithm coefficient and protection factor area of the current host security quantization model, the algorithm coefficients and protection factors of the host security quantization model caused by the fusion algorithm error are optimized and disappear; If the maximum standard threshold is reached, it is directly regarded that the algorithm coefficients and protection factors of the host security quantization model disappear, and the system deletes the algorithm coefficient and protection factor data training data of this host security quantization model; If the algorithm coefficients and protection factors of the host security quantization model reappear during the period when the maximum standard threshold is not reached, it is optimized that the algorithm coefficients and protection factors of the host security quantization model are not fused for a short time.

6. The fusion calculation method for host security quantification according to claim 5, characterized in that: The time to obtain the power system host standard compliant operation data to be managed is set according to the standard compliant operation requirements, and different numbers of power system host standard compliant operation data are obtained per second.

7. A fusion computing method for host security quantification according to any one of claims 1-6, characterized in that: This method is implemented through the algorithm coefficient and protection factor fusion component of the host security quantization model, the power system host monitoring component, and the feature non-fusion optimization component of the host security quantization model; The algorithm coefficient and protection factor fusion component of the host security quantization model is used to fuse the algorithm coefficients and protection factors of the host security quantization model for the power system host standard compliant operation data to be managed obtained by the power system host monitoring component and obtain the calculation parameters and calculation parameter data of the algorithm coefficients and protection factors of the host security quantization model; The power system host monitoring component is used to obtain the power system host standard compliant operation data to be managed, set the supervision area position, and after matching the specific fusion code with the algorithm coefficient and protection factor calculation parameters and calculation parameter data of the host security quantization model obtained by the algorithm coefficient and protection factor fusion component of the host security quantization model, transmit them to the feature non-fusion optimization component of the host security quantization model; The feature non-fusion optimization component of the host security quantization model is used to output data trainer data after receiving the algorithm coefficients, protection factor calculation parameters, and calculation parameter data of the host security quantization model that have been matched with specific fusion codes, search for the data trainer data at different levels, optimize whether the algorithm coefficients and protection factors of this host security quantization model are located within the supervision area according to the supervision area location, and update the data trainer data again. According to the updated data trainer data, optimize whether the algorithm coefficients and protection factors of the host security quantization model should issue an alarm.

8. A fusion calculation method for host security quantification according to claim 7, characterized in that: The data trainer data includes: the algorithm coefficient and protection factor fusion code of the host security quantization model, the algorithm coefficient and protection factor calculation parameters of the host security quantization model, whether the algorithm coefficient and protection factor of the host security quantization model enter the algorithm coefficient and protection factor calculation parameters of the host security quantization model, the time when the algorithm coefficient and protection factor of the host security quantization model enter the algorithm coefficient and protection factor calculation parameters of the host security quantization model, and whether the algorithm coefficient and protection factor of the host security quantization model have issued an alarm.

9. A fusion calculation method for host security quantification according to claim 7, characterized in that: The power system host data selection library uses a human-computer interaction algorithm to perform timed supervision on the data to be optimized of the power system host.

Images