Electric power resource scheduling method and system based on cloud edge collaboration

Abstract

The invention is suitable for the technical field of electric power resource scheduling, and provides an electric power resource scheduling method and system based on cloud edge collaboration, and the method comprises the steps: each edge end obtains the operation data of electric power equipment according to a preset collection frequency and a communication protocol, and carries out the preprocessing of the obtained data; each edge end uses a principal component analysis method to extract a feature vector based on the preprocessed power equipment operation data; each edge end inputs the feature vector into a trained support vector regression model for power resource demand prediction to obtain a target resource demand prediction value; and each edge end encrypts the target resource demand prediction value and then sends the encrypted target resource demand prediction value 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 invention relates to the technical field of power resource scheduling, and in particular to a power resource scheduling method and system based on cloud-edge collaboration. Background Art

[0002] In the IoT environment, a large number of devices continuously generate a large amount of data, such as smart meters that record electricity consumption data in real time, and various sensors that monitor the operating parameters of power grid equipment. If all this data is directly transmitted to the cloud for processing, many problems will arise. On the one hand, network congestion will occur. For example, in the power grid IoT system of a large industrial park, thousands of devices transmit data at the same time, which may cause congestion in the local network or even the entire wide area network, affecting the normal transmission of data. On the other hand, a large amount of data transmission needs to queue up for processing, which will cause significant delays. For the power grid, many real-time monitoring and control operations require extremely high response speeds, such as the rapid detection and isolation of power grid faults. A slight delay may cause the scope of the fault to expand, endangering the safe and stable operation of the power grid, and even causing serious consequences such as power outages.

[0003] In view of this, a power resource scheduling method and system based on cloud-edge collaboration is needed. Summary of the invention

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

[0005] A first aspect of an embodiment of the present application provides a power resource scheduling method based on cloud-edge collaboration, including:

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

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

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

[0009] Each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy.

[0010] Furthermore, each edge terminal inputs the feature vector into a trained support vector regression model to predict power resource demand, and obtains a target resource demand prediction value, including:

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

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

[0013] The optimization algorithm is used for optimization to obtain the trained support vector regression model.

[0014] Furthermore, the support vector regression model architecture of selecting the radial basis function kernel, inputting the training set into the support vector regression model for training, includes:

[0015]

[0016] Where: f(z i ) is the input sample z i The resource demand prediction value, l is the number of support vectors, α j and is the Lagrange multiplier, K(z i ,z j ) is the kernel function and b is the bias term.

[0017] Furthermore, each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy, including:

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

[0019] Each edge terminal splits its own power resource demand forecast value into a plurality of shares based on the multi-party secure computing protocol, and sends the plurality of shares to other edge terminals and the cloud;

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

[0021] The cloud performs aggregate calculations on all received encrypted shares based on homomorphic encryption properties and a multi-party secure computing protocol, and obtains the sum of the predicted values ​​of power resource demand at each edge terminal after decryption.

[0022] Furthermore, each edge terminal splits its own power resource demand forecast value into a plurality of shares based on the multi-party secure computing protocol, and sends the plurality of shares to other edge terminals and the cloud, including:

[0023]

[0024] Where: v i is the power resource demand forecast value of the ith (i=1,2,…,n) edge terminal, s ij The private share generated for the i-th edge node, where (j=1, 2, …, n) and M is the modulus.

[0025] Furthermore, the cloud performs aggregate calculation on all received encrypted shares based on homomorphic encryption properties and multi-party secure computing protocols, and obtains the sum of the power resource demand forecast values ​​of each edge terminal after decryption, including:

[0026]

[0027] Where: Enc(s ij ) is the secret share from the jth edge end received and encrypted by the ith edge end, and Dec is the decryption function.

