A cloud computing-based big data privacy computing method and system, electronic equipment and storage medium

Abstract

The application provides a cloud computing-based big data privacy calculation method and system, electronic equipment and storage medium, relates to the technical field of big data privacy calculation, and is based on a lightweight federated learning framework. An edge computing gateway collects intelligent home real-time energy consumption big data, dynamically masks device identification, generates encrypted data streams of type labels and energy consumption data, processes compressed encrypted data streams through Delta encoding, obtains device processing parameters through local processing, and forms global energy scheduling reference information through homomorphic encryption aggregation parameters on a cloud computing platform to optimize strategies through reinforcement learning, combines an alarm mechanism to link devices, adjusts operation through TLS / SSL encrypted transmission instructions, protects data privacy, cooperates with edge and cloud computing through a lightweight federated learning framework, combines encryption and optimization technologies, realizes intelligent home energy consumption data privacy protection and efficient scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data privacy computing, and particularly relates to a big data privacy computing method and system based on cloud computing, an electronic device and a storage medium. BACKGROUND

[0002] In the scene of cloud computing processing big data, such as the field of smart home, financial risk control, etc., it is necessary to aggregate and analyze massive user data to mine value, while the data privacy must be strictly protected to prevent the leakage of user sensitive information. This requires that the computing method can not only realize efficient processing and sharing of data, but also block unauthorized access through technical means to meet the dual needs of data utilization and privacy protection.

[0003] At present, the common solution to this demand is a cloud computing data processing scheme based on centralized encryption. The user data is encrypted and uploaded to the cloud server, and the cloud server decrypts and completes data computation, while an access control strategy is adopted to limit data operation permissions, so as to balance data processing efficiency and privacy security.

[0004] However, the above-mentioned scheme has obvious defects. Centralized decryption makes the cloud server directly contact the plaintext data, and once the server is attacked or the internal personnel violates the operation, it is easy to cause large-scale data leakage. The centralized transmission and decryption process of encrypted data will increase the network load and computing delay, which is difficult to adapt to real-time scenes with high real-time requirements, and the centralized management of the key also has the risk of single point failure. SUMMARY

[0005] The present application aims to provide a big data privacy computing method and system based on cloud computing, an electronic device and a storage medium, to solve the problem of low security of big data privacy computing in the prior art.

[0006] To solve the above technical problems, in a first aspect, the present application provides a big data privacy computing method based on cloud computing, comprising:

[0007] Based on the edge computing gateway deployed on the side of the smart home device in the lightweight federated learning framework, real-time energy consumption big data of each smart home device is collected, and the device identifier in the real-time energy consumption big data is dynamically masked to generate encrypted data stream carrying device type label and energy consumption data, so as to protect data privacy;

[0008] The encrypted data stream is processed by using a Delta encoding compression algorithm to obtain a compressed encrypted data stream, and the compressed encrypted data stream is locally processed to obtain the processing parameters of the smart home devices;

[0009] Based on the cloud computing platform in the lightweight federated learning framework, the processing parameters are aggregated by homomorphic encryption technology to form global energy scheduling reference information;

[0010] Based on the global energy scheduling reference information, a reinforcement learning method is introduced to optimize the energy scheduling strategy, a local device linkage is realized in combination with an alarm mechanism, and a control instruction is transmitted by a TLS / SSL encryption method to adjust the operation strategy of each smart home device, so that the real-time energy consumption big data is always in a privacy protection state during processing and transmission.

[0011] Optionally, based on the global energy scheduling reference information, a reinforcement learning method is introduced to optimize the energy scheduling strategy, a local device linkage is realized in combination with an alarm mechanism, and a control instruction is transmitted by a TLS / SSL encryption method to adjust the operation strategy of each smart home device, including the following steps:

[0012] Based on the lightweight federated learning framework, the processing results of each device type in the global energy scheduling reference information are input to generate an initial energy scheduling strategy;

[0013] The expected energy consumption values of each smart home device in the initial energy scheduling strategy are compared with a preset normal energy consumption range, and when the expected energy consumption value of any of the smart home devices exceeds the preset normal energy consumption range, an alarm mechanism is triggered and the device type of the smart home device whose expected energy consumption value exceeds the preset normal energy consumption range is marked;

[0014] Based on the edge computing gateway linkage marked device type, all local smart home devices under the device type are linked and the current real-time running parameters of all local smart home devices are collected;

[0015] In combination with the marked device type and the corresponding real-time running parameters, the initial energy scheduling strategy is adjusted by the cloud computing platform to form an optimized energy scheduling strategy, and the optimized energy scheduling strategy includes operation adjustment instructions of the smart home devices;

[0016] The operation adjustment instructions are encrypted by a TLS / SSL encryption method to generate encrypted control instructions, and the encrypted control instructions are sent to the edge computing gateway through a preset transmission channel, so that the edge computing gateway decrypts the encrypted control instructions after receiving them and generates decrypted control instructions, and adjusts the operation strategy of the smart home devices according to the decrypted control instructions.

[0017] Optionally, the running adjustment instruction is encrypted by the TLS / SSL encryption mode to generate an encrypted control instruction, and the encrypted control instruction is sent to the edge computing gateway through a preset transmission channel, so that the edge computing gateway decrypts the encrypted control instruction and generates a decrypted control instruction after receiving the encrypted control instruction.

[0018] The running adjustment instruction in the optimized energy scheduling strategy is added with a device unique identifier and a timestamp, and is encrypted by the TLS / SSL encryption mode based on a lightweight federated learning framework to generate an encrypted control instruction containing an encrypted running adjustment instruction, an encrypted device unique identifier segment, and an encrypted timestamp segment.

[0019] The encrypted control instruction is grouped according to a device type label, and the encrypted control instructions in the same group are sent to the edge computing gateway in sequence through a preset transmission channel established based on a lightweight federated learning framework communication protocol, so that the edge computing gateway splits the encrypted control instruction into single encrypted control instructions according to the device type label after receiving the encrypted control instruction.

[0020] The encrypted device unique identifier segment and the encrypted timestamp segment in the encrypted control instruction are decrypted, and the matching of the device unique identifier with the local device and the validity of the timestamp are verified, and the running adjustment instruction in the encrypted control instruction is decrypted to generate a decrypted control instruction in the case that the device unique identifier matches and the timestamp is valid.

[0021] Optionally, the encrypted data stream is processed by using a Delta encoding compression algorithm to obtain a compressed encrypted data stream, and the processing parameters of the smart home devices are obtained by locally processing the compressed encrypted data stream, including the following steps:

[0022] The real-time energy consumption big data corresponding to the device type label in the encrypted data stream is extracted, and the difference values between consecutive numerical values in the real-time energy consumption big data are calculated.

[0023] Based on the difference values, a Delta encoding compression algorithm is called to form a compressed encrypted data stream, and the compressed encrypted data stream retains the association between the device type label and the corresponding real-time energy consumption big data.

[0024] Based on a preset edge computing gateway rule, the compressed encrypted data stream is divided into a plurality of sub-data streams.

[0025] The local processing rule of each smart home device is called, the difference values in the sub-data stream are converted into intermediate parameters conforming to a preset format based on the local processing rule, and the intermediate parameters are combined and encapsulated to form the processing parameters of the smart home devices.

