A cloud-edge collaborative-based power resource scheduling method and system

Abstract

The application is suitable for the technical field of power resource scheduling, and provides a power resource scheduling method and system based on cloud edge cooperation, which comprises the following steps: each edge end obtains power equipment operation data according to a preset acquisition frequency and a communication protocol, and pre-processes the obtained data; each edge end extracts a characteristic vector based on the pre-processed power equipment operation data by using a principal component analysis method; each edge end inputs the characteristic vector into a trained support vector regression model to perform power resource demand prediction, and obtains a target resource demand prediction value; each edge end sends the target resource demand prediction value after encryption to a cloud end, so that the cloud end analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy, and economic, efficient and sustainable power resource scheduling can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power resource scheduling, in particular to a power resource scheduling method and system based on cloud-edge collaboration. BACKGROUND

[0002] In the Internet of Things environment, a large number of devices continuously generate a large amount of data, such as smart meters recording real-time power consumption data and various sensors monitoring the operating parameters of power grid equipment. If all these data are directly transmitted to the cloud for processing, it will cause many problems. On the one hand, network congestion will occur, for example, in the power grid Internet of Things system of a large industrial park, thousands of devices simultaneously transmitting data may cause local network or even the entire wide area network to be congested, affecting the normal transmission of data. On the other hand, a large amount of data transmission needs to be queued for processing, which will cause obvious delay. For the power grid, many real-time monitoring and control operations have very high response speed requirements, such as rapid detection and isolation of power grid faults. A slight delay may cause the fault influence range to expand, endanger the safe and stable operation of the power grid, and even cause power outages and other serious consequences.

[0003] Therefore, a power resource scheduling method and system based on cloud-edge collaboration are needed. SUMMARY

[0004] The embodiments of the present application provide a power resource scheduling method and system based on cloud-edge collaboration, which are used to solve the problem of unreasonable allocation of power resources.

[0005] The first aspect of the embodiments of the present application provides a power resource scheduling method based on cloud-edge collaboration, comprising:

[0006] Each edge end obtains power equipment operation data according to a preset acquisition frequency and a communication protocol, and pre-processes the obtained data;

[0007] Each edge end extracts a feature vector based on the pre-processed power equipment operation data by using a principal component analysis method;

[0008] Each edge end inputs the feature vector into a trained support vector regression model to predict power resource demand and obtain a target resource demand prediction value;

[0009] Each edge end sends the target resource demand prediction value after encryption to the cloud end, so that the cloud end analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy.

[0010] Further, each edge end inputs the feature vector into the trained support vector regression model to predict power resource demand and obtain a target resource demand prediction value, comprising:

[0011] Obtain historical power equipment operation data and perform data preprocessing, and divide the preprocessed data into a training set and a test set;

[0012] Select a support vector regression model architecture with a radial basis function kernel, and input the training set into the support vector regression model for training;

[0013] Optimize using an optimization algorithm to obtain a trained support vector regression model.

[0014] Further, the selection of the support vector regression model architecture with the radial basis function kernel, and the input of the training set into the support vector regression model for training, comprises:

[0015]

[0016] Wherein: is the resource demand prediction value of the input sample , is the number of support vectors, and is the Lagrange multiplier, is the kernel function, is the bias term.

[0017] Further, the target resource demand prediction value is encrypted and sent to the cloud by each edge terminal, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy, comprising:

[0018] The edge terminal and the cloud determine a multi-party secure computing protocol for secret sharing and homomorphic encryption;

[0019] The edge terminal splits its own power resource demand prediction value into several shares based on the multi-party secure computing protocol, and sends the several shares to other edge terminals and the cloud;

[0020] After receiving the shares of other edge terminals, the edge terminal uses a public key to homomorphically encrypt the received shares and sends them to the cloud;

[0021] The cloud aggregates and calculates all the encrypted shares based on the homomorphic encryption property and the multi-party secure computing protocol, and decrypts to obtain the sum of the power resource demand prediction values of each edge terminal.

[0022] Further, the edge terminal splits its own power resource demand prediction value into several shares based on the multi-party secure computing protocol, and sends the several shares to other edge terminals and the cloud, comprising:

[0023]

[0024] wherein: is the power resource demand prediction value of the nth edge terminal, is the secret share generated by the nth edge terminal, wherein , is a modulus.