[0028] Furthermore, each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy, including:

[0029] Obtain historical resource demand values ​​and use the DBSCAN algorithm to automatically cluster them 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 forecast value in all cluster center nodes in the historical resource demand clustering cluster by using the Pearson similarity algorithm;

[0031] Sorting the similarity array in ascending order, and determining the power dispatching strategy that is used most times in the cluster with the first array sequence number as the first power dispatching strategy;

[0032] The objective function is constructed with maximizing the utilization rate of power resource scheduling and minimizing resource waste as target parameters, and constraint conditions are determined, wherein the constraint conditions include power generation constraint, transmission capacity constraint and power balance constraint;

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

[0034] A target scheduling strategy is determined according to the first scheduling strategy and the second scheduling strategy.

[0035] Furthermore, the objective function is constructed by taking the maximization of the power resource scheduling utilization rate and the minimization of the resource waste as the objective parameters, including:

[0036]

[0037] Among them: U is the effectively utilized resources, S is the total resource supply, W is the actually effectively utilized resources, is the power generation unit g k is the maximum power generation, α and β are weight coefficients.

[0038] The expressions of the constraints are as follows:

[0039] Power generation constraints:

[0040]

[0041] The power generation unit itself has minimum and maximum power generation limits, which ensures that the power generation unit operates within a safe and stable power generation range;

[0042] Transmission capacity constraints:

[0043]

[0044] The transmission lines from the power generation unit to the power consumption area have a maximum transmission capacity limit, which avoids overloading of the transmission lines and ensures the stable transmission of electric energy;

[0045] Power balance constraints:

[0046]

[0047] The amount of electricity received by each power consumption area must match its predicted demand at that point in time to achieve an accurate balance between supply and demand, ensuring that the power supply in the power consumption area can meet its needs and avoid power shortages or surpluses;

[0048] Non-negativity constraints:

[0049]

[0050] The physical meaning of the transmitted electricity is limited to non-negative.

[0051] A second aspect of an embodiment of the present application provides a power resource scheduling system based on cloud-edge collaboration, including:

[0052] The data acquisition and preprocessing unit is used for each edge terminal to acquire the power equipment operation data according to the preset acquisition frequency and communication protocol, and preprocess the acquired data;

[0053] A feature vector extraction unit is used for each edge end to extract a feature vector based on the preprocessed power equipment operation data by using a principal component analysis method;

[0054] A target resource demand prediction value determination unit is used for each edge end to input the feature vector into a trained support vector regression model to perform power resource demand prediction and obtain a target resource demand prediction value;

[0055] The power resource scheduling strategy determination unit is used for each edge terminal to encrypt the target resource demand prediction value and send it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy.

[0056] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0057] Each edge end of the present invention obtains power equipment operation data according to a preset acquisition frequency and communication protocol, and pre-processes the acquired data, and uses the principal component analysis method to extract feature vectors based on the pre-processed power equipment operation data; the feature vectors are input into a trained support vector regression model to predict power resource demand and obtain a target resource demand prediction value; the target resource demand prediction value is encrypted and sent to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy. The present invention obtains and pre-processes power equipment data according to preset acquisition and communication specifications through the edge end, extracts features through principal component analysis, and uses a support vector regression model to predict power resource demand, encrypts the prediction value and sends it to the cloud, so that the cloud can obtain an effective power resource scheduling strategy through decryption analysis, and realizes an efficient and secure process of power resource management from data acquisition to scheduling strategy generation.

[0058] Other advantages, objectives, and features of the present invention will be set forth in part in the following description, and in part will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of an embodiment of a power resource scheduling method based on cloud-edge collaboration in the present invention. DETAILED DESCRIPTION

[0060] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0061] Embodiment 1

[0062] See also Figure 1 , an embodiment of a power resource scheduling method based on cloud-edge collaboration in the present invention includes the following steps:

[0063] S11. Each edge terminal obtains the power equipment operation data according to the preset collection frequency and communication protocol, and pre-processes the obtained data;

[0064] Rich and high-frequency power equipment operation data are collected from each edge end. These data cover 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 busbar current; distributed energy generation equipment such as photovoltaic panels, statistics on light intensity, power generation, and battery panel temperature, etc., to form an initial data set X, whose dimension is m×n, where n represents the number of samples, data items accumulated at different time points according to the set collection frequency, such as collecting by minute, accumulating 60 samples for one hour, and m represents the number of features, that is, the dimension after the operation parameters of various types of equipment are summarized. At the same time, the actual value of power resource demand at each sample time is recorded and summarized into a vector y (dimension n×1). The resource demand value is quantified in units of power (kilowatts) or electricity (kilowatt-hours), reflecting the actual amount of power resources required by the corresponding equipment group or local power grid at the time of collection, which may be derived from historical power consumption records, meter measurement, and grid dispatch instructions.