[0026] Optionally, based on the difference value, a Delta encoding compression algorithm is called to form a compressed encrypted data stream, including the following steps:

[0027] For each difference value, a corresponding numerical position identifier is added to the difference value, which is used to represent the order of the difference value in the continuous numerical values in the real-time energy consumption big data;

[0028] The difference value added with the numerical position identifier is paired with a preset reference value, and for each difference value, the corresponding preset reference value is labeled as an associated marker to form a marked difference value;

[0029] Based on the lightweight federated learning framework, the Delta encoding compression algorithm is called, the data order is maintained through the numerical position identifier, and the corresponding continuous numerical values in the real-time energy consumption big data are replaced one by one with the marked difference value to form a replaced data set, and the original correspondence between the device type label and the real-time energy consumption big data is preserved during replacement;

[0030] The replaced data set is classified according to the device type label, and the marked difference value and the corresponding numerical position identifier under the same device type label are integrated into a data block, wherein each data block is associated with a corresponding device type label;

[0031] All the data blocks are combined according to the preset sorting rule of the device type label to form a compressed encrypted data stream, and the compressed encrypted data stream maintains the association with the real-time energy consumption big data through the numerical position identifier, the associated marker and the device type label in the data block.

[0032] Optionally, the real-time energy consumption big data is dynamically masked to generate an encrypted data stream carrying a device type label and energy consumption data, including the following steps:

[0033] Based on the dynamic masking processing rule called by the lightweight federated learning framework, the serial number, factory number field and device type field used to distinguish different smart home devices in the device identifier are replaced with a random character sequence that changes over time as a device type label;

[0034] The device type label and the corresponding energy consumption data in the real-time energy consumption big data are combined to form an original data unit, and the original data unit is rearranged based on the lightweight federated learning framework to form a new data unit;

[0035] According to a preset encryption conversion processing rule, the new data unit is encrypted to generate an encrypted data stream.

[0036] Optionally, the cloud computing platform in the lightweight federated learning framework aggregates the processing parameters by homomorphic encryption technology to form the global energy scheduling reference information, including the following steps:

[0037] The cloud computing platform based on the lightweight federated learning framework receives the processing parameters sent by each smart home device, and the processing parameters are encapsulation results containing device type labels and corresponding intermediate parameters;

[0038] The homomorphic encryption technology is called to encrypt and convert the intermediate parameters to form encrypted intermediate parameters, and a corresponding device type label is added as an additional identifier for each encrypted intermediate parameter;

[0039] The encrypted intermediate parameters corresponding to the same device type label are divided into the same data group, and the encrypted intermediate parameters in each data group are merged to obtain type aggregation parameters corresponding to each data group;

[0040] All the type aggregation parameters are integrated to form the global energy scheduling reference information.

[0041] In a second aspect, the present application provides a cloud computing-based big data privacy computing system, comprising:

[0042] The acquisition module is configured to acquire real-time energy consumption big data of each smart home device based on an edge computing gateway deployed on a smart home device side in a lightweight federated learning framework, and perform dynamic mask processing on device identifiers in the real-time energy consumption big data to generate encrypted data streams carrying device type labels and energy consumption data to protect data privacy;

[0043] The compression module is configured to process the encrypted data streams by using a Delta encoding compression algorithm to obtain compressed encrypted data streams, and perform local processing on the compressed encrypted data streams to obtain processing parameters of the smart home devices;

[0044] The aggregation module is configured to aggregate the processing parameters by homomorphic encryption technology based on a cloud computing platform in the lightweight federated learning framework to form a global energy scheduling reference information;

[0045] The encryption module is configured to introduce a reinforcement learning method to optimize an energy scheduling strategy based on the global energy scheduling reference information, combine an alarm mechanism to realize local device linkage, and transmit control instructions by a TLS / SSL encryption method to adjust the operation strategies of the smart home devices, so that the real-time energy consumption big data is always in a privacy protection state during processing and transmission.

[0046] In a third aspect, the present application provides an electronic device, comprising:

[0047] a memory for storing the computer program;

[0048] a processor for implementing the steps of the method for big data privacy computing based on cloud computing when executing the computer program.

[0049] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program, when executed by a processor, can implement the steps of the method for big data privacy computing based on cloud computing according to the first aspect.

[0050] The method for big data privacy computing based on cloud computing provided by the present application can collect real-time energy consumption big data of each smart home device through an edge computing gateway deployed on a smart home device side in a lightweight federated learning framework, and perform dynamic mask processing on device identifiers in the real-time energy consumption big data to generate encrypted data streams carrying device type labels and energy consumption data, so as to protect data privacy. A Delta encoding compression algorithm is used to process the encrypted data streams to obtain compressed encrypted data streams, and the compressed encrypted data streams are processed locally to obtain processing parameters of the smart home devices. A cloud computing platform in the lightweight federated learning framework is used to aggregate the processing parameters through homomorphic encryption technology to form global energy scheduling reference information. Based on the global energy scheduling reference information, a reinforcement learning method is introduced to optimize an energy scheduling strategy, a local device linkage is realized in combination with an alarm mechanism, and a control instruction is transmitted through a TLS / SSL encryption method to adjust the operation strategy of the smart home devices, so that the real-time energy consumption big data is always in a privacy protection state during processing and transmission.

[0051] The technical solution of the present application has the following beneficial effects:

[0052] The application collects real-time energy consumption big data through an edge computing gateway, and performs dynamic mask processing on device identification to generate encrypted data stream with device type label and energy consumption data, which can avoid direct exposure of device identification, protect privacy from the data source, and lay a secure foundation for subsequent data transmission and processing through encryption processing; the encrypted data stream is processed by a Delta encoding compression algorithm to obtain compressed encrypted data stream, and processing parameters of each device are obtained through local processing, which can reduce data transmission to improve efficiency, and local processing can also reduce the risk of privacy leakage in cloud centralized processing; the cloud computing platform aggregates processing parameters to form global energy scheduling reference information through homomorphic encryption technology, which can realize data aggregation without leaking plaintext of processing parameters, thus protecting data privacy and providing basis for global energy scheduling; based on the global energy scheduling reference information, a reinforcement learning method is introduced to optimize the energy scheduling strategy, combined with an alarm mechanism to link local devices, and control instructions are transmitted through TLS / SSL encryption to adjust the device operation strategy, which can realize dynamic optimization of energy scheduling, timely response to abnormal situations, and ensure privacy and security of data processing and transmission through full encryption.

[0053] Further, the application calls a reinforcement learning method based on a lightweight federated learning framework, inputs the processing results of each device type in the global energy scheduling reference information to generate an initial energy scheduling strategy; compares the expected energy consumption value of each device with the preset normal range, and triggers an alarm and marks the device type when the value exceeds the range; the edge computing gateway links the local devices of the marked type to collect their real-time operation parameters; the cloud computing platform adjusts the initial strategy in combination with the marked type and corresponding parameters to form an optimized strategy containing operation adjustment instructions; the instructions are encrypted through TLS / SSL to generate encrypted control instructions, which are sent to the edge computing gateway through a preset channel, and the operation strategy of each device is adjusted after decryption.