[0025] Further, the cloud aggregates all the encrypted shares based on the homomorphic encryption property and the multi-party secure computation protocol, and obtains the sum of the power resource demand prediction values of the edge terminals after decryption, including:

[0026]

[0027]

[0028] wherein: is the secret share received and encrypted by the nth edge terminal from the mth edge terminal, is a decryption function. Further, the edge terminals send the target resource demand prediction values to the cloud after encryption, so that the cloud analyzes the decrypted target resource demand prediction values to obtain the corresponding power resource scheduling strategy, including:

[0029] Obtain historical resource demand values and use the DBSCAN algorithm to automatically cluster to obtain historical resource demand clusters, and statistically obtain the maximum boundary value of each cluster;

[0030] Calculate the similarity array of the target resource demand prediction value to all cluster center nodes in the historical resource demand cluster by the Pearson similarity algorithm;

[0031] Sort the similarity array in ascending order, and determine the most frequent power scheduling strategy in the cluster with the first array number as the first power scheduling strategy;

[0032] Respectively construct the objective function with the maximum power resource scheduling utilization rate and the minimum resource waste as the target parameters, and determine the constraint conditions, including the power generation power constraint, the power transmission capacity constraint and the power balance constraint;

[0033] Determine the first power scheduling strategy based on the objective function and the constraint conditions;

[0034] Determine the target scheduling strategy according to the first scheduling strategy and the second scheduling strategy.

[0035] Determine the target scheduling strategy according to the first scheduling strategy and the second scheduling strategy.

[0036] ​​​Further, the objective function is constructed by taking the maximum utilization rate of power resource scheduling and the minimum resource waste as target parameters, including:

[0037]

[0038] wherein: is the effective resource, is the total resource supply, is the actual effective resource, is the maximum power generation of the power generation unit , and is the weight coefficient.

[0039] The expression of the constraint condition is as follows:

[0040] Power generation constraint:

[0041]

[0042] The power generation unit itself has a minimum and maximum power generation limit, which ensures that the power generation unit operates within a safe and stable power generation range;

[0043] Transmission capacity constraint:

[0044]

[0045] The transmission line from the power generation unit to the power consumption area has a maximum transmission capacity limit, which avoids overloading the transmission line and ensures stable transmission of electric energy;

[0046] Power balance constraint:

[0047]

[0048] Each power consumption area receives an amount of electricity that matches its predicted demand at the time, achieving accurate supply and demand balance, ensuring that the power supply of the power consumption area can meet its demand, avoiding power shortage or excess;

[0049] Non-negative constraint:

[0050]

[0051] The physical meaning of the transmission power is non-negative.

[0052] The second aspect of the embodiment of the application provides a power resource scheduling system based on cloud edge collaboration, comprising:

[0053] A data acquisition and preprocessing unit is configured to acquire power equipment operation data according to a preset acquisition frequency and a communication protocol, and to preprocess the acquired data.

[0054] a feature vector extraction unit configured to extract, by each edge terminal, a feature vector based on the preprocessed power equipment operation data by using principal component analysis;

[0055] a target resource demand prediction value determination unit configured to input the feature vector into a trained support vector regression model by each edge terminal to predict power resource demand and obtain a target resource demand prediction value;

[0056] a power resource scheduling strategy determination unit configured to send the target resource demand prediction value to the cloud after encryption by each edge terminal, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy.

[0057] From the above technical solutions, the embodiments of the present application have the following advantages:

[0058] In the present application, each edge terminal acquires power equipment operation data according to a preset acquisition frequency and a communication protocol, pre-processes the acquired data, extracts a feature vector based on the preprocessed power equipment operation data by using principal component analysis, inputs the feature vector into a trained support vector regression model to predict power resource demand and obtain a target resource demand prediction value, and sends the target resource demand prediction value to the cloud after encryption, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy. In the present application, the edge terminal acquires and pre-processes power equipment data according to a preset acquisition and communication specification, extracts features by principal component analysis, predicts power resource demand by using a support vector regression model, sends the encrypted prediction value to the cloud, and then the cloud analyzes the decrypted value to obtain an effective power resource scheduling strategy, thereby realizing an efficient and secure process from data acquisition to scheduling strategy generation in power resource management.