[0065] Preprocess the data set X. For missing values, if the feature value in a sample is missing, or the transformer load current data is missing, fill it with the mean of adjacent time points or based on linear interpolation according to the equipment operation stability and the correlation of surrounding samples; standardize all features for different dimensions, such as uniformly processing transformer oil temperature (degrees Celsius) and photovoltaic power generation (kilowatts).

[0066] S12. Each edge end uses principal component analysis to extract feature vectors based on preprocessed power equipment operation data;

[0067] Calculate the covariance matrix: Based on the standardized data matrix X, calculate the covariance matrix C according to the matrix operation rules. The formula is: The C dimension is m×m. The elements on the diagonal show the variance of each feature, and the off-diagonal elements reflect the correlation between different features. For example, the changes in transformer oil temperature and load current may be closely related, and the corresponding covariance values ​​are prominent, which helps to subsequently explore the hidden structure of the data.

[0068] Eigenvalue decomposition: Use linear algebra algorithms to perform eigenvalue decomposition on the covariance matrix C to obtain a set of eigenvalues ​​λ1, λ2, …, λ m, where the order satisfies λ1≥λ2≥…≥λ m ≥0, and the corresponding eigenvectors v1, v2, …, v m 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 directional dimension. For example, the weights of each component of a certain eigenvector indicate which original features have high weights in the corresponding principal component.

[0069] Principal component selection: Set the cumulative variance contribution rate threshold to determine the number of principal components to retain. Cumulative variance contribution rate calculation formula: In descending order of eigenvalues, the proportions are gradually accumulated, and k is determined when the threshold is reached. The eigenvectors corresponding to the first k eigenvalues ​​are selected to form the projection matrix P. For example, in the data processing at the edge of the regional power grid, this screening may reduce dozens of dimensional original features to 5-8 dimensional principal components, greatly simplifying the data while retaining key information.

[0070] Generate principal component data: Project the standardized data matrix X to the selected principal component space through linear transformation to obtain the principal component matrix Z. The calculation basis is: Z = X × P.

[0071] The above completes the core operation of PCA feature extraction. Each column of Z is the principal component, which integrates the original multi-feature information and is subsequently input into SVR modeling. Each principal component is associated with the power resource demand y at the corresponding moment to be mined by SVR.

[0072] S13. Each edge terminal 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;

[0073] Step S13 includes the following:

[0074] 1. Obtain historical power equipment operation data and perform data preprocessing, and divide the preprocessed data into training sets and test sets;

[0075] 2. Select the support vector regression model architecture with radial basis function kernel, and input the training set into the support vector regression model for training;

[0076] The principal component matrix Z and the corresponding real value vector y of power resource demand are reasonably divided into training set and test set according to time series or random sampling, ensuring that the training set reflects the regular operating rules and the test set verifies the model's generalization and prediction capabilities for new scenarios.

[0077] The radial basis function kernel SVR is selected because it has a strong fit to nonlinear relationships and its basic regression prediction formula is i Output f(z i )as follows:

[0078]

[0079] Where: f(z i ) is the input sample z i The resource demand prediction value, l is the number of support vectors, α j and is the Lagrange multiplier, K(z i ,z j ) is the kernel function and b is the bias term.

[0080] By collaborating with cross-validation and grid search, we test the parameter range, divide the validation set to evaluate the model each time, and lock in the parameter combination that minimizes the MSE of the validation set to enhance the adaptability of the model.