[0054] The application generates an initial strategy through reinforcement learning and triggers an alarm in combination with energy consumption comparison, which can timely discover abnormal energy consumption of devices; the local devices are linked to collect parameters and adjust the strategy, which can optimize energy scheduling; the control instructions are transmitted through TLS / SSL encryption, which can ensure the safety of instruction transmission, finally realize precise adjustment of device operation strategy, and improve the rationality of energy scheduling and the safety of device operation.

[0055] These aspects or other aspects of the application will be more apparent in the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0057] Figure 1 A flowchart of a big data privacy computing method based on cloud computing provided for an embodiment of the present application is shown in FIG. 1.

[0058] Figure 2 A scenario diagram of a big data privacy computing method based on cloud computing provided for an embodiment of the present application is shown in FIG. 2.

[0059] Figure 3 A specific structure diagram of a big data privacy computing system based on cloud computing provided for an embodiment of the present application is shown in FIG. 3.

[0060] Figure 4 A structure diagram of a computing device provided for an embodiment of the present application is shown in FIG. 4. DETAILED DESCRIPTION

[0061] In the scenario of cloud computing processing big data, the existing processing scheme based on centralized encryption has obvious deficiencies. This kind of scheme uploads encrypted data to the cloud for unified decryption processing, so that the cloud directly contacts the plaintext data. Once attacked or improperly operated internally, it is easy to cause large-scale data leakage. At the same time, the centralized transmission and decryption process of encrypted data will increase the network burden and processing delay, which is difficult to adapt to the scenario with high real-time demand, and the centralized management mode of the key also has the security risk of single point failure.

[0062] To solve the above problems, the present application provides a big data privacy computing method based on cloud computing. This method relies on a lightweight federated learning framework, collects real-time energy consumption data on the device side through an edge computing gateway, dynamically hides the device identifier and generates an encrypted data stream to protect privacy from the source. Then, after compression and local processing, the device parameters are obtained, and the cloud computing platform aggregates the parameters in an encrypted state to form global reference information. Finally, based on this information, the energy dispatching strategy is optimized, and the device operation is adjusted through encrypted transmission instructions. This scheme cooperates with the edge and the cloud to avoid the exposure of plaintext data in the cloud, reduce the transmission delay, and decentralize the key management to reduce the security risk. While protecting data privacy, it improves processing efficiency and real-time performance.

[0063] In order for those skilled in the art to better understand the present application, the present application will be further described in detail below in conjunction with the drawings and specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0064] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in one or more embodiments of the present specification are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards, and provide corresponding operation portal for user to choose authorization or refusal.

[0065] The core of the present application is to provide a cloud computing-based big data privacy computing method, and a flowchart of one specific embodiment thereof is shown in Figure 1 The method comprises the following steps:

[0066] S101, based on the edge computing gateway deployed on the smart home device side in the lightweight federated learning framework, collecting real-time energy consumption big data of each smart home device according to a preset time interval, wherein the energy consumption big data contains device identification and energy consumption data, dynamically masking the device identification in the real-time energy consumption big data, and generating encrypted data stream carrying device type label and energy consumption data to protect data privacy.

[0067] Optionally, the step S101 of dynamically masking the device identification in the real-time energy consumption big data to generate the encrypted data stream carrying the device type label and the energy consumption data comprises the following steps:

[0068] Step 1011, based on the dynamic masking processing rule called by the lightweight federated learning framework, replacing the serial number, factory number field and field representing device type in the device identification for distinguishing different smart home devices with a random character sequence varying with time as a device type label.

[0069] Step 1012, combining the device type label and the energy consumption data in the real-time energy consumption big data corresponding to the device type label to form an original data unit, and rearranging the original data unit based on the lightweight federated learning framework to form a new data unit.

[0070] Step 1013, according to a preset encryption conversion processing rule, encrypting the new data unit to generate an encrypted data stream.

[0071] In the above scheme, the lightweight federated learning framework refers to a system that enables edge devices to collaboratively process data with cloud computing platforms and reduces data transmission volume; the edge computing gateway refers to a device installed on the side of the smart home for collecting and preliminarily processing data of various smart home devices; the real-time energy consumption big data refers to the power consumption data generated by various smart home devices during operation and changing over time; the device identifier is information used to distinguish different smart home devices, including a serial number, a factory number, and other fields that can uniquely identify the device, and a field indicating the device type; the dynamic mask processing refers to a replacement processing method that changes over time, replacing specific content in the original identifier with random characters; the device type label is a sequence of random characters used to represent the device type after dynamic mask processing; the energy consumption data refers to real-time power consumption information of the device; and the encrypted data stream is a transmittable data stream formed after the data is encrypted.

[0072] In the example of the present application, first, the device identifier is processed based on the lightweight federated learning framework by step 1011, wherein the device identifier contains a serial number (denoted as S) for distinguishing different devices, a factory number (denoted as F), and a field (denoted as T) representing the device type. During processing, the three fields are collectively replaced with a sequence of random characters that changes over time as a device type label (denoted as L(t)), and the generation method satisfies the formula , wherein is a random character generation function, t is the current timestamp, and the formula ensures that L(t) generated at different times is different.

[0073] Next, L(t) is combined with the energy consumption data (denoted as E) in the corresponding real-time energy consumption big data by step 1012 to form an original data unit , and then D is rearranged based on the lightweight federated learning framework. The rearrangement rule satisfies the formula , wherein represents the new position of the element at position i in the original data unit after bit rearrangement in the new data unit, is the position of the element in the original data unit, K is a preset offset, mod indicates a remainder operation, and N is the total number of elements in the original data unit. For example, when K = 1 and N = 2, i = 0, , it indicates that the element at original position 0 will be placed at new position 1 after rearrangement, and when i = 1, , it indicates that the element at original position 1 will be placed at new position 0 after rearrangement. After calculating the new position according to the formula, the new data unit is obtained.

[0074] Finally, the new data unit D' is encrypted according to a preset encryption conversion processing rule through step 1013, and the encryption process satisfies the formula , where Enc() is an encryption function, and an expression of the encryption function is not specifically limited in the embodiments of the present application, is a preset encryption key, and C is an encrypted data stream, that is, an encryption result. The encrypted data stream is used for processing of a Delta encoding compression algorithm in a subsequent step.

[0075] In actual application, an edge computing gateway in a smart home scene collects real-time energy consumption data E of a device B (type: "refrigerator"). The average power of the device B in a monitoring period is 190 W, and the running time is 20 hours. According to an energy consumption calculation formula E=P*t, E=0.19 kWh*20 h=3.8 kWh, where P is the average power, and t is the running time. The serial number S of the device B is 2002, the factory number F is 2024, and the type field T is "refrigerator". At t=11:00, the device type label L(11:00) is generated according to the formula , where a random function generates the device type label L(11:00)="D5E8F" by performing character mapping conversion on the input serial number 2002, factory number 2024, type field "refrigerator", and timestamp 11:00. According to the real-time energy consumption data E=3.8 kWh, the original data unit D=["D5E8F 3.8] is formed, and the new data unit D' is calculated according to the rearrangement formula , that is, D'=["D5E8F 3.8]. On the basis of Kenc="X9Y2Z", the encrypted data stream C="encrypted [3.8 D5E8F]" is generated through the encryption formula C=Enc(D' Kenc). The encrypted data stream is then transmitted to a local processing module for Delta encoding compression processing.