[0059] Other advantages, objects, and features of the present application will be set forth in part in the following specification taken in conjunction with the accompanying drawings, and in part will become apparent to those skilled in the art from a consideration of the following specification and drawings, or can be learned from the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 FIG. 1 is a flowchart of an embodiment of a power resource scheduling method based on cloud-edge collaboration in the present application. DETAILED DESCRIPTION

[0061] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of these terms herein is to be construed as interchangeable in order to describe the embodiments of the present application as outlined herein in any sequence or order. Furthermore, the terms "comprising", "including", "containing", and "having" and their conjugates herein are used inclusively and not exclusive. That is, they are used to "specify the presence of stated steps, operations, elements, and / or components" but do not preclude the presence or addition of one or more other steps, operations, elements, components, and / or groups thereof that are merely illustrative.

[0062] Embodiment one

[0063] Referring to Figure 1 An embodiment of a power resource scheduling method based on cloud-edge collaboration in the present application includes the following steps:

[0064] S11. Each edge end obtains power equipment operation data according to a preset acquisition frequency and a communication protocol, and pre-processes the obtained data;

[0065] Rich and high-frequency power equipment operation data are collected from each edge end, which covers multiple types of equipment, such as transformers in substations, recording oil temperature, winding temperature, load current, and output voltage; switch cabinets in distribution rooms, monitoring contact temperature and bus current; distributed energy generation equipment such as photovoltaic panels, counting light intensity, power generation, and battery panel temperature, etc., forming an initial data set , the dimension of which is , wherein represents the number of samples, and the data entries accumulated at different time points according to the set acquisition frequency, such as one hour of 60 samples collected per minute, represents the number of features, that is, the dimension after the operation parameters of each type of equipment are summarized. At the same time, the real value of the power resource demand at each sample time is recorded, which is summarized into a vector (dimension ), the resource demand value is quantified in units of power (kW) or electricity (kWh), reflecting the actual power resource amount required by the corresponding equipment group or local power grid at the collection time, which may be derived from historical electricity consumption records, metering combined with power grid scheduling instructions.

[0066] The data set Preprocessing is performed. For missing values, if a feature value is missing in a sample, the transformer load current data is missing, the mean of adjacent time points or linear interpolation is used based on the device running stability and the correlation of the surrounding samples; for different dimensions, all features are standardized, such as transformer oil temperature (degrees Celsius) and photovoltaic power (kilowatts) are uniformly processed.

[0067] S12. Each edge end extracts a feature vector based on the preprocessed power equipment operation data by principal component analysis;

[0068] Calculate the covariance matrix: based on the standardized data matrix , the covariance matrix is calculated according to the matrix operation rules, formula: , The dimension is , the elements on the diagonal show the variance of each feature itself, and the non-diagonal elements reflect the correlation between different features, such as the close correlation between transformer oil temperature and load current change, which corresponds to the prominent covariance value, which helps to mine the potential structure of the data.

[0069] Eigenvalue decomposition: use linear algebra algorithm to decompose the covariance matrix , get a set of eigenvalues , , …, , where the order satisfies , and the corresponding eigenvector , , …, . The size of the eigenvalue directly reflects the amount of information carried by the corresponding principal component, and the eigenvector defines the composition of the principal component from the direction dimension, for example, the weight of each component of the eigenvector implies which original features have high weights in the corresponding principal component.

[0070] Principal component selection: set the cumulative variance contribution rate threshold to determine the number of retained principal components. The cumulative variance contribution rate calculation formula is: , in descending order of eigenvalues, gradually accumulate the proportion, and determine when the threshold is reached , the first eigenvalues are selected to form the projection matrix , for example, in regional power grid edge data processing, the original feature of several tens of dimensions can be reduced to 5-8 principal components through this screening, which greatly simplifies the data and retains the key information.

[0071] Generate principal component data: project the standardized data matrix to the selected principal component space through linear transformation to obtain the principal component matrix , the operation is based on:

[0072] The above completes the core operations of PCA feature extraction. Each column represents a principal component, which integrates the original multi-feature information and is subsequently input into SVR modeling. Furthermore, each principal component corresponds to the power resource demand at the corresponding time point. The correlation needs to be explored using SVR.

[0073] S13. Each edge end inputs the feature vector into the trained support vector regression model to predict the power resource demand and obtain the target resource demand prediction value.