[0081] The training set Z train ,y train Input the SVR model with adjusted parameters, and use iterative optimization methods such as sequential minimum optimization (SMO) to adjust the model parameters (α j , b, etc.), allowing the model to deeply study the intrinsic mapping between the main component characteristics and the power resource demand. The process is in line with the SVR structural risk minimization principle, the complexity is constrained to prevent overfitting, and the fitting curve is close to the actual resource demand trend.

[0082] 3. Use the optimization algorithm to optimize and obtain the trained support vector regression model.

[0083] Based on the trained SVR model, the influence of each principal component is determined. test , randomly permute the principal component features column by column to disrupt the inherent distribution, keep other columns constant, and permute the jth column. The principal component test set is Input the model to get the new prediction value vector Calculate the change in root mean square error (RMSE):

[0084]

[0085] Where: n test is the number of test set samples, RMSE original RMSE for test set prediction when features are not replaced, ΔRMSE j The larger the value is, the more critical the j-th principal component is to predicting the demand for electricity resources. Sorting in this way clarifies the "focus" of the model and helps to focus on core features or detect abnormal principal components.

[0086] The test set Z is trained with a mature and optimized SVR model. test Prediction, get the prediction result vector and the true value y test In contrast, a series of evaluation indicators are used to measure accuracy, such as mean square error (MSE):

[0087]

[0088] The root mean square error (RMSE), mean absolute error (MAE), etc., weigh the model from different dimensions to predict the performance of power resource demand. If the RMSE reaches the expected accuracy, the model can be used for real-time prediction at the front-line edge, and output accurate resource demand estimates based on new data input to assist power grid scheduling.

[0089] S14. Each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy.

[0090] In this embodiment, each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, including the following steps:

[0091] 1. Each edge and cloud end determines the multi-party secure computing protocol of secret sharing and homomorphic encryption;

[0092] 2. Each edge splits its own power resource demand forecast value into several shares based on the multi-party secure computing protocol, and sends several shares to other edge terminals and the cloud;

[0093] 3. After receiving the shares from other edge terminals, each edge terminal uses the public key to homomorphically encrypt the received shares and then sends them to the cloud;

[0094] 4. Based on the homomorphic encryption properties and multi-party secure computing protocol, the cloud aggregates all the encrypted shares received and decrypts them to obtain the total predicted value of power resource demand at each edge.

[0095] Specifically, suppose there are n edge ends, denoted as P1, P2, ..., P n Each edge has its own power resource demand forecast value, which is recorded as v1, v2, ..., v n , we hope to securely aggregate these prediction values ​​in the cloud for subsequent analysis while ensuring the privacy of each edge prediction value.

[0096] First, a common key generation mechanism is established between the edge and the cloud through cryptographic protocols. For example, a key generation algorithm based on homomorphic encryption is used to generate a public-private key pair (pk, sk). The public key pk is used to encrypt data, and the private key sk is used to decrypt data. This key information needs to be securely distributed to all edge terminals and the cloud. At the same time, determine the protocols and algorithms used for multi-party secure computing. For example, a solution 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. Only by combining a sufficient number of shares can the original secret be restored.

[0097] For each edge Pi (i=1,2,…,n), and its power resource demand forecast value v i The (t,n) secret sharing scheme is used. Here t is a threshold, indicating that at least t shares are needed to recover the original secret. For example, if the (2,n) scheme is selected, then the edge end P i will be decomposed into n shares s i1 ,s i2 ,…,s in , so that the following relationship is satisfied:

[0098]

[0099] Where M is a sufficiently large modulus to ensure that the calculation result is within a certain range. This n-share can be generated by Lagrange interpolation method and other methods.

[0100] Edge P i The generated share s ij It is sent to the other n-1 edge terminals, including the cloud terminal as a special participant. In this way, each edge terminal including the cloud terminal receives the share from all other participants.

[0101] After receiving the shares sent by other edge terminals, each edge terminal (including the cloud) uses the public key to homomorphically encrypt the received shares. i Received from other edge terminals P j The share is s ji , the encrypted share is recorded as Enc(s ji ), 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 may be: Where g and p are the parameters of the homomorphic encryption algorithm. Then, each edge (including the cloud) sends the encrypted shares to the cloud. At this point, the cloud receives all the encrypted shares Enc(s ij ).