[0076] The S101 overall scheme generates a device type label that changes over time by dynamically replacing the serial number, factory number and type field in the device identifier, obtains the effect of hiding the specific identity information of the device and retaining only the type characteristics, and at the same time avoids the risk of privacy leakage caused by a fixed identifier; the device type label is combined with the energy consumption data and rearranged bit by bit to obtain the effect of disrupting the original data structure and increasing the difficulty of unauthorized analysis, laying a secure foundation for subsequent encryption processing; the new data unit after rearrangement is encrypted to generate an encrypted data stream, obtaining the effect of ensuring that the data is difficult to be illegally acquired and interpreted during transmission and processing, and at the same time providing a privacy protection prerequisite for subsequent compression and local processing, and overall realizing privacy security protection from the data acquisition source to the preliminary processing stage.

[0077] S102, using a Delta encoding compression algorithm to process the encrypted data stream to obtain a compressed encrypted data stream, and performing local processing on the compressed encrypted data stream to obtain the processing parameters of each smart home device.

[0078] Optionally, in step S102, the encrypted data stream is processed using a Delta encoding compression algorithm to obtain a compressed encrypted data stream, and the compressed encrypted data stream is locally processed to obtain the processing parameters of each smart home device, including the following steps:

[0079] Step 1021, extracting real-time energy consumption big data corresponding to the device type label in the encrypted data stream, and calculating the difference value between consecutive values in the real-time energy consumption big data, wherein consecutive values refer to energy consumption data collected in time sequence in the real-time energy consumption big data. These data are sequences formed after recording the energy consumption of the same device at different time points (such as every minute, every hour), and the values before and after have continuity in time, and the time nodes are multiple, for example, 3. For example, the real-time energy consumption big data of a refrigerator is "3.8kWh (10:00), 4.0kWh (10:05), 3.9kWh (10:10)", wherein "3.8, 4.0, 3.9" are consecutive values arranged in time sequence, and have a time sequence relationship with each other.

[0080] Step 1022, based on the difference value, calling a Delta encoding compression algorithm to form a compressed encrypted data stream, the compressed encrypted data stream retaining the association between the device type label and the real-time energy consumption big data.

[0081] The step 1022 can specifically include the following processes: for each difference value, adding a corresponding numerical position identifier to the difference value, the numerical position identifier being used to represent the order of the difference value in the continuous numerical values in the real-time energy consumption big data; pairing the difference value added with the numerical position identifier with a preset reference value, labeling a corresponding preset reference value as an association mark for each difference value to form a marked difference value; calling a Delta encoding compression algorithm based on a lightweight federated learning framework, keeping data order through the numerical position identifier, replacing the corresponding continuous numerical values in the real-time energy consumption big data with the marked difference value one by one to form a replaced data set, and retaining the original correspondence between the device type label and the real-time energy consumption big data during replacement; classifying the replaced data set according to the device type label, integrating the marked difference value and the corresponding numerical position identifier under the same device type label into a data block, wherein each data block is associated with a corresponding device type label; combining all the data blocks according to a preset sorting rule of the device type label to form a compressed and encrypted data stream, and the compressed and encrypted data stream maintains the association with the real-time energy consumption big data through the numerical position identifier, the association mark and the device type label in the data block.

[0082] Step 1023, based on a preset edge computing gateway rule, dividing the compressed and encrypted data stream into a plurality of sub-data streams.

[0083] Step 1024, calling a local processing rule of each smart home device, converting the difference value in the sub-data stream into an intermediate parameter conforming to a preset format based on the local processing rule, combining and packaging the intermediate parameter to form a processing parameter of each smart home device.

[0084] In the above scheme, encrypted data stream refers to a data stream containing device type tags and energy consumption data that has been encrypted; Delta encoding compression algorithm is a processing method that reduces the amount of data by recording differences between data; compressed encrypted data stream refers to a data stream whose data volume is reduced after Delta encoding processing while still maintaining an encrypted state; local processing refers to the data processing operations performed on the edge computing gateway side; real-time energy consumption big data refers to continuous data related to the real-time power consumption of devices collected by the edge computing gateway; numerical position identifier is a marker used to mark the order of the difference value in the original continuous values; preset baseline value is a pre-set reference value used to associate with the difference value; data block is a collection of marked difference values ​​and numerical position identifiers under the same device type tag; preset sorting rule is a pre-set order rule used to combine data blocks; edge computing gateway rule is a pre-set rule used to divide compressed encrypted data stream; local processing rule is a rule set by each smart home device itself used to convert sub-data streams; preset format is a pre-set format standard used to standardize intermediate parameters.

[0085] In this application example, firstly, the real-time energy consumption big data corresponding to the device type tag in the encrypted data stream is extracted in step 1021 and denoted as E, where E=[E1, E2, ..., Ek], Ek represents the energy consumption value at the k-th time point, k=1, 2, ..., n, and n is the total number of data. The difference between consecutive values ​​is calculated (denoted as...). The calculation formula is: ,in Represents the k-th difference value ( ), For the k-th energy consumption value, This represents the (k-1)th energy consumption value. For example, the actual energy consumption data for device B (refrigerator) is... The unit is ,but , , To obtain the set of difference values .

[0086] Next, the above difference values ​​are processed through step 1022, first by setting each... Add a numerical position identifier (denoted as P, (This represents the k-th position in the original data), such as correspond , correspond , correspond Then, the difference value with location identifier is paired with the preset reference value (denoted as B, where B=3.5), and the association mark is added (denoted as M). ), forming a marked difference value (denoted as D, ), such as , , ; calling the Delta encoding compression algorithm, according to maintaining the order, replacing the values in the original E from the second with , forming a replaced data set (denoted as S), that is , while keeping the association between the device type label "refrigerator" and S; classifying according to the device type label, integrating the same type of and into a data block (denoted as ), such as "refrigerator" corresponding to , combining all according to the preset sorting rule (such as the first letter order of the device type) to form a compressed encrypted data stream (denoted as C).

[0087] Then, through step 1023, based on the edge computing gateway rule (such as dividing according to the number of contained in , every 2 is a subunit), C is divided into sub-data streams (denoted as ), for example contains 3 , then divided into and .

[0088] Finally, through step 1024, call the device local processing rule (such as the rule of the refrigerator is , where is an intermediate parameter), convert in to : in =0.2 corresponds to , corresponds to ; in =0.2 corresponds to . Combine and package according to the preset format (such as ) to form a processing parameter (denoted as ), that is refrigerator, , used for subsequent cloud computing platform aggregation processing.

[0089] In practical applications, in the smart home scene, the edge computing gateway processes the encrypted data stream of device B (refrigerator), and the real-time energy consumption big data of device B is , unit , the difference value is calculated according to the formula , the difference value is calculated according to the formula , the difference value is calculated according to the formula , , the difference value is calculated according to the formula ; then, the original data is replaced by calling the Delta encoding compression algorithm to form S , and integrated into the "refrigerator" data block , combined to form compressed encrypted data C; according to the edge computing gateway rule (every 2 is a subunit), C is divided into and ; according to the local processing rule of the refrigerator , , , combined to form the processing parameter of the refrigerator , which will be transmitted to the cloud computing platform for aggregation of global energy scheduling reference information.