[0074] Step S13 includes the following:

[0075] 1. Acquire historical power equipment operation data and perform data preprocessing, then divide the preprocessed data into training and testing sets;

[0076] 2. Select a support vector regression model architecture with radial basis function kernels, and input the training set into the support vector regression model for training;

[0077] Principal component matrix and the corresponding real value vector of electricity resource demand The model is reasonably divided into training and test sets according to time series or random sampling methods to ensure that the training set reflects the normal operating rules and the test set tests the model's generalization and prediction capabilities for new scenarios.

[0078] The SVR with selected radial basis function kernels is chosen because it has a strong fit to nonlinear relationships, and its basic regression prediction formula is applicable to the input. Output as follows:

[0079]

[0080] in: For input samples The resource demand forecast, The number of support vectors, and For Lagrange multipliers, For kernel function, This is a bias term.

[0081] By combining cross-validation and grid search, the model is tested within the parameter range, and the validation set is divided for evaluation each time. The combination of parameters that minimizes the MSE of the validation set is then locked in, thereby enhancing the model's adaptability.

[0082] training set , Input a pre-tuned SVR model, and use iterative optimization techniques such as Sequence Minimum Optimization (SMO) to adjust the model parameters. , , The model deeply studies the internal mapping between the principal component features and the power resource demand, and the process conforms to the principle of structural risk minimization of SVR, and the complexity constraint prevents overfitting, and the fitting curve approximates the real resource demand trend.

[0083] 3. Optimization is performed by using an optimization algorithm to obtain a trained support vector regression model.

[0084] Based on the trained SVR model, the influence of each principal component is determined, and for the test set , the principal component features are randomly replaced to disturb the inherent distribution, and the other columns are kept constant. After replacing the first column, the principal component test set is , the new predicted value vector is input into the model , and the root mean square error (RMSE) change is calculated:

[0085]

[0086] Among them: is the number of test set samples, is the predicted RMSE of the test set when the features are not replaced, is the predicted RMSE of the test set when the features are not replaced, is the predicted RMSE of the test set when the features are not replaced,

[0087] The trained and explained SVR model is used to predict the test set , and the predicted result vector is obtained , which is compared with the true value , and a series of evaluation indicators are used to measure the accuracy, such as mean square error (MSE):

[0088]

[0089] Root mean square error (RMSE), mean absolute error (MAE), etc. From different dimensions, the model's performance in predicting power resource demand is evaluated. If the RMSE reaches the expected accuracy, the model can be used for real-time prediction on the front edge, and the accurate resource demand prediction value is output to assist power grid dispatching.

[0090] S14. Each edge sends the encrypted target resource demand prediction value to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy.

[0091] In this embodiment, each edge sends the encrypted target resource demand prediction value to the cloud, including the following steps:

[0092] 1. Each edge end and cloud end determines a multi-party secure computation protocol of secret sharing and homomorphic encryption;

[0093] 2. Each edge end splits its own power resource demand prediction value into several shares based on the multi-party secure computation protocol, and sends the several shares to other edge ends and the cloud end;

[0094] 3. After receiving the shares of other edge ends, each edge end sends the received shares to the cloud end after homomorphic encryption using the public key;

[0095] 4. The cloud end aggregates all the encrypted shares based on the homomorphic encryption property and the multi-party secure computation protocol, and decrypts to obtain the sum of the power resource demand prediction values of each edge end.

[0096] Specifically, assume that there are edge ends, denoted as Each edge end has its own power resource demand prediction value, denoted as , and hopes to securely aggregate these prediction values in the cloud for subsequent analysis while ensuring the privacy of each edge end's prediction value.

[0097] First, a common key generation mechanism is established between the edge ends and the cloud end through a cryptographic protocol. For example, a homomorphic encryption-based key generation algorithm is used to generate a pair of public and private keys , the public key is used for encrypting data, and the private key is used for decrypting data. This key information needs to be securely distributed to all edge ends and the cloud end. At the same time, the protocol and algorithm used for multi-party secure computation are determined, for example, a scheme based on secret sharing and homomorphic encryption is selected. Secret sharing is to divide a secret into multiple shares, each share itself has no meaning, and only a sufficient number of shares can be combined to restore the original secret.

[0098] For each edge end , its power resource demand prediction value is processed using the secret sharing scheme. Here is a threshold, indicating that at least shares are needed to restore the original secret. For example, if the scheme is selected, then edge end will be decomposed into shares , so that the following relationship is satisfied:

[0099]

[0100] where is a large enough modulus to ensure that the calculation result is within a certain range. This The shares can be generated by Lagrange interpolation method or the like.