[0102] The cloud performs aggregate calculations on the received encrypted shares based on the multi-party secure computation protocol and the properties of homomorphic encryption. Since homomorphic encryption allows specific calculations to be performed on ciphertext, the cloud can calculate the sum of all encrypted shares. For example, for additive homomorphic encryption, the cloud calculates:

[0103]

[0104] According to the properties of homomorphic encryption, this sum should be equal to the sum of all edge predictions after decryption. That is:

[0105]

[0106] Among them: Dec is the decryption function, and the private key sk is used for decryption.

[0107] Through the above multi-party secure computing process, the cloud can calculate the sum of all edge prediction values ​​without knowing the specific prediction value of each edge, thereby realizing privacy-preserving aggregate computing of the prediction value. At the same time, the entire process ensures the security of the key and the reliability of communication to ensure data privacy and calculation accuracy.

[0108] Specifically, step S14 also includes the following:

[0109] 1. Obtain historical resource demand values ​​and use the DBSCAN algorithm to automatically cluster them to obtain historical resource demand clusters, and statistically obtain the maximum boundary value of each cluster;

[0110] The DBSCAN algorithm is applied to the historical resource demand data set. According to the set parameters, the core points, boundary points and noise points are identified, and the data points are divided into different clusters. Each cluster represents a group of data points with similar historical resource demand characteristics. The boundary values ​​of each dimension are found by analyzing the coordinates of the points within the cluster.

[0111] 2. Calculate the similarity array of the target resource demand forecast value in all cluster center nodes in the historical resource demand clustering cluster through the Pearson similarity algorithm;

[0112] According to the calculation formula of the Pearson similarity algorithm, the similarity between the target resource demand forecast value and each cluster center node is calculated respectively. The value range of similarity is between -1 and 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, and the closer the value is to 0, the weaker the correlation is.

[0113] 3. Sort the similarity array in ascending order, and determine that the power dispatching strategy that is used most times in the cluster with the first array number is the first power dispatching strategy;

[0114] After sorting the obtained similarity array in ascending order, the first element in the array corresponds to the cluster with the lowest similarity to the target resource demand forecast value. In the cluster with the first array number, the number of times different power dispatch strategies are adopted is counted. The power dispatch strategy that is adopted the most times is determined as the first power dispatch strategy, which is used as the reference dispatch strategy when the current target resource demand forecast value is similar to the cluster situation, so as to reasonably arrange the dispatch of power resources.

[0115] 4. Construct objective functions with maximizing the utilization rate of power resource scheduling and minimizing resource waste as target parameters, and determine constraints, including power generation constraints, transmission capacity constraints, and power balance constraints;

[0116] set up is the power generation unit g at time t k Transmit to the power consumption area j of electricity, The power consumption area r j The predicted value of power resource demand in time period t, is the power generation unit g k The maximum power generation capacity of The resources actually used are Waste of resources Where (a―b) + represents max(a―b,0). That is, the objective function is as follows:

[0117]

[0118] Among them: α and β are weight coefficients, which are used to balance the relative importance of the two goals of maximizing resource utilization and minimizing resource waste. For example, if more emphasis is placed on resource utilization, the value of α can be relatively large; if more attention is paid to reducing resource waste, the value of β can be relatively large, and the goal is to maximize Z.

[0119] Power generation constraints:

[0120]

[0121] The power generation unit itself has minimum and maximum power generation limits, which ensures that the power generation unit operates within a safe and stable power generation range.

[0122] Transmission capacity constraints:

[0123]

[0124] The transmission lines from power generation units to power consumption areas are limited in maximum transmission capacity, which avoids overload of transmission lines and ensures stable transmission of electric energy.

[0125] Power balance constraints:

[0126]

[0127] The amount of electricity received by each electricity consumption area must match its predicted demand at that point in time to achieve an accurate balance between supply and demand, ensuring that the electricity supply in the electricity consumption area can meet its needs and avoid power shortages or surpluses.