[0090] The above S102 overall scheme extracts the difference value of real-time energy consumption big data and performs Delta encoding compression, reducing the data transmission amount while retaining the data correlation; the compressed encrypted data stream is divided into sub-data streams and processed locally, avoiding the privacy leakage risk of centralized processing in the cloud, and the local processing can quickly generate processing parameters in the correct format; the whole process improves the data processing and transmission efficiency while maintaining data privacy, providing a reliable foundation for subsequent cloud aggregation and energy scheduling strategy optimization.

[0091] S103, based on the cloud computing platform in the lightweight federated learning framework, the processing parameters are aggregated by homomorphic encryption technology to form global energy scheduling reference information.

[0092] Optionally, step S103, based on the cloud computing platform in the lightweight federated learning framework, the processing parameters are aggregated by homomorphic encryption technology to form global energy scheduling reference information, comprising the following steps:

[0093] Step 1031, based on the lightweight federated learning framework, the cloud computing platform receives the processing parameters sent by each smart home device, and the processing parameters are the encapsulation results containing device type labels and corresponding intermediate parameters.

[0094] Step 1032, call homomorphic encryption technology to encrypt and convert the intermediate parameters to form encrypted intermediate parameters, and add the corresponding device type label as an additional identifier for each encrypted intermediate parameter.

[0095] ​Step 1033: Divide the encrypted intermediate parameters corresponding to the same device type label into the same data group, and perform a merging operation on the encrypted intermediate parameters in each data group to obtain the type aggregation parameter corresponding to each data group.

[0096] Step 1034: Integrate all the above-mentioned aggregation parameters to form global energy scheduling reference information.

[0097] In the above scheme, processing parameters refer to the encapsulated data generated by each smart home device after local processing, which includes device type tags and corresponding intermediate parameters; homomorphic encryption technology is an encryption technology that can directly perform operations on encrypted data, and the decrypted result is consistent with the plaintext operation result; encrypted intermediate parameters are data formed after homomorphic encryption of intermediate parameters; device type tags are labels used to identify device types; additional identifiers are device type tags added to encrypted intermediate parameters; data groups are sets of encrypted intermediate parameters corresponding to the same device type tag; type aggregation parameters are the results obtained after merging encrypted intermediate parameters in the same data group; global energy scheduling reference information is reference data for global energy scheduling formed by integrating all type aggregation parameters.

[0098] In this application example, firstly, through step 1031 based on a lightweight federated learning framework, the cloud computing platform is controlled to receive processing parameters (denoted as...) sent by each smart home device. The format is ), where T is the device type label, Let m be the j-th intermediate parameter, and m be the number of intermediate parameters. For example, the receiving device B (refrigerator) has... refrigerator, And equipment C (air conditioner) air conditioner, Where T = "refrigerator", =5.5, =2.5, =5.5, m=3, T=“Air Conditioner” =3.0, =1.8, m=2, these processing parameters will be used for subsequent encryption processing.

[0099] Next, homomorphic encryption is invoked in step 1032 to... In Perform encryption conversion; the encryption formula is: ,in Here, K is a homomorphic encryption function (used to encrypt plaintext parameters), and K is a preset encryption key. For the j-th encrypted intermediate parameter of the t-th type of device, for example, using K to pair of =5.5 encrypted , =2.5 encrypted , =5.5 encrypted ; to of =3.0 encrypted , =1.8 encrypted . Then, for each , the corresponding T is added as an additional identifier (e.g. "Refrigerator" is added), and these labeled encrypted intermediate parameters will be used for grouping processing.

[0100] Then, by step 1033, the corresponding to the same T are divided into the same data group (denoted as , where t is the device type). For example, the , , corresponding to "Refrigerator" form , and the , corresponding to "Air Conditioner" form . The in each are merged and the operation formula is , where is the type aggregation parameter of the t-th type of device, () is a homomorphic decryption function (used to decrypt the encrypted operation result), is the number of in , i.e. =m, for example in , then = ; in , these will be used to generate global reference information.

[0101] Finally, by step 1034, all are integrated to form the global energy dispatching reference information (denoted as R, in the format of ). For example, integrating =4.5, =2.4, we get Refrigerator, 4.5 Air Conditioner, 2.4 , which will be used for subsequent optimization of energy dispatching strategy.

[0102] The S103 overall scheme receives processing parameters through the cloud computing platform and encrypts intermediate parameters using homomorphic encryption technology, ensuring that data is always in an encrypted state during aggregation, protecting data privacy; grouping and combining encrypted parameters of the same type of device, effectively aggregating data of the same type of device; integrating all types of aggregated parameters to form global energy scheduling reference information, providing comprehensive and reliable basis for subsequent optimization of energy scheduling strategies, while avoiding exposure of plaintext data, balancing the needs of data utilization and privacy protection.

[0103] S104, based on the global energy scheduling reference information, introducing a reinforcement learning method to optimize the energy scheduling strategy, combining an alarm mechanism to realize local device linkage, and transmitting control instructions through a TLS / SSL encryption method to adjust the operation strategy of each smart home device, so that the real-time energy consumption big data is always in a privacy protection state during processing and transmission.

[0104] The operation strategy and the energy scheduling strategy both serve the energy management of the smart home device, the energy scheduling strategy is the guiding basis of the operation strategy, and the operation strategy is the specific form of the energy scheduling strategy. That is, the energy scheduling strategy directly affects and adjusts the operation strategy of each device by generating operation adjustment instructions, and the two work together to achieve global energy optimization and efficient operation of devices.

[0105] Optionally, in step S104, based on the global energy scheduling reference information, a reinforcement learning method is introduced to optimize the energy scheduling strategy, and an alarm mechanism is combined to realize local device linkage, and an encrypted control instruction is transmitted through a TLS / SSL encryption method to adjust the operation strategy of each smart home device, wherein the local device refers to all local smart home devices of a device type, including the following steps:

[0106] Step 1041, based on the lightweight federated learning framework, calling a reinforcement learning method, inputting the processing results of each device type in the global energy scheduling reference information into the reinforcement learning method model, and generating an initial energy scheduling strategy.

[0107] Step 1042, comparing the energy consumption expected value of each smart home device in the initial energy scheduling strategy with the preset normal energy consumption range, when the energy consumption expected value of any of the smart home devices exceeds the preset normal energy consumption range, triggering an alarm mechanism and marking the device type of the smart home device whose energy consumption expected value exceeds the preset normal energy consumption range.

[0108] Step 1043, based on the edge computing gateway, linking all local smart home devices of the marked device type, and collecting the current real-time operation parameters of the all local smart home devices.

[0109] Step 1044, in combination with the labeled device type and the corresponding real-time operation parameter, the initial energy scheduling strategy is adjusted through the cloud computing platform to form an optimized energy scheduling strategy, and the optimized energy scheduling strategy comprises operation adjustment instructions of the smart home devices.