[0101] edge end The generated shares are sent to other edge ends, including the cloud end as a special participant. In this way, each edge end, including the cloud end, receives shares from all other participants.

[0102] After receiving the shares sent by other edge ends, each edge end (including the cloud end) uses the public key to homomorphically encrypt the received shares. Assuming that the edge end receives shares from other edge ends , the encrypted shares are denoted as , and the encryption process can be performed according to the specific formula of the homomorphic encryption algorithm. For example, for some homomorphic encryption algorithms, the encryption formula can be: , where and are parameters of the homomorphic encryption algorithm. Then, each edge end (including the cloud end) sends the encrypted shares to the cloud end. At this time, the cloud end receives all the encrypted shares .

[0103] The cloud end aggregates the received encrypted shares according to the multi-party secure computation protocol and the properties of homomorphic encryption. Since homomorphic encryption allows certain calculations to be performed on ciphertext, the cloud end can calculate the sum of all encrypted shares. For example, for additive homomorphic encryption, the cloud end calculates:

[0104]

[0105] According to the properties of homomorphic encryption, this sum should equal the sum of all edge end prediction values after decryption. That is:

[0106]

[0107] where: is a decryption function that uses the private key to decrypt.

[0108] Through the above multi-party secure computation process, the cloud end can calculate the sum of all edge end prediction values without knowing the specific prediction values of each edge end, thereby achieving privacy-preserving aggregation calculation of prediction values. At the same time, the entire process ensures the security of the key and the reliability of the communication to ensure data privacy and the accuracy of the calculation.

[0109] Specifically, step S14 further includes the following:​

[0110] 1. Obtain historical resource demand values and automatically cluster using DBSCAN algorithm to obtain historical resource demand cluster, and count the maximum boundary value of each cluster;

[0111] Apply DBSCAN algorithm to the obtained historical resource demand dataset, identify core points, boundary points and noise points according to the set parameters, and divide the data points into different clusters. Each cluster represents a group of data points with similar historical resource demand characteristics. The boundary value in each dimension is found by analyzing the coordinates of the points in the cluster.

[0112] 2. Calculate the similarity array of the target resource demand prediction value in all cluster center nodes in the historical resource demand cluster by Pearson similarity algorithm;

[0113] According to the calculation formula of Pearson similarity algorithm, the similarity of the target resource demand prediction value and each cluster center node is calculated. The value of similarity ranges from -1 to 1. The closer the value is to 1, the stronger the positive correlation is. The closer the value is to -1, the stronger the negative correlation is. The closer the value is to 0, the weaker the correlation is.

[0114] 3. Sort the similarity array in ascending order, and determine the most frequently used power dispatch strategy in the cluster with the first array number as the first power dispatch strategy;

[0115] After sorting the obtained similarity array in ascending order, the element corresponding to the first array number is the cluster with the lowest similarity to the target resource demand prediction value. In the cluster with the first array number, the number of times different power dispatch strategies are used is counted. The power dispatch strategy that is used the most times is determined as the first power dispatch strategy, which serves as the reference dispatch strategy when the current target resource demand prediction value is similar to the cluster, so as to reasonably arrange the dispatch of power resources.

[0116] 4. Construct the objective function with power resource dispatch utilization maximization and resource waste minimization as target parameters, and determine the constraint conditions, including power generation power constraint, power transmission capacity constraint and power balance constraint;

[0117] Let be the power transmitted from the power generation unit to the power consumption area at time , be the power resource demand prediction value of the power consumption area in the time period , be the maximum power generation of the power generation unit . First, calculate the total resource supply , the actual effective utilization of resources is , the resource waste is , wherein represents . That is, the objective function is as follows:

[0118]

[0119] wherein: and are weight coefficients for balancing the relative importance of the two goals of maximizing resource utilization and minimizing resource waste. For example, if more attention is paid to resource utilization, the value of may be relatively large; if more attention is paid to the reduction of resource waste, the value of may be relatively large, and the goal is to maximize .

[0120] Generation power constraint:

[0121]

[0122] The generation unit itself has a minimum and maximum generation power limit, which ensures that the generation unit operates within a safe and stable generation power range.