[0128] Non-negativity constraints:

[0129]

[0130] The physical meaning of the transmitted electricity is limited to non-negative.

[0131] In the above formula: is the power generation unit g k The minimum power generation capacity, From the power generation unit g k The maximum transmission capacity of the transmission line to the power consumption area.

[0132] 5. Determine a first power dispatching strategy based on the objective function and the constraint conditions;

[0133] 6. Determine the target scheduling strategy according to the first scheduling strategy and the second scheduling strategy.

[0134] Under the premise of using the clustering algorithm to dispatch power resources, we can also consider the problem of maximizing the utilization rate of power resource dispatch and minimizing resource waste, that is, when the above dispatch is met, the second level of confirmation is performed. The following steps can be used as an option. When the clustering algorithm has met the dispatch requirements, there is no need to calculate the optimization parameters.

[0135] The two goals of maximizing resource utilization and minimizing resource waste are interrelated and complementary. By combining these two goals, while ensuring the efficient operation of the power system, more attention can be paid to the rational allocation and utilization of resources, avoiding waste caused by excessive power generation and unmet demand due to insufficient power generation, which will help achieve economical, efficient and sustainable power resource scheduling.

[0136] Embodiment 2

[0137] An embodiment of a power resource dispatching system based on cloud-edge collaboration in the present invention includes the following steps:

[0138] The data acquisition and preprocessing unit is used for each edge terminal to acquire the power equipment operation data according to the preset acquisition frequency and communication protocol, and preprocess the acquired data;

[0139] A feature vector extraction unit is used for extracting feature vectors based on preprocessed power equipment operation data at each edge terminal using a principal component analysis method;

[0140] A target resource demand prediction value determination unit is used for each edge end to input the feature vector into a trained support vector regression model to perform power resource demand prediction and obtain a target resource demand prediction value;

[0141] The power resource scheduling strategy determination unit is used for each edge terminal to encrypt the target resource demand prediction value and send it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy.

[0142] For the specific definition of the system, please refer to the definition of the method above, which will not be repeated here. Each module in the above system can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0143] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0144] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored. In addition, each functional unit in each embodiment of the present invention can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0145] If the integrated unit is implemented 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 this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0146] It can be understood that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. 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 invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A power resource scheduling method based on cloud-edge collaboration, characterized in that: include: Each edge terminal obtains the power equipment operation data according to the preset collection frequency and communication protocol, and pre-processes the obtained data; Each edge end extracts a feature vector based on the preprocessed power equipment operation data using a principal component analysis method; Each edge terminal inputs the feature vector into a trained support vector regression model to perform power resource demand prediction to obtain a target resource demand prediction value; Each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy.

2. The power resource scheduling method based on cloud-edge collaboration according to claim 1 is characterized in that: Each edge terminal inputs the feature vector into the trained support vector regression model to predict the power resource demand, and obtains the target resource demand prediction value, including: Obtain historical power equipment operation data and perform data preprocessing, and divide the preprocessed data into training sets and test sets; Select the support vector regression model architecture with radial basis function kernel, and input the training set into the support vector regression model for training; The optimization algorithm is used for optimization to obtain the trained support vector regression model.

3. The power resource scheduling method based on cloud-edge collaboration according to claim 2 is characterized in that: The support vector regression model architecture of selecting the radial basis function kernel, inputting the training set into the support vector regression model for training, comprises: Where: f(z i ) is the input sample z i The resource demand prediction value, l is the number of support vectors, α j and is the Lagrange multiplier, K(z i ,z j ) is the kernel function and b is the bias term.

4. The power resource scheduling method based on cloud-edge collaboration according to claim 1 is characterized in that: Each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy, including: Each edge terminal and cloud terminal determine a multi-party secure computing protocol of secret sharing and homomorphic encryption; Each edge terminal splits its own power resource demand forecast value into a plurality of shares based on the multi-party secure computing protocol, and sends the plurality of shares to other edge terminals and the cloud; After receiving the shares from other edge terminals, each edge terminal uses the public key to homomorphically encrypt the received shares and then sends them to the cloud; The cloud performs aggregate calculations on all received encrypted shares based on homomorphic encryption properties and a multi-party secure computing protocol, and obtains the sum of the predicted values ​​of power resource demand at each edge terminal after decryption.