[0110] Step 1045, the operation adjustment instructions are encrypted by a TLS / SSL encryption method to generate encrypted control instructions, and the encrypted control instructions are sent to an edge computing gateway through a preset transmission channel, so that the edge computing gateway decrypts the encrypted control instructions and generates decrypted control instructions after receiving the encrypted control instructions, and adjusts the operation strategy of the smart home devices according to the decrypted control instructions.

[0111] Among them, step 1045 can specifically include the following process: adding device unique identifier and timestamp to the operation adjustment instructions in the optimized energy scheduling strategy, and calling the TLS / SSL encryption method based on the lightweight federated learning framework to generate encrypted control instructions containing encrypted operation adjustment instructions, device unique identifier encrypted segment and timestamp encrypted segment; group the encrypted control instructions according to the device type label, and send the encrypted control instructions of the same group to the edge computing gateway through the preset transmission channel established based on the lightweight federated learning framework communication protocol, so that the edge computing gateway splits the encrypted control instructions into single encrypted control instructions according to the device type label after receiving the encrypted control instructions; decrypt the device unique identifier encrypted segment and the timestamp encrypted segment in the encrypted control instructions, and verify the matching of the device unique identifier and the local device, and the validity of the timestamp, and decrypt the operation adjustment instructions in the encrypted control instructions to generate decrypted control instructions when the device unique identifier matches and the timestamp is valid.

[0112] In the above scheme, the global energy scheduling reference information is the global energy management reference data formed by integrating the device type aggregation parameters, which contains the device type and the corresponding energy consumption control target; the reinforcement learning method is a machine learning method that continuously optimizes the strategy by interacting with the environment, and its model can generate a decision scheme based on input data; the initial energy scheduling strategy is a preliminary energy allocation scheme generated by the reinforcement learning model, which contains the energy consumption expectation value of each device; the energy consumption expectation value is the future energy consumption target of the device preset in the strategy; the TLS / SSL encryption method is a security protocol for network transmission encryption, which can protect the privacy in the data transmission process; the encrypted control instruction is the operation adjustment instruction encrypted by TLS / SSL, which contains encrypted instruction content, device identifier and timestamp; the decrypted control instruction is the executable instruction obtained by decrypting the encrypted control instruction; the timestamp is information recording the generation time of the instruction, which is used to verify the validity of the instruction;

[0113] In this application example, firstly, step 1041 calls the reinforcement learning model M based on the lightweight federated learning framework, inputting the processing results (i.e., A) of each device type in the global energy scheduling reference information R into M. M then generates an initial energy scheduling strategy based on A. , Among them The calculation formula is , where k is the strategy adjustment coefficient, for example, K is set to 1.1. For example, Refrigerator, 4.5 Air conditioning, 2.4 Refrigerator Rounded to one decimal place, 5.0, for air conditioners. Rounded to one decimal place, it is 2.6. Refrigerator, 5.0 Air conditioner, 2.6 .

[0114] Next, through step 1042, In Compared with the preset normal energy consumption range The comparison rule is: if or This will trigger an alarm. For example, the refrigerator's... ,That Trigger alarm and mark "Refrigerator", air conditioner ,That =2.6 is within the range, no alarm is triggered.

[0115] Then, through step 1043, local devices (such as refrigerators B1 and B2) under the tag type "refrigerator" are linked based on the edge computing gateway to collect data. : power ,temperature ; power ,temperature ,in Will be used for adjustment .

[0116] Then, through step 1044, the cloud computing platform combines the type of "refrigerator" and Adjustment ,generate Adjust the formula to ,in , Average power of the equipment, standard power =1.0kW. Calculated... =1.15, then the refrigerator's I: temperature to temperature to 7℃.

[0117] Finally, by step 1045, I adds ID (I “001”, “002”) and =“16:30:00”, call TLS / SSL encryption to generate , the encryption formula is , where K is the encryption key, after grouping according to the “refrigerator” type, sent to the gateway through the preset channel; after receiving, the gateway decrypts the ID and , verifies that “001” matches “002” matches and not expired, and then decrypts to obtain , which is used to adjust the refrigerator operation strategy.

[0118] The above 104 overall scheme generates an energy scheduling strategy through reinforcement learning, realizes intelligent allocation of global energy; combined with the alarm mechanism, it can discover energy consumption anomalies in time to ensure the safety of equipment operation; link local equipment to collect real-time parameters to provide accurate basis for strategy optimization; through TLS / SSL encrypted transmission of control instructions, it ensures that the instructions are not leaked or tampered with during transmission, protecting data privacy; the whole process realizes the whole process closed loop of energy scheduling strategy from generation, optimization to execution, which not only improves energy utilization efficiency, but also guarantees the privacy and security of real-time energy consumption data in the whole link.

[0119] The following is a complete example for steps S101-S104, such as Figure 2 ​​As shown, first in the smart home scene, the edge computing gateway is deployed in the user's home, responsible for collecting the real-time energy consumption big data of device B (refrigerator) and device C (air conditioner). The serial number S of device B is 2003, the factory number F is 2025, the serial number S of device C is 2004, and the factory number F is 2026; the gateway generates a dynamic mask tag for the device according to the formula L(t) = Rand(S, F, T, t), where the type T of device B is "refrigerator", and the tag L1 = "A3B7C" is generated at t = 10:00, and the type T of device C is "air conditioner", and the tag L2 = "D8E2G" is generated at t = 10:00, and the energy consumption data of device B is collected as [3.6, 3.8, 3.7] kWh, and the energy consumption data of device C is collected as [1.1, 1.3, 1.2] kWh, and after AES encryption, the encrypted data stream is formed, where the encrypted data stream of device B is "encrypted content: [A3B7C, 3.6, 3.8, 3.7]", and the encrypted data stream of device C is "encrypted content: [D8E2G, 1.1, 1.3, 1.2]".

[0120] Next, the edge computing gateway calls the Delta encoding compression algorithm to process the above encrypted data stream, calculates the continuous numerical difference value of the energy consumption data of device B, and can obtain Δ1 = 3.8-3.6 = 0.2, Δ2 = 3.7-3.8 = -0.1 according to the formula Δi = Ei-Ei-1, and the difference value is added to the position identifier =1、 =2, and the preset reference value B = 3.5 to form the marked difference value : 0.2 (B = 3.5), : -0.1 (B = 3.5)], and after replacing the original data, it is integrated into the data block "[A3B7C, P1:0.2, P2:-0.1]", and device C is processed in the same way to obtain difference values Δ1 = 1.3-1.1 = 0.2, Δ2 = 1.2-1.3 = -0.1, and form a data block "[D8E2G, P1:0.2, P2:-0.1]", and sort and combine according to the first letter of the device type to form a compressed encrypted data stream; based on the edge computing gateway rule (1 data block = 1 sub-data stream), it is divided into sub-data stream 1 (refrigerator data block) and sub-data stream 2 (air conditioner data block), and then converted according to the local processing rule (refrigerator: Δ × 20 + B; air conditioner: Δ × 30 + B, where B = 1.0), the intermediate parameters of device B are 0.2 × 20 + 3.5 = 7.5, -0.1 × 20 + 3.5 = 1.5, and the processing parameters of device C are 0.2 × 30 + 1.0 = 7.0, -0.1 × 30 + 1.0 = -2.0, and the processing parameters of device C are [refrigerator, [7.5, 1.5], and the processing parameters of device C are [air conditioner, [7.0, -2.0].