[0123] Transmission capacity constraint:

[0124]

[0125] The transmission line from the generation unit to the power consumption area has a maximum transmission capacity limit, which avoids overloading the transmission line and ensures stable transmission of electric energy.

[0126] Power balance constraint:

[0127]

[0128] Each power consumption area receives an amount of electricity that matches its point-in-time forecast demand, achieving accurate supply and demand balance, ensuring that the power supply of the power consumption area can meet its demand, avoiding power shortage or excess.

[0129] Non-negative constraint:

[0130]

[0131] The physical meaning of the delivered power is non-negative.

[0132] In the above formula: is the minimum generation power of the generation unit , and is the maximum transmission capacity of the transmission line from the generation unit to the power consumption area.

[0133] 5. determining a second power dispatching strategy based on the objective function and the constraint condition;

[0134] 6. determining a target dispatching strategy based on the first power dispatching strategy and the second power dispatching strategy.

[0135] Under the above premise of using the clustering algorithm for power resource dispatching, the maximization of resource utilization and the minimization of resource waste can also be considered, that is, a second layer of confirmation is made under the premise of meeting the above dispatching. The following steps can be used as an optional scheme, when the clustering algorithm has met the dispatching requirements, there is no need to calculate the optimization parameters.

[0136] The two goals of maximizing resource utilization and minimizing resource waste are interrelated and complementary. By combining the two goals, the efficient operation of the power system can be ensured while paying more attention to the rational allocation and utilization of resources, avoiding waste caused by excessive power generation and the situation that demand cannot be met due to insufficient power generation, which helps to achieve economic, efficient and sustainable power resource dispatching.

[0137] Embodiment two

[0138] An embodiment of a power resource dispatching system based on cloud-edge collaboration in the application includes the following steps:

[0139] A data acquisition and preprocessing unit is configured to acquire power equipment operation data according to a preset acquisition frequency and a communication protocol at each edge terminal, and to preprocess the acquired data.

[0140] A feature vector extraction unit is configured to extract feature vectors based on the preprocessed power equipment operation data at each edge terminal using a principal component analysis method.

[0141] A target resource demand prediction value determination unit is configured to input the feature vectors into a trained support vector regression model to predict power resource demand at each edge terminal, and to obtain a target resource demand prediction value.

[0142] A power resource dispatching strategy determination unit is configured to encrypt the target resource demand prediction value at each edge terminal and send it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value and obtains a corresponding power resource dispatching strategy.

[0143] The specific limitations of the system can be referred to the limitations of the method described above, which will not be repeated here. Each module in the above system can be realized by software, hardware and their combination, in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor calls and executes the operations corresponding to the above modules.

[0144] Those skilled in the art can understand that the units of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components of the examples have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0145] In the embodiments provided by the present application, it should be understood that the division of units is only a logical functional division, and there can be another division manner when actually implemented, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0146] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0147] It can be understood that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the description of the present application.