5. The power resource scheduling method based on cloud-edge collaboration according to claim 4 is characterized in that: Each edge terminal splits its own power resource demand forecast value into a plurality of shares based on the multi-party secure computing protocol, and sends the plurality of shares to other edge terminals and the cloud, including: Where: v i is the power resource demand forecast value of the ith (i=1,2,…,n) edge terminal, s ij The private share generated for the i-th edge node, where (j=1, 2, …, n) and M is the modulus.

6. The power resource scheduling method based on cloud-edge collaboration according to claim 5 is characterized in that: The cloud performs aggregate calculation on all received encrypted shares based on homomorphic encryption properties and multi-party secure computing protocols, and obtains the sum of the power resource demand forecast values ​​of each edge terminal after decryption, including: Where: Enc(s ij ) is the secret share from the jth edge end received and encrypted by the ith edge end, and Dec is the decryption function.

7. The power resource scheduling method based on cloud-edge collaboration according to claim 4 is characterized in that: Each edge terminal encrypts the target resource demand prediction value and sends it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain a corresponding power resource scheduling strategy, including: Obtain historical resource demand values ​​and use the DBSCAN algorithm to automatically cluster them to obtain historical resource demand clusters, and statistically obtain the maximum boundary value of each cluster; Calculate the similarity array of the target resource demand forecast value in all cluster center nodes in the historical resource demand clustering cluster by using the Pearson similarity algorithm; Sorting the similarity array in ascending order, and determining the power dispatching strategy that is used most times in the cluster with the first array sequence number as the first power dispatching strategy; The objective function is constructed with maximizing the utilization rate of power resource scheduling and minimizing resource waste as target parameters, and constraint conditions are determined, wherein the constraint conditions include power generation constraint, transmission capacity constraint and power balance constraint; Determine a first power dispatch strategy based on the objective function and the constraint conditions; A target scheduling strategy is determined according to the first scheduling strategy and the second scheduling strategy.

8. The power resource scheduling method based on cloud-edge collaboration according to claim 7 is characterized in that: The objective function is constructed by taking the maximization of the utilization rate of power resource scheduling and the minimization of resource waste as the objective parameters, including: Among them: U is the effectively utilized resources, S is the total resource supply, W is the actually effectively utilized resources, is the power generation unit g k is the maximum power generation, α and β are weight coefficients.

9. The power resource scheduling method based on cloud-edge collaboration according to claim 7 is characterized in that: The expressions of the constraints are as follows: Power generation constraints: The power generation unit itself has minimum and maximum power generation limits, which ensures that the power generation unit operates within a safe and stable power generation range; Transmission capacity constraints: The transmission lines from the power generation unit to the power consumption area have a maximum transmission capacity limit, which avoids overloading of the transmission lines and ensures the stable transmission of electric energy; Power balance constraints: The amount of electricity received by each power consumption area must match its predicted demand at that point in time to achieve an accurate balance between supply and demand, ensuring that the power supply in the power consumption area can meet its needs and avoid power shortages or surpluses; Non-negativity constraints: The physical meaning of the transmitted electricity is limited to non-negative.

10. A power resource dispatching system based on cloud-edge collaboration, characterized in that: include: The data acquisition and preprocessing unit is used for each edge terminal to acquire the power equipment operation data according to the preset acquisition frequency and communication protocol, and preprocess the acquired data; A feature vector extraction unit is used for each edge end to extract a feature vector based on the preprocessed power equipment operation data by using a principal component analysis method; A target resource demand prediction value determination unit is used for each edge end to input the feature vector into a trained support vector regression model to perform power resource demand prediction and obtain a target resource demand prediction value; The power resource scheduling strategy determination unit is used for each edge terminal to encrypt the target resource demand prediction value and send it to the cloud, so that the cloud analyzes the decrypted target resource demand prediction value to obtain the corresponding power resource scheduling strategy.

Images