[0121] Then, after receiving the above processing parameters, the cloud computing platform calls homomorphic encryption technology to encrypt the intermediate parameters. The encryption formula is Ct,j=HE(Mj,K), where... Here, Mj represents the intermediate parameters to be encrypted, which are specific values ​​derived from the processing parameters of various smart home devices. K is the encryption key. The encryption intermediate parameters for device B are... , 1 = HE(7.5, K) , 2 = HE(1.5, K), and the value of device C is , 1 = HE(7.0, K) 2 = HE(-2.0, K), after adding type tags, grouped as follows ={ ,1, ,2}、 ={ ,1, ,2}; according to the aggregation formula The calculation, where At represents the aggregation result for a certain device type, is the overall energy scheduling reference value for that type of device obtained after homomorphic encryption, decryption, and calculation. For homomorphic decryption functions, For each set of parameters, the number of refrigerators in device B is n = 2. The number of air conditioners in device C is n = 2. The integrated energy dispatch reference information is R={[refrigerator, 4.5],[air conditioner, 2.5]}.

[0122] Finally, based on the global energy scheduling reference information R, the reinforcement learning model M generates the initial energy scheduling strategy. Among them, the expected energy consumption value =A×k (k=1.2), Device B =4.5 × 1.2 = 5.4 kWh / day, Equipment C =2.5 × 1.2 = 3.0 kWh / day; Compared to the preset normal energy consumption range, the normal range for device B is [3.0, 5.0]. Because 5.4 > 5.0, an alarm is triggered and the device is marked as a "refrigerator". The normal range for device C is [1.0, 3.5], and since 3.0 is within the range, no alarm is needed. The edge computing gateway links all refrigerator devices (device B) and collects real-time operating parameters: power 1.3kW, temperature 4℃. The cloud computing platform, combined with parameter adjustment strategies, adjusts the parameters using the following formula: = -( - )×0.2( The real-time power is 1.3kW. For the standard power 1.0kW), get =5.4-(1.3-1.0)×0.2=5.34≈5.3, the generated optimization strategy contains the operation adjustment instruction "device B temperature adjustment to 5℃"; add the device unique identifier ID="B001" and the timestamp ="15:30:00" to the instruction according to the formula =TLS / SSL(instruction+ID , ) encryption ( for transmission key), and the packet is sent to the gateway; the gateway decrypts and verifies that the ID matches and the timestamp is valid, obtains the decrypted instruction, adjusts the operation temperature of device B to 5℃, and completes the entire process.

[0123] Figure 3 is a structural schematic diagram of a specific embodiment of a cloud computing-based big data privacy computing system provided by the embodiment of the application, referring to Figure 3 , the system can include:

[0124] The acquisition module 31 is configured to acquire real-time energy consumption big data of each smart home device based on an edge computing gateway deployed on a smart home device side in a lightweight federated learning framework, and perform dynamic mask processing on device identifiers in the real-time energy consumption big data to generate encrypted data streams carrying device type labels and energy consumption data, so as to protect data privacy.

[0125] The compression module 32 is configured to process the encrypted data streams by using a Delta encoding compression algorithm to obtain compressed encrypted data streams, and perform local processing on the compressed encrypted data streams to obtain processing parameters of the smart home devices.

[0126] The aggregation module 33 is configured to aggregate the processing parameters by using a homomorphic encryption technology based on a cloud computing platform in the lightweight federated learning framework to form global energy scheduling reference information.

[0127] The encryption module 34 is configured to introduce a reinforcement learning mode to optimize an energy scheduling strategy based on the global energy scheduling reference information, combine an alarm mechanism to realize local device linkage, and transmit a control instruction by using a TLS / SSL encryption mode to adjust operation strategies of the smart home devices, so that the real-time energy consumption big data is always in a privacy protection state in the processing and transmission process.

[0128] The cloud computing-based big data privacy computing system according to an embodiment of the present application is used to implement the cloud computing-based big data privacy computing method described above, and the specific embodiments of the cloud computing-based big data privacy computing system can be found in the foregoing embodiment part of the cloud computing-based big data privacy computing method, and the specific embodiments can be referred to the description of the corresponding embodiment part, which will not be repeated here.

[0129] As shown in Figure 4 The present application further provides an electronic device, comprising: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the cloud computing-based big data privacy computing methods described above.

[0130] The present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of any of the cloud computing-based big data privacy computing methods described above.

[0131] In an exemplary embodiment, the computer readable storage medium described above can include, but is not limited to, a U disk, a read-only memory, a random access memory, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0132] The embodiments of the present application further provide a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps of any of the cloud computing-based big data privacy computing method embodiments described above.

[0133] The skilled person can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the foregoing description in a general manner. 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.

[0134] The above describes in detail the cloud computing-based big data privacy calculation method and system, the electronic device, and the storage medium provided by the present application. The principles and implementation modes of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, some improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the present application.

Claims

1. A cloud computing-based big data privacy computing method, characterized in that, The application comprises the following steps: Based on the edge computing gateway deployed on the smart home device side in the lightweight federated learning framework, the real-time energy consumption big data of each smart home device is collected, and the device identifier in the real-time energy consumption big data is dynamically masked to generate encrypted data stream carrying device type label and energy consumption data to protect data privacy; The encrypted data stream is processed by using the Delta encoding compression algorithm to obtain compressed encrypted data stream, and the compressed encrypted data stream is locally processed to obtain the processing parameters of each smart home device; Based on the cloud computing platform in the lightweight federated learning framework, the processing parameters are aggregated by using homomorphic encryption technology to form global energy scheduling reference information; Based on the global energy scheduling reference information, the energy scheduling strategy is optimized by introducing reinforcement learning mode, the local device linkage is realized in combination with alarm mechanism, and the operation strategy of each smart home device is adjusted by transmitting control instruction through TLS / SSL encryption mode; The step of introducing reinforcement learning mode to optimize the energy scheduling strategy based on the global energy scheduling reference information and realizing local device linkage in combination with alarm mechanism comprises the following steps: Based on the lightweight federated learning framework, the processing results of each device type in the global energy scheduling reference information are inputted to generate an initial energy scheduling strategy; The expected energy consumption value of each smart home device in the initial energy scheduling strategy is compared with the preset normal energy consumption range, and when the expected energy consumption value of any of the smart home devices exceeds the preset normal energy consumption range, the alarm mechanism is triggered and the device type of the smart home device whose expected energy consumption value exceeds the preset normal energy consumption range is marked; Based on the edge computing gateway, all local smart home devices under the marked device type are linked and the current real-time operation parameters of all local smart home devices are collected; In combination with the marked device type and the corresponding real-time operation parameters, the initial energy scheduling strategy is adjusted through the cloud computing platform to form an optimized energy scheduling strategy, and the optimized energy scheduling strategy contains operation adjustment instructions of each smart home device.

2. The method of claim 1, wherein, The step of adjusting the operation strategy of each smart home device by transmitting control instruction through TLS / SSL encryption mode comprises the following steps: The operation adjustment instructions are encrypted by using TLS / SSL encryption mode to generate encrypted control instructions, and the encrypted control instructions are sent to the edge computing gateway through a preset transmission channel, so that the edge computing gateway decrypts the encrypted control instructions after receiving the encrypted control instructions and generates decrypted control instructions, and adjusts the operation strategy of each smart home device according to the decrypted control instructions.