Claims

1. A power resource scheduling method based on cloud-edge collaboration, characterized in that, The method comprises the following steps: Each edge end acquires power equipment operation data according to a preset acquisition frequency and a communication protocol, and pre-processes the acquired data; Each edge end extracts a feature vector based on the pre-processed power equipment operation data by using a principal component analysis method; Each edge end inputs the feature vector into a trained support vector regression model to perform power resource demand prediction, and obtains a target resource demand prediction value; Each edge end encrypts the target resource demand prediction value and sends it to the cloud end, so that the cloud end analyzes the decrypted target resource demand prediction value and obtains a corresponding power resource scheduling strategy; each edge end encrypts the target resource demand prediction value and sends it to the cloud end, so that the cloud end analyzes the decrypted target resource demand prediction value and obtains a corresponding power resource scheduling strategy, which comprises the following steps: Obtain historical resource demand values and use a DBSCAN algorithm to automatically cluster, obtain historical resource demand clusters, and count the maximum boundary value of each cluster; Calculate the similarity array of the target resource demand prediction value to all cluster center nodes in the historical resource demand cluster by using a Pearson similarity algorithm; Sort the similarity array in ascending order, and determine the most frequent power scheduling strategy in the cluster with the first array number as the first power scheduling strategy; Construct a target function with the maximum utilization rate of power resource scheduling and the minimum waste of resources as target parameters, and determine the constraint conditions, which include power generation power constraint, power transmission capacity constraint and power balance constraint; Determine the second power scheduling strategy based on the target function and the constraint conditions; Determine the target scheduling strategy according to the first power scheduling strategy and the second power scheduling strategy. 2.The cloud-edge collaboration based power resource scheduling method of claim 1, wherein, Each edge end inputs the feature vector into a trained support vector regression model to perform power resource demand prediction, and obtains a target resource demand prediction value, which comprises the following steps: Obtain historical power equipment operation data and pre-process the data, and divide the pre-processed data into a training set and a test set; Select a support vector regression model architecture with a radial basis function kernel, input the training set into the support vector regression model for training; Optimize by using an optimization algorithm to obtain a trained support vector regression model. 3.The cloud-edge collaboration based power resource scheduling method of claim 2, wherein, Select a support vector regression model architecture with a radial basis function kernel, input the training set into the support vector regression model for training, which comprises the following steps: wherein: is a resource requirement prediction value for an input sample , is a number of support vectors, and are Lagrange multipliers, is a kernel function, is a bias term. 4.The cloud-edge collaboration based power resource scheduling method of claim 1, wherein, Each edge end encrypts the target resource demand prediction value and sends it to the cloud end, so that the cloud end analyzes the decrypted target resource demand prediction value and obtains a corresponding power resource scheduling strategy, which comprises the following steps: Each edge end and the cloud end determine a multi-party secure computing protocol of secret sharing and homomorphic encryption; Each edge end splits its power resource demand prediction value into several shares based on the multi-party secure computing protocol, and sends the several shares to other edge ends and the cloud end; After receiving the shares of other edge ends, each edge end homomorphically encrypts the received shares by using a public key and sends them to the cloud end; The cloud aggregates all encrypted shares received based on the homomorphic encryption property and the multi-party secure computation protocol, and decrypts to obtain the sum of the power resource demand prediction values of the edge ends.

5. The cloud-edge collaboration based power resource scheduling method of claim 4, wherein, Each edge end splits its power resource demand prediction value into several shares based on the multi-party secure computation protocol, and sends the shares to other edge ends and the cloud, including: wherein: is the predicted value of the power resource demand for the th edge end, is the private share generated for the th edge end, wherein , is the modulus.

6. The cloud-edge collaboration based power resource scheduling method of claim 5, wherein, The cloud aggregates all encrypted shares received based on the homomorphic encryption property and the multi-party secure computation protocol, and decrypts to obtain the sum of the power resource demand prediction values of the edge ends, including: wherein: is the th edge end received and encrypted secret share from the th edge end, is a decryption function. 7.The cloud-edge collaboration based power resource scheduling method of claim 1, wherein, The objective function is constructed with power resource scheduling utilization maximization and resource waste minimization as target parameters, including: wherein: is the resource effectively utilized, is the total resource supplied, is the resource actually effectively utilized, is the maximum power generated by the power generation unit is the power generated by the power generation unit, and is the weight coefficient. 8.The cloud-edge collaboration based power resource scheduling method of claim 1, wherein, The expressions of the constraint conditions are as follows: Generation power constraint: The generation unit itself has minimum and maximum generation power limits, which ensures that the generation unit operates within a safe and stable generation power range; Transmission capacity constraint: The transmission line from the generation unit to the power consumption area has a maximum transmission capacity limit, which avoids overload of the transmission line and ensures stable transmission of electric energy; Power balance constraint: Each power consumption area receives an amount of power that matches its point prediction demand, achieving accurate supply and demand balance, ensuring that the power supply of the power consumption area can meet its demand, avoiding power shortage or excess; Non-negative constraint: The physical meaning of the transmission power is non-negative.

9. A cloud-edge collaboration based power resource scheduling system, characterized in that, The method of any one of claims 1-8, including: A data acquisition and preprocessing unit for each edge end to acquire power equipment operation data according to a preset acquisition frequency and communication protocol, and to preprocess the acquired data; A feature vector extraction unit for each edge end to extract a feature vector based on the preprocessed power equipment operation data using principal component analysis; An objective resource demand prediction value determination unit for each edge end to input the feature vector into a trained support vector regression model to predict power resource demand, obtaining an objective resource demand prediction value; A power resource scheduling strategy determination unit for each edge end to encrypt the objective resource demand prediction value and send it to the cloud, so that the cloud analyzes the decrypted objective resource demand prediction value to obtain a corresponding power resource scheduling strategy.

Images