3. The method of claim 2, wherein, The step of encrypting the operation adjustment instructions by using TLS / SSL encryption mode to generate encrypted control instructions, and sending the encrypted control instructions to the edge computing gateway through a preset transmission channel, so that the edge computing gateway decrypts the encrypted control instructions after receiving the encrypted control instructions and generates decrypted control instructions, comprises the following steps: The running adjustment instruction in the optimized energy scheduling strategy is added with a device unique identifier and a timestamp, and is encrypted by calling a TLS / SSL encryption mode based on a lightweight federated learning framework to generate an encrypted control instruction containing the encrypted running adjustment instruction, the device unique identifier encrypted segment and the timestamp encrypted segment; The encrypted control instruction is grouped according to a device type label, and the encrypted control instruction of the same group is sequentially sent to an edge computing gateway through a preset transmission channel established based on a lightweight federated learning framework communication protocol, so that the edge computing gateway splits the encrypted control instruction into a single encrypted control instruction according to the device type label after receiving the encrypted control instruction; The device unique identifier encrypted segment and the timestamp encrypted segment in the encrypted control instruction are decrypted, and the matching of the device unique identifier with the local device and the validity of the timestamp are verified, and in the case that the device unique identifier matches and the timestamp is valid, the running adjustment instruction in the encrypted control instruction is decrypted to generate a decrypted control instruction.

4. The method of claim 1, wherein, The encrypted data stream is processed by using a Delta encoding compression algorithm to obtain a compressed encrypted data stream, and the compressed encrypted data stream is locally processed to obtain the processing parameters of the smart home devices, including the following steps: Real-time energy consumption big data corresponding to the device type label in the encrypted data stream is extracted, and difference values between consecutive values in the real-time energy consumption big data are calculated; Based on the difference values, a Delta encoding compression algorithm is called to form a compressed encrypted data stream, and the compressed encrypted data stream retains the association between the device type label and the corresponding real-time energy consumption big data; Based on a preset edge computing gateway rule, the compressed encrypted data stream is divided into a plurality of sub-data streams; The local processing rule of each smart home device is called, the difference values in the sub-data stream are converted into intermediate parameters conforming to a preset format based on the local processing rule, and the intermediate parameters are combined and encapsulated to form the processing parameters of the smart home devices.

5. The method of claim 4, wherein, Based on the difference values, a Delta encoding compression algorithm is called to form a compressed encrypted data stream, including the following steps: For each difference value, a corresponding numerical position identifier is added to the difference value, and the numerical position identifier is used to represent the order of the difference value in the consecutive values in the real-time energy consumption big data; The difference value added with the numerical position identifier is paired with a preset reference value, and for each difference value, the corresponding preset reference value is labeled as an association marker to form a marked difference value; Based on a lightweight federated learning framework, a Delta encoding compression algorithm is called to replace the corresponding consecutive values in the real-time energy consumption big data with the marked difference value one by one through the numerical position identifier to maintain the data order, and form a replaced data set, and the original correspondence between the device type label and the real-time energy consumption big data is retained during replacement; The replaced data set is classified according to the device type label, and the marked difference value and the corresponding numerical position under the same device type label are integrated into a data block, wherein each data block is associated with the corresponding device type label; All the data blocks are combined according to the preset sorting rule of the device type label to form a compressed and encrypted data stream, which maintains an association with real-time energy consumption big data through the numerical position identifier, associated mark and device type label in the data block.

6. The method of claim 1, wherein, The dynamic masking processing of the device identifier in the real-time energy consumption big data to generate an encrypted data stream carrying a device type label and energy consumption data includes the following steps: Based on the dynamic masking processing rule called by the lightweight federated learning framework, the serial number, factory number field and device type field used to distinguish different smart home devices in the device identifier are replaced with a time-varying random character sequence as a device type label; The device type label and the energy consumption data in the real-time energy consumption big data are combined to form an original data unit, and the original data unit is rearranged based on the lightweight federated learning framework to form a new data unit; According to the preset encryption conversion processing rule, the new data unit is encrypted to generate an encrypted data stream.

7. The method of claim 1, wherein, Based on the cloud computing platform in the lightweight federated learning framework, the processing parameters are aggregated by homomorphic encryption technology to form global energy scheduling reference information, including the following steps: Based on the lightweight federated learning framework, the cloud computing platform receives the processing parameters sent by each smart home device, and the processing parameters are the encapsulation results containing the device type label and the corresponding intermediate parameters; Call homomorphic encryption technology to encrypt and convert the intermediate parameters to form encrypted intermediate parameters, and add the corresponding device type label as additional identification for each encrypted intermediate parameter; The encrypted intermediate parameters corresponding to the same device type label are divided into the same data group, and the encrypted intermediate parameters in each data group are merged to obtain the type aggregation parameter corresponding to each data group; Integrate all the type aggregation parameters to form global energy scheduling reference information. 8.A cloud computing-based big data privacy computing system, characterized in that, It includes: The acquisition module is used for collecting real-time energy consumption big data of each smart home device based on the edge computing gateway deployed on the smart home device side in the lightweight federated learning framework, and dynamically masking the device identifier in the real-time energy consumption big data to generate an encrypted data stream carrying a device type label and energy consumption data to protect data privacy; The compression module is used to process the encrypted data stream by using the Delta encoding compression algorithm to obtain a compressed and encrypted data stream, and to locally process the compressed and encrypted data stream to obtain the processing parameters of each smart home device; The aggregation module is used to aggregate the processing parameters by homomorphic encryption technology based on the cloud computing platform in the lightweight federated learning framework to form global energy scheduling reference information; The encryption module is used to introduce a reinforcement learning mode to optimize the energy scheduling strategy based on the global energy scheduling reference information, realize local device linkage through an alarm mechanism, and transmit control instructions to adjust the operation strategy of the smart home devices through a TLS / SSL encryption mode. Based on the lightweight federated learning framework, the reinforcement learning mode is called, the processing results of each device type in the global energy scheduling reference information are input, and an initial energy scheduling strategy is generated. The expected energy consumption values of each smart home device in the initial energy scheduling strategy are compared with a preset normal energy consumption range. When the expected energy consumption value of any of the smart home devices exceeds the preset normal energy consumption range, an alarm mechanism is triggered, and the device type of the smart home device whose expected energy consumption value exceeds the preset normal energy consumption range is marked. Based on the edge computing gateway, all local smart home devices of the marked device type are linked, and the current real-time operation parameters of all the local smart home devices are collected. Based on the marked device type and the corresponding real-time operation parameters, the initial energy scheduling strategy is adjusted through the cloud computing platform to form an optimized energy scheduling strategy. The optimized energy scheduling strategy includes operation adjustment instructions of the smart home devices.

9. An electronic device, comprising: It comprises: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the cloud computing-based big data privacy computing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored in the computer readable storage medium, and the computer program is executed by the processor to implement the cloud computing-based big data privacy computing method according to any one of claims 1 to 7.

Images